ArticlePDF Available

Progress and direction in the use of stable isotopes to understand complex coral reef ecosystems: a review.

Authors:

Abstract and Figures

Coral reef ecosystems are exceptionally complex with a myriad of trophic pathways and consumer relationships. The application of stable isotopes (SIs) offers numerous advantages over traditional methods towards understanding these intricate systems. We summarize current knowledge derived from the rapidly increasing SI literature base and identify potential gaps and future directions for the use of SI in coral reef ecosystem studies. Using topic modelling, a form of text mining, on 236 identified published works, we determined that SI research on coral reefs broadly falls into five major topics. 1) Organic matter dynamics: SI analyses (SIA) have quantified substantial variability in autochthonous (internal) and allochthonous (external) fluxes across coral reefs. 2) Holobiont metabolism: Coral nutrient acquisition, translocation and partitioning, and coral responses to various endogenous and exogenous factors, have been explored through SIA. 3) Trophic niches: SIA has indicated that considerable variation in resource use facilitates co-occurrence of high densities of consumers, emphasising that many trophic categorisations on reefs are often too simplistic. 4) Fish diet variation and habitat connectivity: SIA has revealed how ontogenetic, larval, and mobile predator movements link adjacent ecosystems. 5) Environmental drivers (both natural and anthropogenic): SIA can track anthropogenic nutrient inputs, revealing impacts of human-derived pollutants on reef systems. There are a number of important knowledge gaps however. Few studies compare feeding strategies across guilds and the literature is biased towards reef fish and hard corals. Furthermore, few studies examine multiple taxonomic groups in situ or consider multiple environmental drivers. Studies also tend to ignore the underlying, but potentially substantial, spatiotemporal variation in SI baselines as demonstrated from 741 mean SI values extracted from the literature, making inferences based on small variations in SI values problematic. Given that coral reefs face global decline, knowledge gaps need to be addressed while acknowledging the limitations of SIA; careful application of SIs can enhance understanding of processes driving environmental change in these iconic marine ecosystems.
Content may be subject to copyright.
This is an author produced version of an Open Access paper accepted for publication
in Oceanography and Marine Biology: An Annual Review, Volume 60. This paper has
been peer-reviewed but does not include the final publisher proof-corrections or
journal pagination.
Citation for the accepted paper:
Skinner, C., Cobain, M.R.D., Zhu, Y., Wyatt, A.S.J., Polunin, N.V.C.
“Progress and direction in the use of stable isotopes to understand complex coral reef
ecosystems: a review.”
Oceanography and Marine Biology: An Annual Review, 2022, Volume 60. In press.
Published online with permission from: Routledge, Taylor & Francis Group
1
Progress and direction in the use of stable isotopes to understand complex coral reef
1
ecosystems: a review.
2
3
C. Skinner1*§, M. R. D. Cobain2,3*, Y. Zhu2,4, A. S. J. Wyatt1, N. V. C. Polunin2
4
1 = Department of Ocean Science and Hong Kong Branch of the Southern Marine Science and Engineering,
5
Guangdong Laboratory (Guangzhou), The Hong Kong University of Science and Technology, Kowloon, Hong
6
Kong.
7
2 = School of Natural and Environmental Sciences, Newcastle University, Newcastle upon Tyne, NE1 7RU,
8
UK.
9
3 = School of Natural Sciences, Trinity College Dublin, The University of Dublin, Dublin 2, Ireland.
10
4 = Institute of Marine Research, NO-5005 Bergen, Norway
11
* the first two authors contributed equally to this work.
12
§ = corresponding author = christina.skinner@live.com
13
Abstract
14
Coral reef ecosystems are exceptionally complex with a myriad of trophic pathways and consumer
15
relationships. The application of stable isotopes (SIs) offers numerous advantages over traditional methods
16
towards understanding these intricate systems. We summarize current knowledge derived from the rapidly
17
increasing SI literature base and identify potential gaps and future directions for the use of SI in coral reef
18
ecosystem studies. Using topic modelling, a form of text mining, on 236 identified published works, we
19
determined that SI research on coral reefs broadly falls into five major topics. 1) Organic matter dynamics:
20
SI analyses (SIA) have quantified substantial variability in autochthonous (internal) and allochthonous
21
(external) fluxes across coral reefs. 2) Holobiont metabolism: Coral nutrient acquisition, translocation and
22
partitioning, and coral responses to various endogenous and exogenous factors, have been explored
23
through SIA. 3) Trophic niches: SIA has indicated that considerable variation in resource use facilitates co-
24
occurrence of high densities of consumers, emphasising that many trophic categorisations on reefs are
25
2
often too simplistic. 4) Fish diet variation and habitat connectivity: SIA has revealed how ontogenetic,
26
larval, and mobile predator movements link adjacent ecosystems. 5) Environmental drivers (both natural
27
and anthropogenic): SIA can track anthropogenic nutrient inputs, revealing impacts of human-derived
28
pollutants on reef systems. There are a number of important knowledge gaps however. Few studies
29
compare feeding strategies across guilds and the literature is biased towards reef fish and hard corals.
30
Furthermore, few studies examine multiple taxonomic groups in situ or consider multiple environmental
31
drivers. Studies also tend to ignore the underlying, but potentially substantial, spatiotemporal variation in
32
SI baselines as demonstrated from 741 mean SI values extracted from the literature, making inferences
33
based on small variations in SI values problematic. Given that coral reefs face global decline, knowledge
34
gaps need to be addressed while acknowledging the limitations of SIA; careful application of SIs can
35
enhance understanding of processes driving environmental change in these iconic marine ecosystems.
36
3
CONTENTS
37
1. INTRODUCTION ........................................................................................................................................... 3
38
2. KEY PATTERNS AND TOPICS IN SI APPROACHES TO CORAL REEFS ................................................................ 8
39
3. ORGANIC MATTER DYNAMICS ................................................................................................................... 10
40
3.1. BACKGROUND................................................................................................................................. 10
41
3.2. INTERNAL AND EXTERNAL FLUXES OF DOM AND POM .......................................................................... 12
42
3.3. SPATIAL AND TEMPORAL VARIATIONS .................................................................................................. 14
43
3.4. ISOTOPIC INSIGHTS INTO THE ROLE OF DETRITUS ................................................................................... 16
44
4. HOLOBIONT METABOLISM......................................................................................................................... 17
45
4.1. BACKGROUND................................................................................................................................. 17
46
4.2. EXTERNAL NUTRIENT ACQUISITION ..................................................................................................... 18
47
4.3. INTERNAL NUTRIENT TRANSLOCATION AND PARTITIONING ...................................................................... 21
48
4.4. DRIVERS OF MIXOTROPHY ................................................................................................................. 23
49
4.4.1. Exogenous factors .............................................................................................................................. 23
50
4.4.2. Endogenous factors ............................................................................................................................ 24
51
4.5. METABOLISM IN NON-HARD CORAL SYMBIOSES .................................................................................. 26
52
5. TROPHIC NICHES ........................................................................................................................................ 28
53
5.1. BACKGROUND................................................................................................................................. 28
54
5.2. ISOTOPIC NICHES ............................................................................................................................. 29
55
5.3. DRIVERS OF ISOTOPIC NICHE VARIATION .............................................................................................. 35
56
6. FISH DIET VARIATION AND HABITAT CONNECTIVITY .................................................................................. 36
57
6.1. BACKGROUND................................................................................................................................. 36
58
6.2. ONTOGENETIC HABITAT SHIFTS .......................................................................................................... 37
59
6.3. BODY SIZE ...................................................................................................................................... 38
60
6.4. RESIDENCY AND POPULATION CONNECTIVITY ........................................................................................ 40
61
6.5. HABITAT CONNECTIVITY .................................................................................................................... 41
62
7. ENVIRONMENTAL DRIVERS ........................................................................................................................ 44
63
7.1. BACKGROUND................................................................................................................................. 44
64
7.2. NATURAL DRIVERS ........................................................................................................................... 45
65
7.3. ANTHROPOGENIC DRIVERS ................................................................................................................ 49
66
8. KNOWLEDGE GAPS, CAVEATS, LIMITATIONS, AND FUTURE DIRECTIONS ................................................... 53
67
8.1 KNOWLEDGE GAPS AND OPPORTUNITIES .............................................................................................. 53
68
8.2 CAVEATS AND LIMITATIONS ................................................................................................................ 57
69
9. CONCLUSIONS ........................................................................................................................................... 59
70
10. ACKNOWLEDGEMENTS ............................................................................................................................ 61
71
11. REFERENCES ............................................................................................................................................. 61
72
73
74
1. Introduction
75
4
Knowledge of the energy fluxes and networks of consumer relationships that constitute whole ecosystems
76
has long been considered crucial to understanding their structure (e.g. Lindeman 1942, Teal 1962),
77
functioning (e.g. Patten 1959, Slobodkin 1959, Ulanowicz 1972, Bellwood et al. 2019) and sustainable use
78
(e.g. Ryther 1969 , Friedland et al. 2012, Link and Watson 2019). This is particularly challenging where the
79
range of sources, diversity of consumers, and number of trophic levels are large (Link 2002). The high
80
species diversity of coral reefs therefore presents a huge obstacle to any attempt to understand the
81
ecosystem, especially as the ecologies of the majority of species are known only from short bursts of
82
research activity at particular locations and points in time (e.g. Odum and Odum 1955, Hiatt and Strasburg
83
1960, Randall 1967, Vivien 1973, Hobson 1974, Harmelin-Vivien 1981, Sano et al. 1984). This creates great
84
uncertainty in understanding coral reef functionality.
85
Ecosystem function relies on the movement and storage of energy and nutrients. As such, the exceptional
86
productivity and biodiversity of coral reef ecosystems in mostly oligotrophic tropical surface waters has
87
long been considered a paradox (Darwin 1842). The array of potential pathways sustaining these iconic
88
ecosystems are only now being explored in the detail that their present plight demands there is a
89
plethora of threats to coral reefs, including multiple consequences of climate change (Hughes et al. 2003,
90
Graham et al. 2011, Hoegh-Guldberg et al. 2017, Hughes et al. 2017) and impacts of fishery exploitation
91
(Newton et al. 2007, MacNeil et al. 2015, Bozec et al. 2016).
92
Coral reef primary production sources (all bold words appear in the glossary, Table 1) can include algae,
93
phytoplankton, various sponges and cnidarians which are partially photosynthetic or chemosynthetic, and
94
suspended and sedimentary particulate and dissolved organic materials derived from these. The nature of
95
these sources is complex, so characterising consumer relationships is non-trivial. For example, the
96
macroalgal matrix that nominally herbivorous fish feed on may also include microbes, detritus, and animal
97
material (Wilson and Bellwood 1997). Many herbivores may also feed opportunistically on the faeces of
98
zooplanktivorous fishes (Robertson 1982). Corallivores and spongivores, which include nominal herbivores
99
(e.g. Burkepile et al. 2019), actually ingest material of mixotrophic origin. This is because the coral or
100
sponge holobiont is composed of the host animal tissue but also symbiotic dinoflagellates (often from the
101
5
clade Symbiodiniaceae) and other potential prokaryotic symbionts, hereafter referred to as
102
“endosymbionts”. Marine symbioses such as these are particularly common in nutrient-poor environments
103
such as coral reefs (Ferrier-Pagès and Leal 2019) and present another layer of complexity in understanding
104
these ecosystems. Zooplankton have long been considered an important resource for reefs, but assessing
105
their role is complicated by the fact that pelagic plankton are continuously advected over reefs, in
106
conjunction with distinct reef plankton living amongst the reef substrata, with some only emerging at night
107
(Hobson 1974). Free-living bacteria also represent an important resource in coral reefs. They experience
108
high rates of growth and production through feeding on abundant dissolved organic matter, providing a
109
food source to higher trophic level consumers including zooplankton and corals, thereby transferring
110
energy and contributing to reef productivity (Sorokin 1973a, b, Sorokin et al. 1985, Ferrier-Pagès and
111
Gattuso 1998). While detritus constitutes an important flux (e.g. Crossman et al. 2005), the origins and
112
lability of the materials involved are diverse and little studied.
113
Even at this low level, understanding these diverse and complex relationships is hedged with uncertainty.
114
Traditionally, tools used to tackle these knowledge gaps have included in situ behavioural and stomach-
115
contents analyses, mass-balanced modelling, genetics, and “omics” methods (i.e. proteomics and
116
metabolomics). Feeding observations offer considerable resolution in some aspects of a consumer’s diet
117
but, without costly extension, provide data over only small temporal and spatial scales. They also generally
118
do not allow quantification of some hard to observe but significant dietary components such as microbes
119
and plankton, and they may provide only a modest measure of what is assimilated into consumer tissues.
120
Additionally, for interactions among marine symbioses, it is not possible to track nutrient exchanges
121
visually. For modelling approaches, the scope is vast but severely constrained by the accuracy of
122
parameterisation and key assumptions such as the rate of primary production and trophic transfer
123
efficiencies (e.g. Polovina 1984, Polunin and Klumpp 1990, Arias-González et al. 1997). While “omics”
124
methods are important for disentangling metabolic relationships in marine symbioses, they cannot trace
125
nutrient exchanges or identify original sources well (Ferrier-Pagès and Leal 2019).
126
6
Although coral reef ecosystem functioning has been considered since the early 1800s (Darwin 1842), stable
127
isotope (SI) approaches (see Text box 1) are a relatively recent addition to this endeavour, having only been
128
applied readily since the early 1980s (e.g. Fry et al. 1982), although there are some examples of earlier
129
works (Stephens 1960, Sorokin 1973a, Goreau 1977). These approaches fall into two categories: 1)
130
measuring natural abundances of stable isotopes in a sample, and 2) tracing the artificial addition of
131
heavier isotopes through a system of interest, known as isotope labelling. SI approaches have many
132
strengths (Text box 1; Fry 2006). SI compositions are time-integrated so potentially represent material
133
assimilation by consumers over time scales of days to months, whereas behavioural and gut content
134
analyses capture at most hours to days. SI ratios in a consumer’s tissues also contain information about
135
what has been assimilated from the diet, not merely ingested. This means the nutritional role of a
136
particular source can be clarified rather than having to be assumed. This extends to quantifying the
137
importance of materials such as gelatinous plankton and dissolved organic matter, which may be estimated
138
poorly, or not at all, via traditional gut contents analyses. This also extends to primary producers, with
139
tissue SIs relating to the nutrients taken up for fixation along with the fixation pathway itself. SI analyses
140
(SIA) offer ways of tracking distinct production sources and testing ideas about how the importance of
141
different pathways may vary in space and time, or in relation to other factors such as body size. SI ratios
142
can provide a chemical proxy of trophic position (TP) and a means of estimating trophic niche widths and
143
volumes, providing an opportunity to test ideas about overlaps in these parameters. Because of their
144
snapshot character, gut-contents analyses tend to require much larger sample sizes than those needed for
145
SIA, while SI data can be derived from non-lethal sampling. Inherent in each of these strengths are
146
important constraints, however. This includes those of the isotope ratio mass spectrometry technology
147
involved (e.g. Mill et al. 2008), the fact that SI ratios can only be used to track sources and trophic pathways
148
that are isotopically distinct, and that derived metrics such as TP rely on assumptions regarding isotopic
149
discrimination which can vary across space, time and trophic levels (Hussey et al. 2014).
150
There are still many opportunities to add value to SI studies, including through: 1) the use of elements
151
other than carbon and nitrogen (such as sulphur); 2) the use of fast and slow turnover tissues to represent
152
7
feeding over different time scales; 3) the use of archived materials to test ideas about past changes; 4)
153
improved parameterisation of niche topology (Jackson et al. 2011); 5) development of SI mixing models
154
allowing better differentiation of different production sources and trophic pathways (e.g. Parnell et al.
155
2013); 6) estimation of key processes such as food chain transfer efficiency from relationships between TP
156
and body size (Jennings et al. 2002); and 7) accurate characterization of enrichment factors between
157
consumers and their diet, especially for holobionts. The ability to gain SI information on specific
158
compounds rather than just bulk tissue homogenate has further expanded our ability to understand coral
159
reef ecosystems by better resolving different sources and pathways (e.g. McMahon et al. 2016, Fox et al.
160
2019, Skinner et al. 2021). While important for contextual understanding, the purpose of this review is not
161
to summarise or explain recent advances in SI methodologies and techniques (for that, we point readers to
162
Text box 1 and, among others, e.g. Boecklen et al. 2011, McMahon et al. 2013, Vander Zanden et al. 2015,
163
Pethybridge et al. 2018, Ferrier-Pagès and Leal 2019, McCormack et al. 2019, Whiteman et al. 2019, Shipley
164
and Matich 2020, Tsui et al. 2020). Instead, in light of the expanding but often disparate research occurring
165
on coral reefs that utilises SIs and the lack of a comprehensive synthesis of general findings, we sought to
166
summarise the current knowledge of coral reefs that has been advanced through the application of SI
167
approaches.
168
Undertaking a more systematic approach to guide this review, we conducted an extensive literature search
169
for studies utilising SI approaches in coral reef ecosystems and organisms. We screened and extracted key
170
information pertaining to each study including the focal taxa, when and where the study was conducted,
171
along with SI measurements of baseline sources. We then analysed the text in article abstracts using topic
172
modelling (a form of text mining which identifies clusters of words e.g. topics; Griffiths and Steyvers 2004,
173
Grün and Hornik 2011). This approach facilitates the objective identification of recurring themes in the
174
articles within the extensive literature (see Supplements S1 for detailed methods). Specifically, here we:
175
1) characterise the biological, geographical, and methodological foci of SI applications in published works,
176
2) quantify the isotopic variability observed among production sources across coral reef systems globally,
177
and 3) identify the principal topics to which SI data have been applied. We summarise the extent to which
178
8
current knowledge of coral reef ecosystems has been advanced through SIA, highlight the main areas for
179
growth, emphasise considerations that are essential when interpreting SI data from coral reef ecosystems,
180
and identify future challenges for obtaining fuller understanding of this complex system.
181
2. Key patterns and topics in SI approaches to coral reefs
182
By combining a traditional literature search with mining of pertinent information from identified articles,
183
and topic modelling of their abstracts, we were able to identify patterns and recurring themes in the SI
184
coral reef studies published to date (1980 2019 inclusive). Of 341 articles identified through
185
WebOfScience and 964 identified through GoogleScholar, a total of 238 were retained after careful
186
screening (Table S1). Of those, 236 had abstracts and were used in the subsequent topic modelling analysis
187
(see Supplements S1 for detailed methods). We also extracted key information from all 238 articles (Table
188
S1), including basic article information, study-specific information (e.g. spatial and temporal information;
189
Table S2 for definitions), sample information (e.g. single or multi-tissue, tissue type, isotopes, focal taxa etc;
190
see Table S3 for definitions), and, when reported, SI values of baseline sources (Table S6). It is worth noting
191
here that although we combined extensive literature searches across two different databases, the final 238
192
articles are likely not an exhaustive list of published SI studies on coral reefs. However, our literature search
193
generated an extensive, comprehensive, and representative suite of coral reef SI studies using an objective
194
approach.
195
Topic modelling identified five topics, which were named based on the highest weighted words occurring
196
within each (Table 2). Generally, similar numbers of articles were assigned to each topic, with Topic 2
197
(Holobiont metabolism) having the highest number of articles (n = 55) and Topic 1 (Organic matter
198
dynamics) the lowest number of articles (n = 41). Although topic popularity has fluctuated over time (Fig.
199
1), the distribution of topics across regions is fairly equitable (Fig. 2). These five topics were used as a guide
200
to construct the remainder of this review. The topics appear in a logical order which reflects the structure
201
of the coral reef ecosystem; starting from the basal production sources (Topic 1: Organic Matter
202
Dynamics), moving onto key primary producers and consumers (Topic 2: Holobiont Metabolism), food web
203
9
and community structure (Topic 3: Trophic Niches), higher-level consumers and the energetic linkages they
204
construct (Topic 4: Fish Diet Variation/Habitat Connectivity), and finally considering external drivers of
205
coral reef systems (Topic 5: Environmental Drivers) (Fig. 3). It should be acknowledged however that there
206
is some inherent cross-over between topics.
207
In terms of SI approaches used, natural abundances of SIs of bulk tissue (i.e. measuring a whole tissue type
208
such as muscle or liver) dominate the literature (84% of articles; Fig. 2B), with dual analysis of carbon and
209
nitrogen (i.e. δ13C and δ15N) the most popular approach (54% of articles; Fig 2B), followed by nitrogen on its
210
own (12% of articles; Fig. 2B). While SI studies using compound-specific stable isotope analysis (CSIA, i.e.
211
profiling individual compounds such as amino acids or fatty acids) are less frequent, they are increasing as
212
this methodology becomes more accessible (6%; Fig. 2B). Analyses on organism muscle tissue (n = 99),
213
holobiont host soft tissue (n = 79) and their endosymbionts (n = 46), and algal tissues (n = 66) constitute
214
the majority of studies, but there is a huge diversity of tissues that have been analysed. In addition to
215
various organs (liver n = 7; gonad n = 3; skin n = 3; digestive tract n = 2; heart n = 2), analyses have been
216
conducted on fish scales (n = 2) and gills (n = 2), faeces (n = 2), bone (n = 2), and sea cucumber respiratory
217
trees (n = 1). Most studies are field based (75%), with few relying on laboratory experiments (22%), and
218
fewer still combining both field and lab approaches (3%). While studies have been carried out around the
219
world, there is substantial disparity between regions (Fig. 2): the eastern Atlantic and eastern Pacific remain
220
poorly represented (n = 1 and n = 2, respectively), likely due to a lack of expansive coral reef systems in
221
these regions. Instead, western Atlantic studies dominate (28%), followed by the western Pacific (21%), and
222
the western Indian Ocean (20%), with notable foci localities in the Caribbean and Florida Keys, the Great
223
Barrier Reef, and the Red Sea, respectively.
224
Across the literature, a variety of sources have been sampled in the field, ranging from water nutrients and
225
organic matter through to macrophytes, zooplankton and holobionts, covering both the pelagic and
226
benthic and spanning the mangrove seagrass reef offshore habitat continuum (Table 3). Of the 741
227
extracted values, macroalgae were the most repeatedly measured for both δ13C and δ15N (n = 145 and 203,
228
respectively). POM, zooplankton, algal turfs, SOM, and coral (both homogenate and separated fractions)
229
10
have also been well characterised across studies (n > 45). The global variability observed in δ13C and δ15N
230
across coral reef sources is substantial (Fig. 4), spanning from approximately -30‰ to -4‰ for δ13C and -5
231
to 15‰ for δ15N, with macroalgae specifically exhibiting considerable diversity in expressed SI
232
compositions.
233
Sampling regimes of SI studies are dominated by limited temporal windows (single point, period, or season,
234
approximately 50% of screened articles), with only ~13% of studies designed to observe changes, if any,
235
between distinct seasons or years (multi-season, monthly, annually, and interannual studies) (Fig. 5A). Of
236
considerable concern is that 17% of studies screened failed to provide enough information to discern their
237
sampling regime, with examples found across all major regions of study. The focal taxa of studies are
238
heavily, but unsurprisingly, skewed towards reef fish and hard corals, which combined comprise over 60%
239
of all SI studies globally. This bias is particularly notable in the Western Indian region, whereas studies
240
based in the Western Atlantic appear to be more evenly dispersed across focal taxa groups (Fig. 5B). Barring
241
hard coral, there is limited focus on other coral reef sources (although they are often measured within
242
other studies, Table 3, and Fig. 4). Encouragingly however, studies that explore SIs in multiple components
243
of the ecosystem thereby employing a more holistic approach to understanding whole coral reef systems,
244
consist of over 5% of the observed literature.
245
3. Organic matter dynamics
246
3.1. Background
247
The dynamics of organic matter and remineralized constituents over and within coral reef ecosystems, and
248
hence the food and dissolved nutrient resources available to reef consumers, have long been enigmatic.
249
Odum and Odum (1955) suggested that the “changes in dissolved organic matter (DOM) in the vast flow of
250
water crossing [a] reef” posed a central question in reef productivity. Quantifying variability in material
251
fluxes into and within reefs was recognised in early studies as essential to understanding how reefs
252
function in oligotrophic oceanic settings (Sargent and Austin 1954, Odum and Odum 1955, Sorokin 1973b).
253
However, reef food-webs possess a large range of potential sources, both autochthonous (hereafter
254
11
internal”) and allochthonous (hereafter “external”), and these are likely to vary spatially along with
255
hydrodynamics and zonation of community structure (Fig. 3; Fig. 4; Table 3). SI studies investigating these
256
fluxes have predominantly been conducted in the western (n = 12) and central (n = 9) Pacific, western
257
Atlantic (n = 11), and western Indian Ocean (n = 10), with particular foci on Japan (n = 7), the Caribbean (n =
258
7), and the Red Sea (n = 5) respectively (Fig. 2). The eastern Pacific, eastern Indian, and eastern Atlantic are
259
poorly represented with few, if any, studies conducted (n = 0 of the papers assigned to this topic from our
260
literature search).
261
High reef productivity despite low ambient nutrient concentrations has been attributed to both internal
262
and external processes: high rates of internal recycling or atmospheric nitrogen fixation, and therefore
263
nutrient retention (e.g. Odum and Odum 1955, Johannes et al. 1972, Webb et al. 1975), versus flow-driven
264
inputs of external oceanic nutrients (Odum and Odum 1955, Andrews and Gentien 1982, Atkinson 1992,
265
Atkinson 2011, Wyatt et al. 2012a). The high rate of respiration, with net productivity often close to zero,
266
indicates effective recycling within reef systems (Kinsey 1985, Crossland et al. 1991, Tribble et al. 1994).
267
However, the fluxes of materials over reefs (both internal and external) and their uses by consumers have
268
rarely been examined at ecosystem-scales. This makes it difficult to make accurate predictions regarding
269
the future function of reefs subject to change. Indeed, non-isotope studies, mostly in mesocosms or
270
incubations, have indicated that DOM may be a significant resource (Tanaka et al. 2009, Nakajima et al.
271
2010, Naumann et al. 2010a, Tanaka et al. 2011a, Naumann et al. 2012). However, these are generally
272
hampered by a lack of understanding of hydrodynamic influences on DOM fluxes over natural reef systems
273
(but see Wild et al. 2012, Thibodeau et al. 2013). SI data offer great potential to identify organic matter
274
sources over natural reefs and measure the gross fluxes of this material, providing insights into reef
275
functioning that are essential for predictions for the future. To date, the focal taxa of studies investigating
276
these fluxes (n = 41) are typically reef fish (n = 13), other invertebrates (n = 6), and hard corals (n = 5). While
277
seawater is also frequently sampled (n = 9), there has been little focus on benthic algae (macroalgae n = 4;
278
turf algae n = 0) regarding organic matter dynamics, despite their generally large surface area, important
279
role in nutrient uptake and as a resource for many reef consumers (Fig. 5B).
280
12
3.2. Internal and external fluxes of DOM and POM
281
Organic matter in the ocean is dominated by dissolved organic matter (DOM), concentrations of which are
282
far greater on average (1 to 2 orders of magnitude) than particulate organic matter (POM) (e.g. Libes 2009,
283
Barrón and Duarte 2015). In contrast to the refractory nature and low availability of oceanic DOM to reef
284
consumers, much of the DOM produced by reef communities is labile and rapidly remineralized (e.g., one
285
month for ~80% of total organic carbon in coral mucus to be mineralized; Tanaka et al. 2011b). SI data have
286
demonstrated that corals can release significant amounts of DOM; background concentrations of DOC and
287
DON rose from 100 µM and 15 µM to 300 1700 and 120 µM, respectively, above coral colonies (Ferrier-
288
Pagès et al. 1998b), as well as POM (see below). Isotope labelling is often used to trace DOM fluxes; highly
289
isotopically enriched nutrients are supplied and subsequently traced through the various metabolic
290
pathways within the holobiont. 13C-labelling of reef-building corals Porites cylindrica and Acropora pulchra
291
suggested that the dissolved organic carbon (DOC) they release over reefs is derived from stored (> 90% of
292
total) rather than from newly synthesised organic C (< 10%) (Tanaka et al. 2008). One of the few ecosystem-
293
scale investigations of reef dissolved organic nitrogen (DON) cycling used spatial patterns in δ15N to
294
demonstrate that localised release of DON (potentially supported by N2 fixation in pristine benthic habitats)
295
is an important means of recycling N within reef communities (Thibodeau et al. 2013). The spatial
296
arrangement (see 3.3. Spatial and temporal variations below) of reef communities is therefore important in
297
determining rates of nutrient exchange across whole coral reef systems (Smith and Marsh 1973, Steven and
298
Atkinson 2003, Miyajima et al. 2007, Wyatt et al. 2012a, Wyatt et al. 2013).
299
While concentrations of DOM are greater, POM is more bioavailable (Lønborg et al. 2018), and therefore
300
POM dynamics are more readily studied on coral reefs. Several studies, including non-isotope approaches,
301
have identified significant inputs of external POM from the ocean by analysing gross fluxes of isolated
302
components of the external POM pool (particularly the smallest phyto- and bacterio-plankton) (Genin et al.
303
2009, Wyatt et al. 2010b, Patten et al. 2011, Akhand et al. 2021). While this may suggest a less prominent
304
role for tight recycling within reefs, rates of internal POM production have not been adequately quantified
305
across reef systems. Therefore, robust generalisations regarding the overall relative importance of external
306
13
versus internal POM inputs are lacking. Isotope labelling indicates that release of POM from corals may be
307
of a similar order or higher than that of DOM (average ratios of 0.6 and 0.5 for P. cylindrica and A. pulchra
308
respectively) (Tanaka et al. 2008). Macroalgae can also produce large amounts of POM; small incubation
309
chambers over intact reef habitats in Moorea (French Polynesia) have shown a dominance of algal material
310
within POM, with increased POM δ13C (−16.9 to -11.2‰) above ambient (− 20.6‰) indicative of the
311
contribution of algal exudates relative to 13C-depleted plankton (Haas et al. 2010). More enriched
312
nearshore δ13C values of POM samples (-18.3‰) in the Florida Keys were similarly taken to indicate the
313
degradation of seagrass detritus, while more depleted δ13C of the POM on the outer reef (-21.4‰)
314
suggested it was dominated by plankton (Lamb and Swart 2008), cf. Fig. 4. Few studies have quantified
315
rates of POM production over a natural reef. However, modelling of δ13C and δ15N data from Ningaloo Reef
316
in Western Australia, supported the premise that external POM inputs were balanced by correspondingly
317
high rates of gross internal POM release into water flowing over the reef crest and flat (Wyatt et al. 2013).
318
Measuring net concentration changes of, for example bulk POM, may therefore obscure the dynamics of
319
organic matter uptake and release over reefs, making SI evidence indispensable for disentangling internal
320
from external organic matter fluxes over reefs.
321
Organic mucus produced in abundance by corals may play a role in both internal cycling and external
322
inputs to reefs. Studies have used δ13C analyses to demonstrate that coral mucus can be both exported
323
from reefs but also re-enter reef food webs due to uptake by reef consumers (Naumann et al. 2010b, Wyatt
324
et al. 2013). Mucus itself can be a significant source of organic matter, especially in inner reef habitats with
325
decreased oceanic exposure (Wyatt et al. 2013). For example, it may enhance the flux of external particles
326
by trapping oceanic plankton in mucus aggregates, enhancing sedimentation rates onto reef habitats
327
(Naumann et al. 2009). Moreover, internal sources within reefs are not limited to the corals. For instance,
328
15N-labelling has shown significant rates of POM release from upside-down jellyfish Cassiopea sp. which can
329
be abundant in some reef systems and subsequently assimilated by zooplankton (Niggl et al. 2010).
330
Furthermore, non-isotope work has highlighted the important role of fish communities in supplying
331
(excretion) and storing (biomass) nutrients on coral reefs and adjacent habitats (Allgeier et al. 2014),
332
14
suggesting nutrient cycling is prevalent even at higher trophic levels. To date, there is limited SI work
333
exploring this aspect of organic matter dynamics on reefs (but see 6.5 Habitat connectivity).
334
Isotope labelling has been used to demonstrate a potentially major role for sponges in organic matter
335
uptake and retention within coral reefs. The combination of isotopic labelling (13C and 15N) and nanoscale
336
secondary ion mass spectrometry (NANOSIMS) has shown that sponge host tissues are capable of directly
337
utilising DOM in the water column through filter feeding (Achlatis et al. 2019). Indeed, some sponges are
338
more heterotrophic, as indicated by their bulk δ15N values (Fig. 4) (Weisz et al. 2007). For example, bulk
339
δ13C and δ15N indicate that coral cavity sponges typically feed on coral-derived materials (van Duyl et al.
340
2011, Slattery et al. 2013). Unlike other marine organisms, a few sponges show differential assimilation
341
mechanisms for different DOM sources, as revealed by stable isotope pulse‐chase experiments (13C and
342
15N); e.g. algal DOM is mainly used by symbiotic bacteria while coral DOM is assimilated by sponge cells (Rix
343
et al. 2017). de Goeij et al. (2008) used 13C-labelling to provide evidence that, in addition to POM, a coral
344
reef sponge, Halisarca caerulea, was able to incorporate DOM, thereby accessing an abundant source of
345
organic matter in reef waters. Later, using 13C- and 15N-labelled DOM, de Goeij et al. (2013) demonstrated
346
both in aquaria and in situ that DOM incorporation by sponges may facilitate DOM transfer to higher
347
trophic levels through detritus (POM) production. Using 13C- and 15N-labelling in the laboratory, DOM
348
uptake and POM production by sponges was further demonstrated to facilitate the transfer of coral-mucus-
349
derived DOM to detritivores on both warm and cold-water reefs (e.g. ophiuroids and polychaetes) (Rix et
350
al. 2016, Rix et al. 2018). These isotope studies on reef sponges have revealed an additional mechanism by
351
which organic matter produced on reefs may be internally retained within the system.
352
3.3. Spatial and temporal variations
353
Organic matter fluxes are likely to have strong spatial and temporal dynamics (Fig. 4). These will be
354
influenced by the hydrodynamic conditions and zonation of benthic communities over small spatial scales
355
of hundreds of metres or less (Haas et al. 2011, Kolasinski et al. 2011, Wyatt et al. 2012a), and larger-scale
356
stochastic events such as upwelling, storm disturbances, and coral bleaching events (Leichter et al. 2007,
357
15
Wild et al. 2008, Kolasinski et al. 2011, Radice et al. 2021). Further, large-scale urbanisation-eutrophication
358
gradients can be reflected through reef food webs, particularly as elevated δ15N (Kürten et al. 2014, Duprey
359
et al. 2020; and see 7. Environmental Drivers). Broad scale SI data or ‘isoscapes’, i.e. a spatial pattern of
360
isotopic values across a land or seascape (West et al. 2008, Bowen 2010), can reveal geographic patterns in
361
organic matter and nutrient dynamics over reefs. For instance, in the Red Sea, where 11% of all coral reef SI
362
studies have been conducted (Fig. 2), low organic matter δ15N values in the north (zooplankton δ15N: 1.3‰)
363
reflect the importance of N2 fixation, while the higher δ15N of organic matter in the south (zooplankton
364
δ15N: 5.8‰) reflects N inputs from the Indian Ocean (Kürten et al. 2014). Spatial and temporal variability of
365
organic matter over reefs is poorly understood, yet there is strong SI evidence of habitat-linked variations
366
in fluxes of DOM (Thibodeau et al. 2013), POM (Wyatt et al. 2013), and dissolved nutrients (Leichter et al.
367
2007). While these oceanographic drivers mean that external resource supply can be highly variable in
368
space and time (Wyatt et al. 2012a, Wyatt et al. 2013, Kürten et al. 2014), most SI reef studies are based on
369
just a single time point or period of sampling (44%), with few conducting repeat samplings across multiple
370
time points (19%) (Fig. 5A).
371
Mass coral spawning can periodically increase POM concentrations several fold on many reefs and SI data
372
have proven useful in examining the biogeochemical impact of these events. A lasting influence of
373
spawning on POM δ15N was seen on the Great Barrier Reef over a period of about 10 days (increases of ~ 5
374
, from 0.7 to 4.8-5.7 ), suggesting spawning-derived organic matter is rapidly transferred to higher
375
trophic levels (Wild et al. 2008). Similarly, in Kane’ohe Bay, Hawai’i, POM δ15N remained slightly elevated
376
(2-3 ‰) after spawning (> 10 days), indicating incorporation of spawn-derived organic matter into higher
377
trophic levels, but spawning impacts on water-column POM δ13C were short-lived (2-4 days post-spawning
378
returned to pre-spawning values) (Briggs et al. 2013). Briggs et al. (2013) also found that the δ13C of tissues
379
of the coral Montipora capitata increased by ~ 1 ‰ (-13.3 to -12.2 ‰) over the course of the spawning
380
season, with eggs having lower δ13C than host tissues (-14.5 ‰). This temporal change could reflect the
381
spawning physiology; the concentration of 13C-light wax esters and overall carbon content is higher in M.
382
capitata eggs compared to adults (Padilla-Gamiño et al. 2013), but it may also reflect changes in feeding or
383
16
relative rates of autotrophy and heterotrophy. This underscores the importance of having a good
384
understanding of baseline isotope variations when assessing the impacts of stochastic events like spawning
385
using tissue SI data.
386
The high spatiotemporal variability in resources means that adequately quantifying the isotopic
387
composition of material supplied to reefs can be a non-trivial undertaking (Fig. 4; Table 3; Table S6). For
388
instance, variability in the isotopic composition of dissolved inorganic N (e.g. nitrate) can be high across the
389
water column (3.5 to 5.5 between 50 242 m) just due to small-scale patchiness in N cycling (Leichter et
390
al. 2007). Internal production can also be highly variable, with organic carbon released by benthic algae on
391
reefs at Moorea demonstrating distinct seasonality, including δ13C variation of ~ 5 ‰ (Haas et al. 2010).
392
This variability is likely why few studies to date have used isoscapes on coral reefs, or tropical waters in
393
general (but see MacKenzie et al. 2019); those in marine settings have mostly been conducted in temperate
394
shelf regions (MacKenzie et al. 2014, Kurle and McWhorter 2017, St. John Glew et al. 2019). However,
395
isoscapes can provide spatial (and potentially temporal) context to more general questions about
396
community ecology, animal migration, and nutrient cycling (Hobson et al. 2010, McMahon et al. 2013,
397
Cheesman and Cernusak 2016). In one of the few studies employing an isoscape approach on coral reefs,
398
δ15N values of long-lived benthic bivalves were used to create a nitrogen isoscape in New Caledonia that
399
highlighted regions of eutrophication 15N of 11.7 compared to 4.3 ‰ in lagoon), characterising the
400
anthropogenic nitrogen footprint of the area (Thibault et al. 2020). Thus, isoscapes of coral reef ecosystems
401
that account for sufficient natural variation could offer insight into the complex processes that might
402
influence coral reef trophodynamics and SI values across various locations (Fig. 3; Fig. 4; Table 3). Given
403
that environmental conditions are fluctuating due to climate change, isoscape studies would further benefit
404
from the inclusion of SI data across both space and time where possible (McMahon et al. 2013).
405
3.4. Isotopic insights into the role of detritus
406
Detritus (i.e. non-living organic matter) is an abundant and potentially significant food resource over reefs,
407
but its variable lability makes it challenging to characterise (Fig. 4; Table 3; Table S6). Isotopes offer an
408
17
opportunity to better trace and quantify fluxes of this material. Indeed, there is substantial isotopic
409
evidence of a prominent role for detritus in reef food webs. While detrital pools are higher (1.6 times) in
410
low energy back reef habitats where benthic material accumulates (evidenced by enriched mean δ13C -
411
16.83 ‰), input rates are higher (1.7 2.9 times) over dynamic forereefs due to oceanic detritus supply
412
(evidenced by depleted mean δ13C - 19.84 ‰) (Max et al. 2013). Thus while detritus can be a nutritious
413
food source, its availability fluctuates with hydrodynamics (Max et al. 2013). Isotope labelling (13C and 15N)
414
has demonstrated that detritivores, such as ophiuroid brittle stars, can play an important role in the
415
recycling of nutritionally poor detritus to higher trophic levels (Rix et al. 2018). There is also increasing
416
evidence of a high degree of detritivory in reef fishes previously assumed to be largely herbivorous. For
417
instance, parrotfish like Chlorurus sordidus are often classified as herbivores. Stomach and feeding
418
observations have suggested that this species may be better described as a detritivore (Choat et al. 2002)
419
but bulk and CSIA data indicate it might be principally detritivorous across both oceanic and inshore reefs
420
(Wyatt et al. 2012b, McMahon et al. 2016). In agreement with observational and morphological studies
421
(Choat et al. 2002, Choat et al. 2004), CSIA data identified that the surgeonfish Ctenochaetus striatus is also
422
predominantly detritivorous (73%; McMahon et al. 2016). Dietary plasticity at the individual level
423
(McMahon et al. 2016), or spatial changes in the importance of detritivory (Wyatt et al. 2012b), require
424
more detailed investigation. However, it appears likely that the importance of detritus, and therefore
425
microbial reworking, in reef food webs has been underestimated (McMahon et al. 2016). As described in
426
detail in 3.2 Internal and external fluxes of DOM and POM, sponges also play a key role in reworking
427
detritus on coral reefs through the sponge loop: coral mucus carbon and nitrogen is transferred into
428
sponge tissues and subsequently released as detritus (de Goeij et al. 2013, Rix et al. 2016).
429
4. Holobiont metabolism
430
4.1. Background
431
Hermatypic (reef-building) scleractinian corals play a foundational role providing essential habitat for reef
432
organisms (Coker et al. 2014), so it is only natural that considerable efforts have focussed on their ecology
433
18
and the metabolic dependency between them and their symbiotic dinoflagellates in the family
434
Symbiodiniaceae (Fig. 3). Holobiont metabolism had the largest number of studies principally assigned to it
435
(n = 55) among SI coral reef topics. Likewise, of all 238 identified SI studies, 26% (n = 62) have had coral as
436
their focal taxa, second only to fishes (36%, n = 85) (Fig. 5B). Studies focusing on this topic contributed a
437
major proportion of the earliest SI works on coral reefs (Fig. 1). As SIs act as natural (or artificial) tracers for
438
elucidating energetic pathways, they are ideally placed to resolve questions regarding nutrient acquisition,
439
translocation, and utilisation within and between holobionts and their symbiotic microbiomes (Fig. 3). This
440
is highlighted by the fact that most isotope labelling studies were assigned to the topic of Holobiont
441
metabolism (28 of 37 total). Geographically, there has been a particular focus on holobiont metabolism
442
regionally in the western Indian Ocean (notably experimental work in the Red Sea) and the western Pacific
443
(Fig. 2B; 5B). Key aspects revolve around the nature of nutrients that are taken up by holobionts; how they
444
are proportioned internally between hosts and symbionts; and drivers of variation in the relative utilisation
445
of photo- or chemosynthetically fixed material (autotrophy) and secondarily sourced production
446
(heterotrophy), known as mixotrophy. Here we expand upon these ideas to demonstrate the
447
understanding achieved using SIs, before considering studies whose focus is on the symbiotic relationships
448
found in organisms other than those of hermatypic corals.
449
4.2. External nutrient acquisition
450
As highlighted in the previous section, there is a wealth of external nutrient sources that are potentially
451
available to corals and other holobionts to underpin their metabolism, as well as those of their
452
endosymbionts to form photosynthates. This is despite coral reef ecosystems traditionally being viewed as
453
restricted to oligotrophic tropical surface waters (Darwin 1842). Early work by Risk et al. (1994) suggested
454
increasing reliance on terrigenous carbon sources by corals and their endosymbionts after observing
455
increasing δ13C tissue values (from -16 to -11 ‰) with increasing distance from shore in the Great Barrier
456
Reef. Spatially varying nutrient acquisition has also been observed with reef zonation, with varying coral
457
tissue δ15N values along an offshore to mid-shelf to inshore barrier reef transect attributed to cold-water
458
upwellings, algae-based nitrogen fixation and terrigenous sources respectively (Sammarco et al. 1999, Erler
459
19
et al. 2014). Elsewhere, across a reef atoll system in the Maldives, nitrates from deep ocean water that are
460
enriched in 15N have been shown to be incorporated into coral tissues via seasonal upwelling (Radice et al.
461
2019). In the Cook Islands in the South Pacific Subtropical Gyre, nitrogen inputs to corals also vary
462
seasonally, with δ15N in coral skeletons suggesting nitrogen sources originate from groundwater during wet
463
seasons and from N2 fixation during dry seasons (Erler et al. 2019). These studies highlight the
464
environmental and temporal context dependency of patterns in coral nutrient uptake, but it is worth noting
465
that increases in coral and endosymbiont δ13C values may also be related to spatial gradients in, for
466
example, turbidity affecting coral photosynthesis rates or endosymbiont genera.
467
Direct nutrient uptake from the water column is likely influenced by its form; nitrogen is primarily available
468
as nitrate or ammonium, where ammonium appears to be principally taken up by the endosymbiotic algae.
469
Eight week-long 15N-labelled ammonium enrichment experiments revealed a 10-fold increase in ammonium
470
uptake rates by hosted dinoflagellates compared to coral tissue (Grover et al. 2002). However, overall
471
ammonium uptake is reduced in fed versus starved hosts (Grover et al. 2002), which may suggest
472
secondary reliance on ammonium as a nitrogen source compared to heterotrophic feeding. Similar
473
patterns were found for nitrate using 15N-labelled nitrate and ammonium; nitrate is principally taken up by
474
the endosymbionts, with uptake rate independent of prior nutrient acclimatisation, but significantly lower
475
under high ammonium regimes (Grover et al. 2003). The metabolic response of dinoflagellate
476
endosymbionts to external nutrient enrichment is rapid; significant uptake occurs within an hour of
477
exposure to either ammonium or nitrate 15N-enriched seawater (Pernice et al. 2012, Kopp et al. 2013). This
478
suggests direct uptake and fixation of ammonium by endosymbionts from seawater filling the coelenteron
479
rather than nutrient transfer from hosts to algae (Pernice et al. 2012, Kopp et al. 2013). Such a rapid
480
response is facilitated by temporary nitrogen storage in intracellular uric acid crystals that can then be
481
remobilised in the following hours (Kopp et al. 2013). Similar temporary intracellular storage structures
482
have also been identified for carbon using pulse chase isotopic labelling of 13C-bicarbonate, whereby
483
seawater bicarbonate is rapidly (~ 15 minutes) taken up by dinoflagellates and fixed into lipid droplets and
484
starch granules (Kopp et al. 2015).
485
20
The uptake of water column nutrients can be further facilitated by recycling pathways, as has been
486
revealed by isotope tracer experiments using artificially 15N-labelled bacteria to track their incorporation
487
into coral larvae (Ceh et al. 2013). Isotopic labelling (15N) has further shown that nitrogen-fixing
488
diazotrophic bacteria may also be taken up directly into the epidermal layers of coral larvae (Lema et al.
489
2016), potentially helping to meet their nitrogen demands. Nitrogen fixation can also occur within the
490
endolithic diazotrophs found between coral tissues and their carbonate skeleton, as demonstrated with
491
15N-labelling (Grover et al. 2014, Yang et al. 2019). However, the importance of this pathway varies
492
according to coral metabolic status (i.e. whether they are more auto- or hetero-trophic) and with depth
493
(Bednarz et al. 2017). Bulk SI work 13C and δ15N) has however shown a lack of nutrient exchange between
494
coral polyps and adjoining epilithic algal turfs, demonstrating that at least some potential nutrient
495
pathways are not utilised (Titlyanov et al. 2008).
496
Heterotrophic feeding by coral polyps further expands the potential nutrient sources available for these
497
holobionts, with a diversity of trophic interactions being observed (see Houlbrèque and Ferrier-Pagès 2009
498
for a detailed review on coral heterotrophy). Lai et al. (2013) used 15N-labelled seagrass to experimentally
499
show that corals can directly consume seagrass material, both in particulate and dissolved form. This
500
highlights a direct nutritional link between corals and adjacent seagrass meadows, which can export large
501
quantities of fresh and detrital material. Other experimental bulk SI work 13C and δ18O) has highlighted
502
how grazing on zooplankton can lead to substantial increases in various measures of coral fitness, such as
503
tissue chlorophyll concentrations, compared to when heterotrophic feeding is restricted (e.g. Reynaud et
504
al. 2002). Furthermore, grazing experiments using 3H-labelled bacteria and ciliates showed that although
505
coral grazing rates decrease as light intensity increases, heterotrophy still contributes to coral nutrition,
506
suggesting it complements autotrophy even under high light conditions in Stylophora sp. (Ferrier-Pagès et
507
al. 1998a). More recent SI finger-printing techniques using δ13C of essential amino acids (CSIA)
508
demonstrates the significant contribution of heterotrophic feeding to coral hosts, with an average of 41%
509
contribution to assimilated material determined for the widespread Indo-Pacific scleractinian coral
510
Pocillopora meandrina (Fox et al. 2019). Similar proportions of heterotrophic nutrient uptake have also
511
21
been verified experimentally using 15N-labelled rotifers to quantify nitrogen incorporation rates in another
512
common Indo-Pacific coral Porites lutea (Rangel et al. 2019). However, the degree of heterotrophic feeding
513
may be a species-specific trait; paired bulk 13C and δ15N) and amino acid CSIA 13CAA, δ15NAA) revealed
514
that Montipora capitata did not change their nutritional strategies under different experimental nutrition
515
regimes (Wall et al. 2021), suggesting a lack of trophic plasticity.
516
4.3. Internal nutrient translocation and partitioning
517
The fate of nutrients once assimilated into host tissues underpins the symbiotic relationship between coral
518
and Symbiodiniaceae basal products need to be supplied to algal cells for autotrophy and likewise
519
photosynthates shuttled elsewhere for energetic demands of the host (Fig. 3). Isotopic labelling
520
experiments tracing the various metabolic pathways within the holobiont constitute some of the earliest
521
uses of SIs on coral reefs. For example, feeding experiments with 14C revealed that six species of coral were
522
able to feed on DOM and planktonic bacteria, actively consuming organic phosphorus bound in the cells of
523
the latter (Sorokin 1973a). Other studies have attempted to understand carbon translocation and turnover
524
within coral colonies. Crossland et al. (1980) saw a 50-60% loss of photosynthetically fixed 14C from
525
isotopically labelled and replanted coral colonies during their first 40h back on the reef. However, rather
526
than the coral colonies translocating fixed carbon from outer branches to basal regions, the authors
527
suggested that the coral tissues had released mucus and dissolved organic carbon into the environment
528
(Crossland et al. 1980). Similarly, Rinkevich and Loya (1983) used 14C sodium bicarbonate in the field to
529
demonstrate limited translocation of photosynthetic products across coral colonies over the course of a
530
month (from bases to branch tips), despite lower incorporation rates in the growing tip regions.
531
Under more controlled laboratory settings, high precision tracing can be conducted to elucidate internal
532
translocation of metabolites over short timescales, further revealing the complexities of this symbiotic
533
relationship. Following 13C-labelled bicarbonate enriched seawater incubation, Tremblay et al. (2012)
534
calculated that 60% of carbon fixed by endosymbionts in the scleractinian Stylophora pistillata is
535
translocated to host tissue within 15 minutes, with approximately 50% of fixed carbon being respired by
536
22
the holobiont as a whole. Whole nitrogen budgets have also been determined through 15N enrichment,
537
showing that a majority (50-83%) of nitrogen utilized by endosymbionts is derived from coral hosts, with
538
host species-level differences attributed to different N-biomasses per unit surface area of coral host species
539
(Tanaka et al. 2015, Tanaka et al. 2018). One of the few studies to combine both δ13CAA and δ15NAA to
540
explore the nutritional exchanges between coral hosts and their endosymbionts confirmed that
541
endosymbionts do benefit from host heterotrophy (Ferrier-Pagès et al. 2021). Furthermore, this
542
relationship is not one way but tightly coupled. Combining 13C- and 15N-labelling showed that the coral host
543
derived 99% of its total nitrogen from the endosymbiont, suggesting the host “farms” the endosymbionts
544
to efficiently exploit both C and N (Tanaka et al. 2018).
545
Interestingly, the fate of acquired nutrients appears to depend upon the source of the material. Isotope
546
labelling (13C and 15N) reveals that heterotrophic sources exhibit considerable internal exchange and
547
retention within the coral-algae symbiosis, whereas inorganic nutrients (that are photosynthetically fixed
548
by endosymbionts) are rapidly used and respired to meet more immediate metabolic demands (Hughes et
549
al. 2010, Tremblay et al. 2015). The release of DOM by corals can constitute a considerable loss of fixed
550
material, corresponding to approximately 5% of net photosynthetic production of endosymbionts as shown
551
by 13C tracer accumulation experiments (Tanaka et al. 2009). This has been corroborated elsewhere; similar
552
work has estimated coral carbon losses due to DOM release is equivalent of 28% of gross carbon fixation
553
(Tremblay et al. 2012).
554
With the advent of ever more-sophisticated technology, exploring the spatial structuring of metabolic
555
processes within coral tissues is now possible. Notably, the combination of isotopic labelling and nanoscale
556
secondary ion mass spectrometry (i.e. NANOSIMS) has revealed the variation in net carbon fixation rates
557
between individual cells of Symbiodiniaceae, with an average sixfold decrease between upper and lower
558
tissue layers within individual polyps (Wangpraseurt et al. 2016). These combined technologies have also
559
helped disentangle the rapid nutrient uptake dynamics within endosymbionts (Kopp et al. 2013, Kopp et al.
560
2015). Such approaches can be further expanded with, e.g. simultaneous immunofluorescent microscopy to
561
correlate presence of isotopically-labelled labile nutrients with associated proteins and enzymes to further
562
23
elucidate metabolic pathways (Loussert-Fonta et al. 2020). These technological advances have the potential
563
to greatly expand current understanding of the molecular-level underpinning of coral-algae symbioses.
564
4.4. Drivers of mixotrophy
565
The mechanisms that influence the strength of coral-algae symbiosis are key to the wider coral reef
566
ecosystem. Therefore, understanding how coral mixotrophy changes with different factors provides insight
567
into how they may be best managed and conserved in a changing world. SI studies that explore potential
568
drivers of mixotrophy in corals can generally be categorised by the nature of the driver(s) of interest:
569
whether mixotrophy is impacted by the external environment (exogenous) or influenced by traits that are
570
particular to the studied host or symbiont (endogenous).
571
4.4.1. Exogenous factors
572
Given the photosynthetic underpinning of coral-algae symbiosis, depth represents a strong natural abiotic
573
gradient which can impact autotrophic efficiency due to, among other things, diminishing ambient light
574
levels (see 7.2 Environmental Drivers: Natural Drivers). The nutritional history of host corals (fed versus
575
starved), which is driven by the temporal dynamics of prey availability, may also play a role in mixotrophy.
576
In isolation, 15N labelling indicates that nutritional history does not appear to affect the assimilation
577
efficiency of heterotrophic feeding by corals (Piniak and Lipschultz 2004). However, δ13C of fatty acids
578
reveals that recent starvation, when compounded with thermal stress, leads to reductions in chlorophyll
579
and maximal photosynthetic efficiency in coral tissues, resulting in respiration of storage fatty acids in order
580
to maintain coral metabolism (Tolosa et al. 2011). This highlights the compounding nature of
581
environmental factors (see 7.2 Environmental Drivers: Natural Drivers).
582
Bleaching events, which are induced by a variety of external stresses, are projected to continue to increase
583
with ongoing climate change (Hughes et al. 2017). Given the loss of the autotrophic symbionts during such
584
episodes, the impact on host energetics is likely to be significant. Early SI work suggested however that
585
bleaching does not alter the ratio of heterotrophic vs autotrophic dependency. This was inferred from
586
similar δ13C values in two corals (Porites compressa and Montipora verrucosa) and their endosymbionts
587
24
taken from paired samples of bleached and non-bleached tissue in the field in Hawaii (Grottoli et al. 2004).
588
This was in contrast to pulse-chase experiments (13C) on the same species (P. compressa) and a congener
589
(Montipora verrucosa cf. M. capitata) that showed reduced assimilation of autotrophically derived carbon
590
in bleached corals, but with assimilation of heterotrophic sources remaining similar in bleached versus non-
591
bleached corals (Hughes et al. 2010). More recent work using bulk SI data (δ13C and δ15N) on the same two
592
species also did not indicate increased heterotrophic nutrition post-bleaching (Wall et al. 2019). The δ13C
593
data also suggest corals may undergo biomass compositional changes when bleached and directly after
594
(Wall et al. 2019), possibly as they catabolize their stores of 13C-enriched lipids (Grottoli and Rodrigues
595
2011). Heterotrophic carbon sources are likely important for coral recovery as they are predominantly used
596
to replenish coral lipids (Baumann et al. 2014), which are used to maintain metabolism during thermal
597
stresses (Tolosa et al. 2011). For longer-term responses of corals to bleaching, see 7.2 Environmental
598
Drivers: Natural Drivers.
599
4.4.2. Endogenous factors
600
The diversity in form observed across Scleractinia, and even within species across for example depth
601
gradients (Einbinder et al. 2009), suggests ecological trade-offs associated with different host
602
morphologies. Previous non-SI work exploring coral morphology-feeding relationships suggested that coral
603
surface to volume ratio (S/V) and polyp size might determine the importance of light or zooplankton
604
capture. Branching corals have maximum S/V ratios and small polyps so are best suited for light capture,
605
while corals with lower S/V ratios generally have larger polyps suiting them to zooplankton capture (Porter
606
1974, Porter 1976). Subsequent non-SI work by Wellington (1982) corroborated that tentacle size was
607
important for determining the degree of hetero- or autotrophy (i.e. corals with larger tentacles use more
608
zooplankton), but could not confirm the coral morphology-feeding relationships. Some SI data have now
609
supported this hypothesis. δ13C values in coral tissues and endosymbionts from 14 Red Sea coral species
610
indicated increased relative rates of autotrophy in branching corals with smaller polyps compared to
611
massive species with larger polyps, attributed to reductions in carbon-limitation associated with increasing
612
surface area (Levy et al. 2006). Such a trend with host morphology has been further corroborated by Xu et
613
25
al. (2020) using δ13C. However, in contrast, a non-isotope feeding study found no relationship between
614
coral feeding rates and polyp size (Palardy et al. 2005), while Hoogenboom et al. (2015) using bulk δ13C and
615
δ15N found that feeding rates were instead highest in branching corals with smaller polyp sizes. This
616
suggests more research is needed to disentangle coral morphology-feeding relationships. More recently,
617
corals that were more autotrophic, implied by similarity in host tissue and symbiont bulk δ13C and δ15N,
618
had a negative relationship with polyp size but also bleaching resistance, suggesting they may be more
619
susceptible to increasing water temperatures (Conti-Jerpe et al. 2020).
620
While less conspicuous, the different genera within Symbiodiniaceae that form symbioses with corals can
621
also influence host metabolism due to their differing functional responses to environmental conditions. For
622
example, isotope labelling (13C and 15N) revealed that holobionts hosting former clade C (genus
623
Cladocopium) have increased uptake rates of inorganic nitrogen under non-thermal stress conditions
624
compared to holobionts with former clade D (genus Durusdinium) (Baker et al. 2013a). This can lead to the
625
competitive exclusion of clade D under normal conditions despite this clade resulting in increased carbon
626
acquisition during periods of thermal stress (Baker et al. 2013a). Similarly, Ezzat et al. (2017) demonstrated
627
increased carbon acquisition by holobionts with clade C symbionts compared to clade A under low
628
irradiance levels using isotope labelling (13C and 15N), but this appeared to be due to increased
629
heterotrophic capacity by corals hosting clade C endosymbionts. However, other non-isotope studies did
630
not find an influence of symbiont genus on host metabolism (Matthews et al. 2020). Endosymbiont
631
communities do influence coral SI values though. Coral colonies dominated by clade D have lower δ13C in
632
both host and symbiont tissues compared to colonies dominated by clade C (clade C host and
633
endosymbiont δ13C 1.6 and 1.5 ‰ higher than clade D in summer); but these differences were inferred to
634
be driven by light availability rather than coral feeding (Wall et al. 2020).
635
The strong focus on adult coral colonies can ignore the importance of vulnerable larval stages that support
636
the maintenance of coral reefs through successful recruitment and ontogenetic development. In an effort
637
to better understand planktonic coral larvae metabolism, Alamaru et al. (2009b) conducted feeding
638
experiments with various potential food sources phytoplankton, zooplankton and bacteria - but failed to
639
26
observe active feeding or heterotrophic uptake via δ13C and δ15N SI data. This is despite planulae larvae
640
having an oral opening. They therefore inferred that coral larvae rely wholly on lipid reserves or
641
photosynthates from symbionts that are already present in the endoderm. This was further elaborated by
642
pulse-chase experiments using labelled 13C-bicarbonate and 15N-nitrate conducted by Kopp et al. (2016),
643
who determined that there is minimal contribution by symbionts to host metabolism, demonstrating that
644
coral larvae are essentially lecithotrophic pre-settlement.
645
4.5. Metabolism in Non-Hard Coral Symbioses
646
A significant number of SI studies have explored the metabolic underpinning of other non-coral symbiotic
647
groups, revealing complex physiologies. Sponges possess an intricate symbiotic system due to the diversity
648
of the microbiomes they can host, including photosymbionts (Davy et al. 2002, Weisz et al. 2010). However,
649
the strength of the symbiotic dependency appears to vary between host sponge species, with both tight
650
and weak metabolic couplings being observed (Freeman et al. 2015). Isotopically enriched seawater with
651
13C and 15N tracers has been used to demonstrate nutrient transfer between microbes and host sponges,
652
the rate of which appears dependent on symbiont identity and irradiance rather than overall symbiont
653
abundance (Freeman et al. 2013). Interestingly, labelled 15N-ammonium and 13C-bicarbonate revealed that
654
coral-excavating sponges often host both Symbiodiniaceae and prokaryotic symbionts, with the former
655
undertaking significant inorganic nutrient fixation and transfer to host bioeroding sponges, but the latter
656
not contributing to nutrient assimilation (Achlatis et al. 2018). Conversely, 13C- and 15N-enriched DOM show
657
that sponge hosts can directly take up and utilise DOM and subsequently transfer significant dissolved
658
organic waste products to symbionts (Achlatis et al. 2019). This can constitute the entire nitrogen budget of
659
hosted algae (Davy et al. 2002), in a similar fashion to coral hosts. Pulse chase experiments with isotopically
660
labelled 13C‐ and 15N‐enriched coral‐ and algal‐derived DOM show that various sources of DOM are utilised
661
by sponges, but algal derived sources are predominantly transferred to the microbiome while coral derived
662
DOM is used directly by the sponge host (Rix et al. 2017). Interestingly, bulk SI (δ13C and δ15N) and isotope
663
labelling (13C and 15N) revealed that the encrusting sponge, Terpios hoshinota, which kills corals by
664
overgrowing them, does so for space, not food; in the new sponge tissues only 9.5% of the C and 16.9% of
665
27
the N was derived from the corals underneath (Syue et al. 2021). In contrast, contact association with
666
macroalgae appears to competitively inhibit sponges, with SI enrichment experiments (13C and 15N)
667
demonstrating nitrogen transfer from sponge to macroalgae (Easson et al. 2014). As with coral holobionts,
668
changes in the surrounding environment can alter the dependency of sponge hosts on photoautotrophic
669
symbiont production. Recent bulk and amino acid CSIA work in the Caribbean demonstrated increasing
670
heterotrophy with depth for sponges, but highlighted species-specific trends in host utilisation of POM and
671
DOM, and internal translocation of these (Macartney et al. 2020).
672
The host-symbiont relationship in gorgonians has been the focus of more recent interest in holobiont
673
metabolism utilising SIs, likely due to their increasing presence in some coral reef systems (Rossi et al.
674
2020). While overall autotrophic reliance varies seasonally and with species, endosymbionts can deliver the
675
majority of energetic demands (> 95%), with heterotrophically acquired carbon sources contributing less
676
than 5% year-round for gorgonian species examined in the Caribbean (Rossi et al. 2020). Conversely,
677
gorgonians examined in Taiwan appear to be highly dependent on heterotrophic inputs based on host and
678
symbiont δ13C and δ15N values, with relatively little energetic benefit garnered from their endosymbionts
679
(Hsu et al. 2020). Similarly, recent work on Antipatharian soft corals (black corals) using isotope labelling
680
(13C and 15N) has demonstrated that they can use a range of different food sources, likely allowing them to
681
exploit seasonal fluctuations in prey concentrations (Rakka et al. 2020). These studies highlight the diversity
682
in energy acquisition within this group. While environmental sensitivities to gorgonian symbioses have yet
683
to be thoroughly explored, depth appears to have little impact on gorgonian tissue δ15N values (change in
684
only 1.4 over 20 m), implying limiting physiological effects associated with ambient light (Baker et al.
685
2011). Further, short-term (seven day) nutrient enrichment does not appear to significantly impair the
686
symbiosis between gorgonians and their dinoflagellate endosymbionts, but can result in increased
687
chlorophyll content and algal densities within hosts (McCauley and Goulet 2019).
688
Other forms of symbiotic relationships, while lesser studied, are likely commonplace on coral reefs and
689
constitute important components to ecosystem functioning (Pinnegar and Polunin 2006). Anemones may
690
also host Symbiodinium as a source of autotrophic nutrition, however SI labelling (13C and 15N) highlights
691
28
the competitive nature of resource acquisition by both host and endosymbiont, implying a less stable
692
symbiotic association compared to true coral holobionts (Radecker et al. 2018). Carbon and nitrogen SI
693
data from both lab and field isotope-labelling feeding experiments have been used to demonstrate nutrient
694
transfer from anemonefishes to both fish-hosting anemones and their endosymbionts forming a tripartite
695
symbiosis (Cleveland et al. 2011). Parmentier and Das (2004) examined relationships between four species
696
of apparently parasitic carapid fishes and their echinoderm hosts using bulk SIs, but only found evidence
697
for feeding on host tissues for one species (Echeliophis gracilis), suggesting commensal associations for the
698
other three fishes. Similarly, while shallow-water black corals host a variety of macrosymbionts, bulk δ13C
699
and δ15N reveal that they do not use their host as a main food source, but instead use the coral’s structure
700
to access nutrition from the water column (Terrana et al. 2019). True parasitic-host relationships have also
701
been explored using bulk SIs in reef fish and gnathiid ectoparasites (Demopoulos and Sikkel 2015). Finally,
702
upscaling from individual metabolic processes to whole community metabolic functioning can be facilitated
703
through SI approaches. Spatial differences in isotopic discrimination of overlying seawater DIC are
704
attributable to the varying ratio of calcification:primary production of the underlying ecological community;
705
e.g. ~5‰ higher δ13CDIC discrimination for portions of the reef with more non-calcifiers compared to
706
portions of the reef with more calcifiers (Koweek et al. 2019).
707
5. Trophic niches
708
5.1. Background
709
Ecological niches are multidimensional spaces defined by environmental conditions and resource
710
utilisations that are occupied by an organism (or population) where their survival curves are optimised
711
(Hutchinson 1957). The trophic niche relates to the array of food items consumed by an organism, which
712
constitutes a subset of its overall ecological niche. Understanding an animal’s resource use, and how this
713
varies within and among species and guilds, helps determine its trophic function within an ecosystem and
714
how this might respond to environmental change. Increasingly, studies are using SI to estimate an animal’s
715
trophic niche. Isotope data are typically presented on a bi-plot using the isotope values (δ‐values) as
716
29
coordinates (Fig. 3) (although note that tri-isotope plots are now being used; Skinner et al. 2019a). The area
717
(δ‐space) of these coordinates is determined to be the animal’s isotopic niche, providing an understanding
718
of its trophic ecology by reflecting some aspects of their trophic niche (Fig. 3) (Newsome et al. 2007). For
719
example, the size of the isotopic niche and position of the individual coordinates indicate intraspecific
720
variation in resource use, known as the niche width (Bearhop et al. 2004). SI approaches to trophic niche
721
determination are the newest of our defined topics to emerge in SIs on coral reefs (first paper published in
722
2007) and it has expanded rapidly since then (n = 48; Fig. 1). This is likely due to the instrumental Layman et
723
al. (2007) and Jackson et al. (2011) papers which have brought community SIA into the foreground.
724
However, although related, trophic niches and isotopic niches are not interchangeable, and care should be
725
taken when using these terms (see Reddin et al. 2018, Hette-Tronquart 2019). In some cases, variation in SI
726
values may be independent of diet, e.g. where habitat-derived isotopic baselines differ (Fig. 3; Fig. 4).
727
Organism foraging behaviour and habitat use must also be considered before converting an isotopic niche
728
to an ecological or trophic niche (Flaherty and Ben-David 2010).
729
5.2. Isotopic niches
730
To date, coral reef isotopic niche studies have predominantly explored resource partitioning within guilds,
731
likely as the high densities of species with seemingly similar functional roles raises questions as to the
732
mechanisms of their coexistence. This can be elucidated by SI data, revealing dietary variation and
733
intricacies which were previously overlooked. For example, while traditional techniques suggest that
734
herbivorous surgeonfish and parrotfish consume similar production sources on reefs, bulk δ13C and δ13N SI
735
data reveal complex trophic ecologies, indicating a high level of functional diversity (Plass-Johnson et al.
736
2013, Dromard et al. 2015). Similarly, isotopic niche overlap (see Fig. 3) among sympatric spotted
737
(Panulirus guttatus) and Caribbean (P. argus) spiny lobsters is minimal, with each utilising different food
738
sources and occupying unique TPs (evidenced by δ15N) in the food web (Segura-García et al. 2016). Higher
739
up the food chain, coral reefs have high biomasses of predators with seemingly similar traits and trophic
740
ecologies. Yet δ13C and δ13N reveal that sympatric species of coral trout Plectropomus laevis and P.
741
leopardus, have resource uses that differ (Matley et al. 2017, Matley et al. 2018). They likewise also vary
742
30
from other predatory teleosts (Lethrinus miniatus and Lutjanus carponotatus), implying degrees of trophic
743
specialisation in coral reef mesopredators (Frisch et al. 2014). Indeed, in the Maldives, the resource uses of
744
seven sympatric teleost reef predators across multiple families vary both within and among species and
745
spatially, with δ34S acting as a useful third isotope further differentiating individual feeding behaviours
746
(Skinner et al. 2019a). Furthermore, on Bahamian reefs, while isotope biplot data (δ13C and δ15N) of the
747
invasive lionfish (Pterois spp) and a native snapper Lutjanus griseus appeared to overlap considerably, their
748
core isotopic niches did not, suggesting competition was not as high as initially perceived (Layman and
749
Allgeier 2012). These predators appear to occupy the same functional roles within their guilds, but by
750
partitioning trophic resources both spatially and temporally, inter- and intraspecific competition is likely
751
reduced, thereby facilitating their coexistence and altering previously assumed ecological roles (Dale et al.
752
2011, Gallagher et al. 2017, Matich et al. 2017, Curnick et al. 2019, Skinner et al. 2019a).
753
Individual specialisation within a population is a mechanism through which sympatric individuals within an
754
age or size class may reduce competition for resources by focusing on a narrower set of resources than that
755
of the population as a whole (Bolnick et al. 2003, Araújo et al. 2007, Araújo et al. 2011). Occurrences of
756
individual specialisation are expected to be greater where resource diversity is high, as there is increased
757
ecological opportunity (Semmens et al. 2009, Araújo et al. 2011). Coral reefs, with their high rates of
758
biodiversity, should therefore be prime locations for occurrences of individual specialisation. This is
759
compounded for populations with access to two or more adjacent resource pools (e.g. benthic and pelagic)
760
or habitats (e.g. reef and seagrass), as there is a greater array of potential resources (Araújo et al. 2011,
761
Matich et al. 2019). SIs are a powerful tool which can quantify these specialisations and are less costly and
762
labour-intensive than long-term dietary records (Newsome et al. 2009). To date, studies investigating
763
individual specialisation using SI on coral reefs have concentrated on elasmobranchs and large predatory
764
teleosts (Shipley et al. 2018, Shiffman et al. 2019, Skinner et al. 2019a, Wyatt et al. 2019). The degree of
765
specialisation across lower trophic levels is little explored. However, 34% of SI coral reef studies involve a
766
single species, which suggests there is sufficient opportunity to investigate individual specialisation more
767
generally on coral reefs. Indeed, there is some evidence of such specialisation at lower trophic levels; bulk
768
31
δ13C and δ13N SI show that damselfish, Dascyllus aruanus, are more specialised when colonies are larger,
769
suggesting local abundances drive intra-group competition, which is then modulated by individual
770
specialisation (Frédérich et al. 2010). Groups of damselfish with narrower trophic niches (evidenced by
771
their isotopic niches using δ13C, δ13N, and δ34S) also have lower genetic diversity, highlighting potential links
772
between population- and trophic ecology (Gajdzik et al. 2018). Exceptional variation in feeding strategies
773
from heterotrophy to autotrophy at scales of metres to kilometres among colonies of the hard coral
774
Pocillopora meandrina has also been identified using δ13C of essential amino acids, showing no relationship
775
with site or depth (Fox et al. 2019). Although consumer dietary specialisation does not occur in all systems
776
or between all species (Gallagher et al. 2017), it warrants further study, particularly in systems where
777
resources are fluctuating. If consumers partition resources, their ecological roles may be vastly different,
778
which could be masked by traditional species- or guild-level categorisations.
779
Attempts by ecologists to categorise species functional traits to better understand ecosystem function
780
conflicts with the natural variability inherent in complex systems such as coral reefs. SI data have
781
successfully been used to refute some strict dietary classifications of reef organisms derived from
782
traditional ecological studies using gut contents data and in situ observations. For example, in Papua New
783
Guinea, of seven damselfish species sampled, their δ13C and δ15N SI values and corresponding isotopic
784
niches, indicated that none were strict herbivores, despite traditionally being classified as such (Eurich et al.
785
2019). In fact, using δ13C and δ13N damselfish have relatively distinct isotopic niches, reflecting varying
786
degrees of planktonic to benthic reliance (Frédérich et al. 2009, Lepoint et al. 2016, Olivier et al. 2019).
787
Similarly, cardinalfishes, thought to be generalist carnivores based on stomach contents data, are sustained
788
by production sources from across a benthic-planktonic gradient, and δ13C and δ13N indicate distinct
789
species-specific isotopic niches using (Frederich et al. 2017). Indeed, community-wide δ13C and δ15N data
790
show that many strictly categorised reef fish rely on production sources outside their putative diet source
791
and this spans multiple trophic levels (Zhu et al. 2019), highlighting how narrow functional group
792
categorisations should be applied with caution. However, care should be taken when interpreting isotopic
793
spread as solely due to diet variation; spatiotemporal differences in production source isotopic baselines
794
32
and a myriad of other drivers may also cause such variations (Fig. 3 and Fig. 4; and see 7. Environmental
795
Drivers).
796
In some cases, there may be greater redundancy among different guilds rather than within them; δ13C and
797
δ15N SI values of muscle tissue from reef sharks and large teleost mesopredators indicated that they occupy
798
the same isotopic niche on the Great Barrier Reef, suggesting they perform similar functional roles (Frisch
799
et al. 2016). Similarly, in the Hawaiian archipelago, there was considerable overlap in the δ13C and δ15N
800
values of reef sharks and larger teleosts, highlighting a degree of functional redundancy which could
801
contribute to ecosystem stability (Hilting et al. 2013, Roff et al. 2016). However, a study further north on
802
the Great Barrier Reef using multi-tissue δ13C and δ13N data (muscle, blood, and plasma) to compare
803
resource use of reef sharks and large predatory teleosts identified significant trophic separation (Espinoza
804
et al. 2019). While Frisch et al. (2016) call for apex predator reef sharks to be reassigned as high-level
805
mesopredators, Espinoza et al. (2019) suggested there is substantial diversity in their trophic ecologies and
806
likely wider functional roles. This suggests that different localities may have populations with different
807
behaviours. Beyond these studies of reef mesopredators, few studies compare isotopic niches among
808
differing guilds on coral reefs, likely because they are expected to have different dietary strategies. For
809
example, overall, the structure of the reef fish assemblage remains fairly consistent, with herbivores
810
occupying lower trophic levels (reflected by low δ15N), while carnivores and piscivores feed across a
811
broader range of trophic levels and resources (Carassou et al. 2008, Yang et al. 2012). Most studies
812
comparing niches and categorisations do so across similar species groups; of the 47 articles assigned to the
813
Trophic Niche topic, 64% focus on reef fish and 15% on elasmobranchs. Few, if any, compare resource use
814
across broader groupings (e.g. some invertebrates may feed on the same prey as some reef fish; Zapata-
815
Hernández et al. 2021), so wider competitive feeding relationships may be missed.
816
Multi-tissue SI data can reveal detailed feeding strategies and how these fluctuate over different
817
timescales, particularly regarding individual specialization within a population (Bond et al. 2016). For
818
example, a multi-tissue (plasma, cartilage, and faeces), multi-isotope approach identified consistent
819
individual specialization on either oceanic or coastal prey in whale sharks, Rhincodon typus (Wyatt et al.
820
33
2019). Yet, they are infrequently utilised on coral reefs. Of the 29% of identified SI reef studies employing a
821
multi-tissue approach, only 9% measure multiple whole tissue types, e.g. muscle and liver, of the same
822
individual consumer to elucidate short- and longer-term fluctuations in resource use. Instead, 20% of multi-
823
tissue studies focus on exploring the trophic niches of mixotrophic holobionts, e.g. host soft tissue and
824
endosymbionts. For example, the degree of overlap in the isotopic niches (determined using bulk δ13C and
825
δ15N) of seven different genera of coral hosts and their endosymbionts was linked to coral trophic strategy;
826
higher overlap indicated resource sharing between host and endosymbionts (i.e. greater host autotrophy)
827
while lower overlap indicated less resource sharing (i.e. greater host heterotrophy) (Conti-Jerpe et al.
828
2020). While multi-tissue analyses may offer additional insight, comparing isotopic niches among guilds
829
helps understanding of how resource use and partitioning are structured within the wider community. One
830
of the few studies using SIA to investigate trophic interactions at the community level confirmed previous
831
hypotheses that the sea cucumber Stichopus herrmanni has a top-down influence (consistently higher δ15N
832
across seasons) on its meiobenthic prey in a lagoon system on the Great Barrier Reef (Wolfe et al. 2021).
833
For inferences to be made about resource use and trophic positions with confidence, knowledge on diet-
834
tissue discrimination factors (hereafter ‘trophic discrimination factor’ (TDF), usual notation is Δ) between a
835
consumer and their diet are necessary (see reviews referenced in 1. Introduction; Boecklen et al. 2011,
836
McCormack et al. 2019, Whiteman et al. 2019). The standard TDFs that are typically applied to bulk carbon
837
(~0.5) and nitrogen (~3.4) are averages drawn from broader syntheses of laboratory experiments, but are
838
known to vary among species, by life stage and season, among other factors (Wyatt et al. 2010a, Wyatt et
839
al. 2019). For example, the estimated Δ15N between the hard coral Porites lutea and its food source is only
840
1 (Rangel et al. 2019), three times lower than the commonly applied value. Average TDF values may be
841
appropriate for studies at the ecosystem level, but greater resolution specificity is required for individual
842
species. To date, very few studies (n ≈ 5) have empirically determined species-specific TDF for coral reef
843
organisms, despite a call for more lab studies to do so over a decade ago (Wyatt et al. 2010a). Evidence
844
suggests that typically employed TDFs may not be wholly applicable to coral reef ecosystems: feeding
845
observations and modelling indicated that three herbivorous reef fish have consistently higher Δ15N (model
846
34
estimates ranged from 4.30‰ to 5.68‰) than the accepted value of 3.4‰ (Mill et al. 2007). Furthermore,
847
a comprehensive study using CSIA of >200 samples from 47 species of marine teleosts including many reef
848
fish found that the estimated TDFAA was significantly lower (2‰) than previously accepted values for CSIA
849
(Bradley et al. 2015). Given the huge diversity of organisms on coral reefs, a greater understanding of how
850
these values vary among species and compounds, and with other factors such as seasonality, is required
851
(Whiteman et al. 2019) to improve understanding of coral reef trophic interactions. One such avenue might
852
involve parasites (an understudied group on coral reefs), which often reflect their host’s feeding ecology
853
and TP (indicated by their δ13C and δ15N values) (Jenkins et al. 2018). This suggests future research could
854
explore this relationship to better understand the complex trophic interactions and feeding strategies that
855
could affect reef organism’s TDF values, but SI data between fish hosts and their parasites are likely to be
856
complicated (Pinnegar et al. 2001).
857
As time-integrated values of the major source reliance and TP of an organism in the food web, SI data
858
cannot generally be used to identify specific prey. For example, while bulk δ13C and δ15N values were
859
similar across four species of mesopredatory teleosts, stomach contents data revealed substantial
860
differences in the types of prey taxa consumed (Ashworth et al. 2014). Similarly, among 21 species of
861
butterflyfish, gut contents data and in situ feeding observations were required to resolve finer scale
862
differences in prey items (Nagelkerken et al. 2009). SIs are therefore not necessarily suitable at
863
differentiating ecological niche differences that vary over smaller scales, especially when isotopic
864
differences between sources are small. SIs are also unable to distinguish between behavioural differences
865
in feeding: two sympatric species of fish that have approximately the same diet, but with one feeding
866
nocturnally while the other diurnally, would occupy the same position in isotopic space (given the same
867
TDFs). It is therefore often useful to combine multiple methodologies to obtain a comprehensive overview
868
of the complexities of an organism’s feeding strategies across different temporal and spatial scales on coral
869
reefs. As researchers become more accustomed to using SI data to understand food webs, the number of
870
isotopic niche studies on coral reefs is increasing, but other data, including traditional techniques, often
871
35
complement and add to the inferences deduced from SI data alone (e.g. inclusion of fatty acids, see Dethier
872
et al. 2013).
873
5.3. Drivers of isotopic niche variation
874
Isotopic niches may vary spatially, particularly in diverse systems such as coral reefs, helping to explain how
875
species partition resources (Fig. 3). This variation can occur across relatively large scales: across the
876
Southern Line Islands, isotopic niche widths (from δ13C and δ15N) of reef fish populations change in relation
877
to available primary production (Miller et al. 2019). They may also change at much smaller scales: bulk δ13C
878
and δ15N data show that several reef shark species have different patterns of resource use among coastal
879
areas of Florida (Shiffman et al. 2019) and dolphins have varying levels of coastal vs offshore resource use
880
in Panama (Barragan-Barrera et al. 2019). While isotopic niche variation can be driven by factors other than
881
dietary preference (Fig. 3), understanding the relationships between SI values and these drivers is non-
882
trivial. For example, predator biomass influences the dietary diversity (represented by the δ13C and δ15N
883
isotopic niche area) of two herbivorous reef fish in the Florida Keys, likely by causing a change in their
884
group foraging activities and perceived predation risk (Catano et al. 2014). Other biotic influences on
885
isotopic niches include the presence of parasites: ectoparasitic isopods affect the resource use (inferred
886
through δ13C and δ15N) of grunts, likely causing them to feed in different localities to uninfected fish, and
887
changing their condition (Welicky et al. 2017). On reefs in New Caledonia, partitioning of resources
888
indicated by bulk δ13C and δ15N data by two species of sea krait indicated habitat-based dietary divergence
889
(Brischoux et al. 2010). Environmental conditions may also influence isotopic niches: various flow
890
conditions led to differing physiological adaptations of two sympatric bonefish species based on δ13C and
891
δ18O data (Haak et al. 2018). These studies highlight how a range of factors may influence an organism’s
892
isotopic niche at any one time and underline the inherent difficulty in extrapolating findings from one
893
location to another. Other factors which may influence isotopic niches include growth rate, metabolism,
894
and diet quality, however to our knowledge, there are no studies investigating these relationships on coral
895
reefs to date. With few studies investigating the effect of multiple variables or drivers, or the indirect
896
36
relationships occurring between them, the ability to generalise such patterns across coral reef ecosystems
897
is limited.
898
Anthropogenic impacts can influence isotopic niches and resource use, as habitats degrade or prey groups
899
fluctuate, but such responses vary among species. For example, corallivores had smaller δ13C and δ15N
900
isotopic niches at a degraded reef site compared to a healthy reef site, likely reflecting a narrower pool of
901
available resources from a loss of live coral cover (Letourneur et al. 2017). In contrast, invertivores had
902
larger isotopic niches, and herbivores and zooplanktivores displayed no difference (Letourneur et al. 2017).
903
Similarly, δ15N in spiny lobster muscle tissue did not differ between reefs of different levels of degradation,
904
with no apparent trend with the loss of habitat structural complexity (Lozano-Álvarez et al. 2017). The
905
impact of reef habitat degradation on organic matter sources and the trophic ecology of reefs in the
906
Caribbean (where 14% of SI reef studies have been conducted; Fig. 2) was quantified by Morillo-Velarde et
907
al. (2018), on the basis of switches or widening in carbon sources as evidenced by reef fish δ13C in degraded
908
habitats. This indicates that there is a degree of trophic plasticity, with some groups potentially adapting to
909
degradation while others do not, but there may be a cost to this adaptation. As the biomass of planktonic
910
damselfish decreased on a degraded reef, δ13C and δ15N of the spotted coral grouper, Plectropomus
911
maculatus, indicated it switched to foraging on herbivorous damselfish. This was linked to long-term
912
declines in the groupers’ growth rate, fecundity, and survivorship (Hempson et al. 2017, Hempson et al.
913
2018), highlighting potential negative, long-term consequences of anthropogenic impacts across the wider
914
coral reef food web.
915
6. Fish diet variation and habitat connectivity
916
6.1. Background
917
In coral reef SI studies, reef fish are the dominant focal taxa (36% of all articles; Fig. 5B) and a major area of
918
research involves using SIs to determine how their habitat and resource use varies spatially and throughout
919
their life cycle (n = 44 identified studies; Fig. 3). Although this topic includes one of the earliest known
920
papers using SIs on coral reefs (Fry 1982), much of the bulk of the work on fish trophodynamics was
921
37
conducted from the late 2000s onwards, with a focus on the western Atlantic (n = 16) (Fig. 2B). While there
922
are thousands of fish living on coral reefs, only few have been the focus of multiple SI studies, namely the
923
whitetail damselfish Dascyllus aruanus (n = 5), leopard coral grouper Plectropomus leopardus (n = 6),
924
blackspot snapper Lutjanus ehrenbergii (n = 5) and the red lionfish Pterois volitans (n = 5). Similarly, coral
925
reef SI studies focus on common groups (e.g. damselfish, surgeonfish, groupers, snappers), with few to
926
none investigating the trophodynamics of well understood and often cryptic groups such as blennies and
927
gobies.
928
6.2. Ontogenetic habitat shifts
929
Ontogenetic habitat shifts are common in marine organisms as they strive to maximize their fitness
930
strategies (Schmitt and Holbrook 1985, Holbrook and Schmitt 1988, Dahlgren and Eggleston 2000). In coral
931
reef ecosystems, it is common for juvenile reef fishes to spend time in a nursery habitat such as mangroves
932
or seagrass meadows, prior to migrating to the reef (Fig. 3). SIA can be a cost-effective tool to pinpoint
933
when and how these changes occur by analysing the SI values of juveniles through to adults across the
934
nursery and reef habitats. This allows the timing of the ontogenetic habitat and diet shifts to be estimated
935
(Cocheret de la Morinière et al. 2003, Frédérich et al. 2012, McMahon et al. 2012, Berkström et al. 2013).
936
For example, the δ13C of juvenile snappers in mangroves reflect mangrove habitat (-23 to -17 ‰), while
937
individuals on the reef shift to a reef δ13C signature with increasing body size (-16 to -8 ‰), indicating that
938
the smaller individuals within reefs have migrated there from the mangroves (Nakamura et al. 2008).
939
Acoustic telemetry can be used to further support inferences from SI data; δ13C of fin tissue and acoustic
940
tracking of the schoolmaster snapper, Lutjanus apodus, revealed that they move from bays to coral reefs as
941
they get larger (i.e. smaller fish mean δ13C -16.7 ‰, larger fish -12.2 ‰) (Huijbers et al. 2015).
942
Fish otoliths, as an incrementally grown, metabolically inert biological structure, provide the ability to
943
determine when ontogenetic shifts occur within a single individual, and have been utilised across several
944
coral reef SI studies (n = 13). As surface otolith material is continuously deposited with age, a time-series of
945
SI data can be derived by sampling progressive segments from the core (reflecting larval life stages) to the
946
38
edge (age of the fish just prior to otolith sampling). Changes in SI values will reflect potential changes such
947
as in habitat, diet, or physiology. In the Red Sea, otolith δ13C essential amino acid values of the blackspot
948
snapper, Lutjanus ehrenbergii, across a gradient from coral reefs to seagrass were influenced by the
949
habitats the fish resided in, offering an opportunity to track their movements across the isoscape
950
(McMahon et al. 2011). In addition to δ13C, some studies (6%; Fig 2B) use oxygen isotopes 18O) in otolith
951
segments (Blamart et al. 2002), as they vary with temperature and salinity, thus providing information on
952
the external environment. Both δ13C and δ18O in otolith core and edge segments of the trumpet emperor,
953
Lethrinus miniatus, on the Great Barrier Reef revealed contrasting movements of juveniles across different
954
latitudes and corresponding isotopic environments (Currey et al. 2014). Similarly, δ13C and δ18O of gray,
955
Lutjanus griseus, and yellowtail, Ocyurus chrysurus, snapper sub-adult otoliths in Florida determined the
956
distinct nursery areas in the Florida Bay that the fish had migrated to the reef from (Gerard et al. 2015).
957
Otolith SI values can be incorporated into spatial simulation models, revealing that the geographic
958
distribution of the nursery areas plays an important role driving the spatial distribution of the adults on the
959
reefs (Hujibers et al. 2013). Recently, fish eye lenses have also been advocated as an alternate,
960
metabolically stable, incrementally grown structure with which to track resource and habitat use shifts with
961
SIs across different life stages (Quaeck-Davies et al. 2018, Curtis et al. 2020, Vecchio and Peebles 2020).
962
Clearly, SI data can be used to delineate fish movement patterns, tracking animal movements in the
963
absence of electronic tagging, and identifying nursery habitats that should be prioritized for management;
964
habitat impacts may disproportionately affect certain life stages thus influencing populations elsewhere,
965
e.g. recruitment of juveniles to reefs. However, if the animals move through the habitat before an isotopic
966
signature can be recorded, or if the habitats are isotopically indistinguishable, then movement information
967
may be missed (McMahon et al. 2013). As such, use of traditional approaches in conjunction such as
968
tagging (e.g. Huijbers et al. 2015), spatial simulation (e.g. Hujibers et al. 2013), and fish size distribution
969
modelling (e.g. Mumby et al. 2004) should be considered.
970
6.3. Body size
971
39
Multiple studies indicate that variation in resource use may occur as a function of body size (e.g. Layman et
972
al. 2005, Romanuk et al. 2011). Indeed, many predatory fishes across various biomes tend to feed at higher
973
TP (evidenced by higher δ15N) as body size increases, a pattern that also holds true for coral reef systems
974
(Layman et al. 2005, Greenwood et al. 2010). For example, in Moorea, French Polynesia, plasma of blacktip
975
reef sharks, Carcharhinus melanopterus, was enriched in δ15N with increasing body size but there was no
976
change in δ13C, suggesting they were feeding on higher trophic level prey but reliant on the same basal
977
production sources (Matich et al. 2019). Such trends are however far from ubiquitous in coral reefs. For
978
blacktip reef sharks at Palmyra Atoll, there was a positive relationship between body size and δ15N in one
979
lagoon, but no relationship with body size at all in another, suggesting that the two shark populations have
980
different trophic ecologies despite their proximity to one another (Papastamatiou et al. 2010). Similarly, in
981
the Caribbean reef shark Carcharhinus perezi there was no relationship between δ15N and body size but a
982
significant positive relationship with δ13C, implying larger individuals relied more on lagoonal food sources
983
with increasing size rather than feeding at higher trophic levels (Bond et al. 2018). Clearly, relationships
984
between predator TP and body size are not always positive or absolute (Gallagher et al. 2017, Matley et al.
985
2017, Shipley et al. 2018, Skinner et al. 2019a, Eddy et al. 2020). This is demonstrated by the invasive
986
lionfish in the Caribbean, which are smaller than native Nassau groupers, yet occupy the highest TP in the
987
study region (represented isotopically by higher δ15N) (O'Farrell et al. 2014). These studies highlight how
988
trophic ecology of fishes can vary with body size, indicating that an organism’s ecological role may be
989
complex and life stage-, as well as species-specific. However, for most herbivorous reef fish, TP
990
(represented by δ15N) remains relatively unchanged as body size increases (Cocheret de la Morinière et al.
991
2003, Greenwood et al. 2010, Plass-Johnson et al. 2013, 2015a). This is likely as, although they may access
992
different resources with increasing body size, the δ15N of these basal production sources remains fairly
993
similar.
994
Despite being lesser studied, similar trends with body size have also been observed in invertebrates. The
995
fireworm, Hermodice carunculata, a facultative corallivore, is enriched in δ15N and has fluctuating δ13C with
996
increasing body size, potentially reflecting feeding at higher trophic levels while diversifying resource use
997
40
across ontogeny (Wolf et al. 2014). It should be noted, however, that body size relationships with SIs
998
assume there is minimal change in isotopic baselines and growth/metabolic influences on isotopes
999
independent of diet (e.g. constant TDFs across ontogeny). Intra-specific and life-stage feeding
1000
specialisations, such as these, may help promote population resilience to environmental change, as
1001
individuals and populations are reliant on a wider range of resources across ontogeny. However, given that
1002
many of the anthropogenic impacts on reef-fishes are size- and species-selective, with many targeted fish
1003
often larger and functionally important (Benoît and Swain 2008, Lokrantz et al. 2009, Plass-Johnson et al.
1004
2015a), this may have serious consequences for the trophodynamics of the coral reef food web.
1005
6.4. Residency and population connectivity
1006
SIA can reveal an organism’s residency within a habitat at various life stages, representing a more cost-
1007
effective approach than tagging or tracking. This can be useful for management of certain species when
1008
assessing the efficacy of protected areas. In Australia, δ13C and δ15N values of reef fish liver and muscle
1009
revealed three species were resident in the area, while others had migrated from coastal riverine habitats
1010
(Davis et al. 2015). In western Australia, δ13C and δ15N of muscle tissue of several reef sharks in coastal
1011
habitats confirmed residency was high across four species (Speed et al. 2012). Examination of δ13C and δ15N
1012
in lionfish, Pterois volitans, muscle tissue showed they did not move between mangroves and reefs; in fact
1013
there was no overlap in habitat or resource use of lionfish between habitats, confirming them as site-
1014
attached opportunistic foragers (Pimiento et al. 2015). SI data can also help assess the importance of
1015
different habitats to more mobile consumers that have larger home ranges; for example one of the few
1016
studies to use δ13C, δ15N, and δ34S, to examine habitat residency determined that the reef manta ray,
1017
Manta alfredi, was heavily dependent on lagoonal resources, suggesting long periods of residency in the
1018
lagoon (McCauley et al. 2014). Combining SI data with fatty acids further revealed that M. alfredi are
1019
secondary consumers that rely on both epipelagic and demersal zooplankton, reflecting their ability to
1020
access disparate resources through vertical and horizontal movements (Couterier et al. 2013). While SI data
1021
can provide much needed insight into movement patterns and habitat usage, it should be noted that
1022
41
connectivity ultimately depends on the spatial configuration of the seascape (Nagelkerken et al. 2008,
1023
Rooker et al. 2018).
1024
Larval population dynamics are inherently more difficult to study; larvae are hard to follow due to their
1025
high natural mortality and rapid dispersal by ocean currents (Cowen and Sponaugle 2009). Nevertheless, SI
1026
approaches provide an opportunity to track larval dispersal and understand population connectivity more
1027
easily. Transgenerational isotope labelling involves injecting adult female fish with labelled isotopes (137Ba)
1028
and results in consistently and permanently marked larvae throughout a reproductive season. This method
1029
has minimal impact on the fish, their eggs, and larvae as they develop, or to those that consume them
1030
(Williamson et al. 2009a, Williamson et al. 2009b, Roy et al. 2012, Cuif et al. 2014). By analysing the SI
1031
values of the otolith cores of the new cohorts, the degree of connectivity and self-recruitment within a
1032
population can be determined. In New Caledonia, this approach revealed that self-recruitment of
1033
damselfish Dascyllus aruanus varied significantly between months and years, but was independent of the
1034
proportion of self-recruits within the population (Cuif et al. 2015). This suggests that self-recruitment can
1035
successfully indicate population openness but may not relate to population persistence (Cuif et al. 2015).
1036
6.5. Habitat connectivity
1037
There are important exchanges of organisms and energetic material between coral reefs and other
1038
adjacent habitats that can be difficult to measure (Polis and Strong 1996, Huxel and McCann 1998). SI data
1039
offer distinct opportunities to identify these cross-system linkages and quantify fluxes across habitat
1040
boundaries; the flow of nutrients into and across shallow coral reef ecosystems is increasingly being
1041
documented (Fig. 3). In shallow-water reef habitats in the Caribbean, δ13C suggests that benthic algae and
1042
seagrass contribute 48-76% of carbon to reef fish (Fry et al. 1982), but in Moreton Bay, in eastern Australia,
1043
dietary proportions vary with distance to adjacent mangrove and seagrass habitats (Davis et al. 2014). Such
1044
patterns are not necessarily surprising, however they highlight that contributions of various exogenous
1045
materials to reefs are likely to be site-dependent due to varying seascapes (Briand et al. 2015).
1046
Nevertheless, looking at the wider spatial context can reveal less intuitive habitat linkages. In the Chagos
1047
42
Archipelago, it has been shown that rats interrupt nutrient flows between pelagic, coral reef, and island
1048
ecosystems. Rat-free islands had significantly greater densities of seabirds and therefore larger deposits of
1049
nitrogen and subsequent runoff, leading to higher δ15N in the soil, macroalgae, turf algae, and reef fish
1050
compared to rat-infested islands where seabird densities were lower (Graham et al. 2018). Sulphur isotopes
1051
34S) have rarely been applied to infer habitat connectivity but they may be well placed to distinguish and
1052
help identify sources produced under anaerobic conditions, e.g. decomposition of mangrove organic
1053
matter, due to the strong fractionations that occur during such processes (Okada and Sasaki 1998, Granek
1054
et al. 2009). More widely, mangrove, microalgae, macroalgae, and seagrass exhibited greater separation in
1055
δ34S compared to δ13C in a study conducted in Bocas del Toro (Panama) (Granek et al. 2009). Used together,
1056
δ34S and δ13C suggested that mangrove-coral reef nutrients contributed up to 57% to the biomass of sessile
1057
reef invertebrates (Granek et al. 2009). SI data thus point to mangroves as an important source of nutrients
1058
for adjacent reef consumers (Carreón-Palau et al. 2013, Briand et al. 2015) despite their apparent low labile
1059
organic matter content and nutritional quality (Granek et al. 2009).
1060
While common in the deep-sea, chemosynthesis has rarely been explored as a potential source of nutrient
1061
exchange for shallow water marine food webs (Table 3), yet SI data point to chemosynthesis being
1062
important for some consumer species on coral reefs. δ34S, together with δ13C and δ15N, provides evidence
1063
that chemosynthesis in lucinid clams supports 20% of the diet of Caribbean spiny lobster (Higgs et al. 2016).
1064
They therefore play an important part in transferring the chemosynthetically fixed carbon into reef food
1065
webs (Higgs et al. 2016). While there are undoubtedly many energetic connections between coral reefs and
1066
adjacent habitats, the intricacies of many have yet to be fully identified. This should be addressed going
1067
forward given that such subsidies are expected to contribute to wider ecosystem resilience and stability
1068
(Bascompte et al. 2005).
1069
Predators, typically being more mobile, often have greater opportunity to feed across ecosystem
1070
boundaries, playing an important ecological role connecting distinct food webs (Fig. 3). While tracking
1071
studies can identify these predator movements (e.g. Papastamatiou et al. 2009, Heupel and Simpfendorfer
1072
2015), they cannot in isolation define the role of predators in nutrient cycling. This is in contrast to SI data
1073
43
that offer an opportunity to track these energetic linkages. In Palmyra atoll, δ13C and δ15N from muscle
1074
tissue of blacktip reef sharks, Carcharhinus melanopterus, grey reef sharks, C. amblyrhynchos, and red
1075
snapper, Lutjanus bohar, indicated they relied on pelagic production sources from outside their primary
1076
forereef habitats, playing a key role providing ecological coupling as cross-system foragers (McCauley et al.
1077
2012). The reef manta ray, Manta alfredi, also constructs an important link between adjacent pelagic and
1078
reef/lagoonal systems (evidenced by δ13C and δ15N) by feeding on pelagic zooplankton, which is then
1079
excreted over shallow reefs (Peel et al. 2019). In fact many reef predators are similarly subsidised by pelagic
1080
inputs (evidenced by δ13C, δ15N, and δ34S), likely arising through their feeding on reef-based planktivorous
1081
fish, suggesting they play an important role in reef-pelagic connectivity (Frisch et al. 2014, Frisch et al.
1082
2016, Matley et al. 2018, Skinner et al. 2019b).
1083
Energetic connections across depth ranges have also been identified using δ13C, δ15N, and δ15NAA in
1084
combination with acoustic telemetry. Galapagos sharks, Carcharhinus galapagensis, and giant trevally,
1085
Caranx ignobilis, forage in both shallow and deep water mesophotic reef habitats, transporting nutrients
1086
between them (Papastamatiou et al. 2015). In some cases, these energetic linking movements may be
1087
nocturnal. δ13C and δ15N data combined with acoustic telemetry revealed that the grunt, Haemulon
1088
plumieri, in the Caribbean is predominantly sustained by organic matter from habitats it visited at night,
1089
likely to minimize encounters with barracuda (Rooker et al. 2018). These studies highlight the potential for
1090
SIs to complement and enhance inferences from acoustic telemetry or tracking data, by providing detailed
1091
information on energy fluxes which may help elucidate organism movements. Although SI studies
1092
documenting cross-system linkages are increasing, only 13% of the studies in this topic have focussed on
1093
sharks, despite their acknowledged important role in nutrient transfer (Williams et al. 2018b). There is
1094
therefore much that remains to be understood about energy subsidies to and from coral reefs, particularly
1095
the generality of the role that these mobile organisms play in establishing them.
1096
Production source use in coral reef consumers can be difficult to define compared to other, simpler
1097
systems with fewer energy pathways. Significant overlap between the myriad of reef-associated sources
1098
(Fig. 4; Table 3; and see 3. Organic matter dynamics) often makes it impossible to clearly identify for
1099
44
example carbon sources over a reef using bulk δ13C data. As such, isotope modelling to confidently assign
1100
consumers to energy sources in such ‘underdetermined’ systems becomes impractical (Fry 2013). The use
1101
of CSIA of amino acids (CSIA-AA), particularly essential amino acid carbon isotopes (δ13CEAA), which exhibit
1102
minimal discrimination between resources and consumers across the food web, has great potential to help
1103
resolve carbon resources over reefs due to the increased dimensionality of the data. This technique
1104
effectively separates benthic versus planktonic pathways, further separating the latter into distinct
1105
nearshore reef-associated plankton and offshore pelagic plankton groups (Skinner et al. 2021). Analysis of
1106
consumer tissue δ13CEAA allows the source of primary productivity supporting consumers to be identified
1107
with more precision than is possible with bulk δ13C. Less consideration of variations in isotopic baselines
1108
and TDFs is required when interpreting δ13CEAA data, which can often confound inferences from bulk SI
1109
data. Surprisingly, δ13CEAA data suggest that even highly mobile top predators may rely predominantly on a
1110
single carbon source at the base of the food web. For instance, in the Red Sea, δ13CEAA data suggested that
1111
the snapper, Lutjanus ehrenbergii, and giant moray, Gymnothorax javanicus, may receive >70 % of their C
1112
from a single end-member, phytoplankton, indicating quite tightly linked food chains supporting these
1113
predators (McMahon et al. 2016). Similarly, in the Maldives, several species of grouper were primarily
1114
supported by offshore pelagic plankton across an oceanic atoll (73-86%; Skinner et al. 2021). Contrasting
1115
δ13CEAA within more wide-ranging apex predators like the tiger shark, Galeocerdo cuvier, could provide
1116
important information on the role of mobility in integrating across ocean and reef food webs (Hilting et al.
1117
2013, Frisch et al. 2016).
1118
7. Environmental drivers
1119
7.1. Background
1120
It is intuitive following the identification of spatial and temporal patterns in data for scientists to try and
1121
elucidate the underlying processes and mechanisms that give rise to them. Despite the complexity of coral
1122
reefs, SI-based studies focusing on particular components of these ecosystems often infer the key drivers at
1123
play that underpin observed trends. This is evident in all of the preceding topics, where studies attribute
1124
45
the importance of inherent environmental and biological factors (see, for example, 3.3. Organic matter
1125
dynamics: Spatial and temporal variations and 4.4 Holobiont metabolism: Drivers of mixotrophy). However,
1126
there are a collective of SI studies that specifically investigate the general role spatial and temporal forcings
1127
play on coral reefs, providing context for other studies utilising SI approaches (n = 48 identified topics).
1128
These environmental drivers can be broadly separated into two categories: natural drivers and
1129
anthropogenic drivers (Fig. 3). The first category investigates how the isotopic composition of various
1130
components of coral reef ecosystems responds to a range of natural variables such as depth, resource
1131
availability, and salinity. Hard coral (n = 17) and reef fish (n = 5) are the dominant focal taxa in this regard,
1132
with most studies conducted in the Caribbean (n = 10), the Red Sea (n = 6), or the Central Pacific (n = 6) (Fig.
1133
2B). The latter category almost exclusively focusses on measuring SIs in structure forming organisms, that is
1134
marine plants (n = 9), soft corals (n = 6), and hard corals (n = 6), to explore anthropogenic nutrient inputs to
1135
coral reef communities but also considers the impacts of thermal stress. These are conducted mostly in the
1136
Caribbean (n = 7) or the western Pacific (n = 6) (Fig. 2B).
1137
7.2. Natural drivers
1138
Environmental conditions and nutrient availability fluctuate across both fine and broad spatial and
1139
temporal scales. Much like those of the holobionts (see 4. Holobiont metabolism), primary producer and
1140
higher consumer SI values may reflect this variability (Fig. 4; Table 3; Table S6), as organisms alter their
1141
resource use accordingly and/or integrate isotopic differences in production baselines. For example,
1142
gradients in resource availability exist across coral reef habitats depending on their proximity to open
1143
ocean resources (Wyatt et al. 2013). Primary producers have higher δ15N baseline values on outer reefs
1144
compared to lagoon regions (Page et al. 2013) and many reef fish are increasingly reliant on oceanic
1145
nutrients with proximity to the open ocean (Wyatt et al. 2012b, Gajdzik et al. 2016, McMahon et al. 2016).
1146
In New Caledonia, δ13C provided clear evidence of spatial changes in the primary sources of carbon across a
1147
reef habitat, with increased oceanic inputs on the outer slopes (grouped teleosts mean δ13C -17.9 )
1148
compared to increased internal subsidies in the lagoon (grouped teleosts mean -14.8 ) (Le Bourg et al.
1149
2017). Interestingly, δ15N data did not demonstrate any differences in food chain length or trophic level,
1150
46
supporting the idea of consistent food web structure across the reef scale despite differences in production
1151
sources supporting that structure (Le Bourg et al. 2017). At a regional spatial scale, Zgliczynski et al. (2019)
1152
similarly did not find strong isotopic evidence of shifts in foraging patterns in a range of species and
1153
functional groups across the central Pacific. Isotopic changes were attributed to baseline changes across
1154
the ~ 1000 km scale oceanographic gradient studied and taken as indicative of regional consistency in
1155
foraging (Zgliczynski et al. 2019). This demonstrates that coral reef food webs will often reflect fluctuations
1156
in resource availability and illustrates how SI data can be used to test ideas about spatial and temporal
1157
variation in key trophodynamic processes at different scales, and the factors likely to be driving these.
1158
If assessed in the context of well quantified temporal and spatial isotope baselines, SI data can
1159
demonstrate spatial changes in organic resource use that have significant implications for reef food web
1160
structure (Glass et al. 2020). Such changes can be exceedingly difficult to observe with traditional
1161
techniques such as feeding studies. As an example, bulk δ13C and δ15N data suggested a switch between
1162
planktivory and herbivory between fore and back reef environments for both nominally planktivorous and
1163
herbivorous species (Ho et al. 2009, Wyatt et al. 2012b). At Ningaloo Reef (western Australia), δ13C and
1164
δ15N modelling suggested that nominally herbivorous Stegastes spp derived over half their diet from the
1165
plankton in more oceanic habitats while planktivores increased herbivory in the back reef (Wyatt et al.
1166
2012b). On a reef flat in Okinawa (Japan), δ13C, δ15N, and fatty acid data showed that the algal gardening
1167
damselfish Stegastes nigricans supplemented its diet with unexpectedly large amounts of animal material
1168
for a herbivore (Hata and Umezawa 2011). This suggests that notions of resource reliance in reef
1169
consumers may need to be more flexible. Using δ13CEAA, McMahon et al. (2016) showed that while S.
1170
nigricans gained the majority (75%) of their dietary C from macroalgae, in more oceanic habitats 13%, and
1171
up to 30% for some individuals, came from phytoplankton. Importantly, the error on the estimates from
1172
the δ13CEAA data was much lower than from bulk SIA observations. It is not clear whether such changes
1173
reflect real spatial differences in actual food web structure however (McMahon et al. 2016), as SI data
1174
cannot distinguish between active consumption or incidental ingestion of particular sources that can be
1175
influenced by exposure rates and therefore spatial proximity and foraging strategies.
1176
47
Interactions between the complex topography of reefs and hydrodynamic processes that alter physical and
1177
biogeochemical conditions such as internal waves and upwelling can lead to highly variable isotope ratios in
1178
reef organisms across depth ranges at the same location. As light availability decreases with depth so,
1179
generally, does coral autotrophy; coral host tissue δ13C values thus become closer to oceanic carbon values
1180
through feeding (Muscatine et al. 1989, Gattuso et al. 1993, Lesser et al. 2010, Crandall et al. 2016). Coral
1181
skeletal δ13C values also reflect fluctuations in light and heterotrophy because they are influenced by
1182
metabolic fractionation; decreasing light and zooplankton levels resulted in significant decreases in coral
1183
skeleton δ13C (Grottoli and Wellington 1999, Grottoli 2000). Furthermore, both host and endosymbionts
1184
are depleted in 15N with depth, with impacts on endosymbiont growth rates (Muscatine and Kaplan 1994).
1185
This pattern of trophic zonation (i.e. where coral heterotrophy increases with depth) is not always
1186
consistent, however. Some evidence suggests that increasing depth does not necessarily result in increased
1187
heterotrophy as is typically expected (Einbinder et al. 2009), with observed changes in δ13C along depth
1188
axes potentially driven by internal carbon cycling processes once other biological traits are accounted for.
1189
Controlled feeding experiments of Reynaud et al. (2009) found no effect of ambient light levels on internal
1190
nutrient metabolism. They did however identify increased nitrogen cycling of host metabolic waste
1191
products to endosymbionts when heterotrophic feeding was limited by tracing bulk δ15N through different
1192
tissues. This is in contrast to carbon metabolism, which appears to change with irradiance levels when
1193
heterotrophic feeding occurs, with increasing rates of photosynthesis and carbon translocation from
1194
endosymbionts to coral host under high light conditions (Tremblay et al. 2014). At Palmyra Atoll, δ13C and
1195
δ15N data suggested that increases in coral heterotrophy with depth were absent at sites where resources
1196
were assumed to be readily available from extensive water mixing (Williams et al. 2018a). Similarly, in one
1197
of the few studies to use both δ13CAA and δ15NAA, while autotrophy was the dominant source of carbon to
1198
the hard coral Stylophora pistillata, heterotrophic energy contributions were equal across shallow (5 m)
1199
and deep (60 m) reefs (Martinez et al. 2020). Furthermore, SI data reveal that some species are
1200
heterotrophs throughout their depth ranges (Alamaru et al. 2009a, Crandall et al. 2016, Radice et al. 2019)
1201
or vary in their degree of heterotrophy across depths based on sampling region (Santos et al. 2021). Even
1202
48
within the same genus and at the same location, SI values (δ13C and δ18O) reveal that three species of
1203
Madracis differ in their depth adaptations and ecological plasticity (Maier et al. 2003), indicating that this
1204
phenomenon could depend on prey encounter, resource availability and resource acquisition (Maier et al.
1205
2010, Plass-Johnson et al. 2015b, Fox et al. 2018, Santos et al. 2021). Indeed, Leichter et al. (2007)
1206
suggested that depth gradients in reef macroalgal SI values can reflect gradients of exposure to offshore
1207
nutrient sources, such as increased use of deep-water nitrate by macroalgae exposed to high-frequency
1208
upwelling. Conversely, variability in SI values (δ13C, δ15N, and δ18O) in two abundant macroalgal species over
1209
scales of 10s of metres were perhaps independent of depth and instead reflected large amounts of spatial
1210
heterogeneity (Stokes et al. 2011). Clearly there are limited universal trends with depth that can be
1211
identified thus far on coral reefs, which suggests that biological traits and their plasticity may play a
1212
significant role buffering effects of physical drivers over a species’ observed distribution.
1213
CSIA can be especially powerful in examining variability in resource use across environmental gradients that
1214
could influence tissue SI signatures independent of diet changes. While primary producer bulk SI values
1215
vary across regions and seasons (Fig. 4; Table 3; Table S6), which can confound interpretation of dietary
1216
changes, CSIA measures individual compounds which are shaped by the biochemical processes of the
1217
primary producers and are thought not to vary spatially, temporally, or with growth rates (Whiteman et al.
1218
2019). Using CSIA, Papastamatiou et al. (2015) were able to discount the isotope baseline as a source of
1219
variation in the bulk SI data of giant trevally (Caranx ignobilis) from deep water reefs on a Pacific atoll;
1220
differences were due to individual trophic flexibility, with trophic positions determined with CSIA-AA
1221
ranging from 3.54.6. Acoustic tracking demonstrated individual variability in diel migration and feeding
1222
behaviour leading to a range of trophic positions, perhaps reflecting individual foraging preferences and
1223
intra-specific competition (Papastamatiou et al. 2015). By demonstrating changes in δ15NAA derived TP of
1224
spiny and slipper lobsters across large spatial scales in the north-western Hawaiian Islands, which were thus
1225
independent of baseline isotope variation, O’Malley et al. (2012) robustly demonstrated that spatial
1226
variability in growth was due to different responses between the two species to limited prey availability.
1227
Resource availability across spatial gradients thus may be a driver of consumer SI values. Of two obligate
1228
49
corallivore butterflyfish that both preferentially feed on Acropora coral, δ13CAA reveal that the specialist
1229
Chaetodon baronessa is more selective with depth and continually seeks out Acropora despite decreased
1230
availability, while the generalist Chaetodon octofasciatus becomes more flexible with depth (MacDonald et
1231
al. 2019). Furthermore, while planktivores have a consistent feeding strategy across shallow to mesophotic
1232
reefs, benthic invertivores and omnivores have significantly broader niches (with benthic invertivores also
1233
occupying a higher TP) (Bradley et al. 2016). These studies highlight the trophic versatility of many reef
1234
organisms along spatial gradients.
1235
7.3. Anthropogenic drivers
1236
Land and seascapes are under increasing pressure from human activities worldwide, and shallow coral reef
1237
ecosystems are particularly vulnerable to these threats (Fig. 3); coral bleaching events are now occurring
1238
every six years or less (Hughes et al. 2017). Bulk δ18O SI values can reveal when a coral has been subjected
1239
to extreme thermal stress by directly relating the δ18O values to in situ temperature (Porter et al. 1989,
1240
Mayal et al. 2009). SIs can also identify how the coral animal host and endosymbionts alter their trophic
1241
strategies and resource use in response to bleaching events (see 4.4.1. Holobiont metabolism: Exogenous
1242
factors). Over the longer term, pulse-chase experiments using 13C-labelled bicarbonate revealed that some
1243
corals may increase their use of heterotrophic carbon for up to a year after bleaching, but it is not known
1244
whether this is a sign of resilience or prolonged stress (Hughes and Grottoli 2013). Generally, it seems that
1245
different coral species have different bleaching responses, with some maintaining energy reserves or
1246
heterotrophic capacity, but most recovering within a year if the bleaching is a mild and isolated event
1247
(Grottoli et al. 2017, Levas et al. 2018). There is currently little understanding of how coral and
1248
endosymbionts trophic strategies will change through successive bleaching events, despite predictions that
1249
these will occur ever more frequently. Furthermore, studies assessing thermal stress effects using SIs have
1250
focussed on corals, with few, if any, investigating these impacts on other holobionts. For non-symbiotic
1251
organisms, Vaughan et al. (2021) suggested increases in δ15N (~ 1 ‰) in natural and transplanted
1252
macroalgae, Sargassum mangarevense, were linked to the release of coral-derived nutrients post-
1253
bleaching. While a better understanding of SI changes in reef biogeochemical cycles post-bleaching is
1254
50
required (but see Radice et al. 2021), it is crucial to understand and measure the underlying natural
1255
variation in SI source data when making inferences, i.e. by measuring the underlying δ15N baseline.
1256
Global bleaching events and declines in live coral cover have been linked to declines in structural
1257
complexity, coral biodiversity, and the abundance and diversity of reef-associated fishes (Jones et al. 2004,
1258
Carpenter et al. 2008, Pratchett et al. 2018). Overfishing also contributes to the latter, while simultaneously
1259
reducing fish-mediated storage and supply of nutrients by up to 50% (Allgeier et al. 2016). These drivers
1260
may be reflected in consumer SI values. For example, as the amount of rubble increases along a habitat
1261
disturbance gradient, there are significant differences in the δ13C and δ15N ranges and the isotopic niche
1262
area of the parrotfish Chlororus bleekeri, the foraging of which varies according to the surrounding habitat
1263
(Plass-Johnson et al. 2018). Similarly, as prey densities fluctuate in response to declines in structural
1264
complexity, the peacock grouper Cephalopholis argus maintains their TP (as indicated by their δ15N) by
1265
switching foraging modes from ambush to widely active foraging (Karkarey et al. 2017). Loss of habitat
1266
structural complexity is predicted to cause a 3-fold reduction in fishery productivity (Rogers et al. 2014).
1267
However, some species such as the peacock grouper may be more resilient to habitat loss than previously
1268
thought as their foraging plasticity may enable them to adapt to coral degradation (Karkarey et al. 2017). In
1269
addition to habitat loss, ocean acidification (i.e. reductions in oceanic pH through elevations in partial
1270
pressure of seawater CO2 (pCO2) due to increasing global carbon emissions) may negatively impact
1271
calcifying marine organisms. By studying coral skeleton δ13C and δ18O values, Zhou et al. (2016) determined
1272
that Acropora gemmifera photosynthesis and calcification were only impaired at the highest pCO2
1273
treatment, with their microbial communities remaining stable. More recently, benthic communities were
1274
exposed to increased pCO2 and warming in mesocosms to determine how trophic architecture (represented
1275
by organism δ13C and δ15N) would respond under future scenarios (Nagelkerken et al. 2020). While trophic
1276
pyramids and community structure (i.e. biomass and productivity) shifted, the food web architecture
1277
remained inflexible and stabilising processes were absent suggesting a lack of adaptive capacity in the
1278
ecosystem (Nagelkerken et al. 2020). SIs therefore offer a unique opportunity to study potential impacts of
1279
51
changing environmental conditions on organism and whole food web trophic ecology (Plass-Johnson et al.
1280
2018).
1281
SI data offer great potential for exploring other anthropogenic impacts on coral reef food webs, notably the
1282
role of human-derived pollutants in reef systems. By measuring SI values of various reef organisms, these
1283
nutrient inputs can be identified to determine the extent of their impact on the local reef communities
1284
(Yamamuro et al. 2003). Elevated δ15N values in sampled organisms have suggested artificial nutrient inputs
1285
to reef communities pertaining to aquaculture (Herbeck et al. 2014), shoreline sewage (Todd et al. 2009,
1286
Baker et al. 2017, Abaya et al. 2018, Lachs et al. 2019), stormwater discharge (Lapointe and Bedford 2011),
1287
seepage water (Mwaura et al. 2017), river plume pollution (Risk et al. 2014), and even wood pulp effluent
1288
(Schleyer et al. 2006). This highlights how spatially explicit context is required to determine the role of
1289
pollution in nutrient cycling of nearshore reefs and flow on effects for other taxa like coral (Umezawa et al.
1290
2002, Huang et al. 2013, Adam et al. 2021). While most of these studies use macroalgae as bio-indicators of
1291
pollution, seagrass δ15N has also been proposed as a tool to monitor time-integrated changes over coral
1292
reef habitats, as there are fewer seagrass species than the diverse macroalgae that are found across reefs,
1293
facilitating both identification and standardisation (Yamamuro et al. 2003). Indirect incorporation of
1294
anthropogenic inputs can result from connectivity to the pelagic food chain and reef planktivore grazing.
1295
Lower δ13C and higher δ15N in two damselfishes demonstrated how POM released from nearby fish farms in
1296
southern Taiwan can enter coral reef food webs (Jan et al. 2014). One of the few studies to employ sulphur
1297
isotopes to examine anthropogenic inputs attributed decreasing δ34S in coral skeletons off Yoron Island to
1298
rain-driven inputs of low δ34S terrestrial material from sugar cane dominated farmland (Otani and
1299
Nakanishi 2019). Oxygen isotopes may also be useful in anthropogenic impact monitoring, with shifts in
1300
δ13C and δ18O in coral skeleton correlating with local oil spills (Xu et al. 2018). However, studies need to be
1301
cautious when using taxon-specific variations in tissue isotopes to infer nutrient fluxes based on SI data,
1302
especially across small ranges that might be explained by metabolic or hydrodynamic variations (see 7.2
1303
Natural drivers), which is especially the case for coral skeleton. Relationships between anthropogenic
1304
nutrient inputs and SI data, and subsequent ecological effects may be more nuanced. For example, a long-
1305
52
term experiment involving δ13C and δ15N in the hard corals Acropora palmata and Porites porites showed
1306
that moderate doses of anthropogenic nutrients had no additional effects and the corals continued growing
1307
(Allgeier et al. 2020). However, models revealed that nutrient and carbon flows were dominated by the
1308
symbiont, leading to algal dominance in the holobiont and greater algal demand on coral resources, likely
1309
increasing the corals’ future vulnerability to bleaching due to stress (Allgeier et al. 2020).
1310
It should be emphasised that elevated δ15N values alone are insufficient evidence of anthropogenic
1311
impacts over reef ecosystems. There is an extensive range in natural isotope abundances in the ocean in
1312
the absence of anthropogenic inputs; isotope effects across the marine nitrogen cycle span approximately -
1313
0.5 ‰ to +38 ‰ (Sigman et al. 2009). Increased macroalgal δ15N with depth can be indicative of upwelling
1314
increasing dissolved nitrogen availability, rather than a sewage impact (Huang et al. 2013). Instead of
1315
anthropogenic sources (e.g. Lapointe 1997, Lapointe et al. 2005), POM isotopic variations may reflect
1316
mineralization of organic material and nitrification along with inputs of DIN from upwelling, runoff,
1317
sediments, and the atmosphere (Lamb and Swart 2008). Natural N cycling and resultant isotopic variation
1318
could account for δ15N variations in benthic components of the reef. For instance, Lapointe et al. (2005)
1319
reported algal δ15N elevated by +2 ‰ as indicative of land-based pollution, however the range in POM δ15N
1320
across the Florida Keys varies over a 20 ‰ range with a standard deviation of ± 3.6 ‰, apparently
1321
independent of human influences (Lamb and Swart 2008). This underscores the importance of
1322
understanding spatial and temporal variations in potential source isotopes (Fig. 4; Table 3), due to e.g.
1323
water column sources and upwelling (Leichter et al. 2007), before invoking anthropogenic perturbation.
1324
This is especially important where the variations are small (e.g. a few per mil or less) relative to a potential
1325
natural range of SI data variations.
1326
Due to their incremental growth, corals, and in particular gorgonians, provide an opportunity to investigate
1327
long-term trends in anthropogenic nutrient inputs. The biochemistry of skeletal banding reflects the
1328
ambient nutrient levels and can therefore be used to infer long-term pollution trends (Ward-Paige et al.
1329
2005). Indeed, gorgonian δ15N values accurately reflect tourism levels over multiple years, with declines in
1330
δ15N linked to declines in tourism and increases in δ15N linked to its recovery (Baker et al. 2013b). Hard
1331
53
coral skeletons can track anthropogenic nitrogen fluctuations over decades with 15N enrichment linked to
1332
increasing sewage levels and population density (Duprey et al. 2017, Duprey et al. 2020). Archived museum
1333
samples may provide another means to monitor human pollution levels over longer time spans. Gorgonian
1334
samples from a 143 year time span had δ13C and δ15N values which reflected increasing atmospheric CO2
1335
and use of agricultural fertilizers respectively (Baker et al. 2010), while samples in Bermuda spanning 50
1336
years reflected changes in management policies which were effective in reducing local pollution levels
1337
(Baker et al. 2017). Despite the wealth of information they can convey on prior isotopic baselines, the use
1338
of archival samples is exceedingly low across SI reef studies (n ≈ 3).
1339
8. Knowledge gaps, caveats, limitations, and future directions
1340
Throughout this review we have summarised the key findings of studies that have used SIs to understand
1341
coral reef ecosystems. In doing so, we have highlighted that there are still areas where research,
1342
knowledge, and understanding are lacking. This should be taken as an opportunity to focus future research.
1343
Below, we have included a numbered list of identified gaps and opportunities for future SI studies. Given
1344
the logistical ease for researchers to now obtain and analyse SI data, we also emphasise why care must be
1345
taken not to misuse or overinterpret SI data. The Knowledge gaps and opportunities section is followed by
1346
important caveats and limitations which must be considered. We hope that these points serve as a
1347
roadmap to direct future research involving SIs on coral reefs.
1348
8.1 Knowledge gaps and opportunities
1349
1. Improve characterisation of basal sources. Corals have been the focus of much SI research due to their
1350
foundational importance, but the living proportion of whole reefs that they cover is generally low.
1351
Conversely, turf algae’ cover a greater surface area and are intensively grazed, but there is sparse
1352
understanding of the fate of turf-derived fluxes in the ecosystem. Similarly, substantial fluxes of detritus
1353
derive from benthic reef sources both directly (e.g. exudates) and secondarily (e.g. grazer defecation),
1354
but the detrital pool, its origins and fluxes are little resolved. SI data, both bulk and CSIA, have a lot to
1355
54
offer well-focussed investigations which can advance our understanding of these lesser studied fluxes
1356
on reefs.
1357
2. Extend sampling over longer time frames. Resource availability on coral reefs fluctuates through time
1358
and consumer resource use may vary accordingly. However, to date, many coral reef SI studies sample
1359
during a single time point or limited temporal window (Fig. 5A). Future studies exploring the dynamics
1360
and stability of coral reef ecosystems should consider longer time periods over which inference can be
1361
made. Multi-tissue approaches are one technique which may reveal dietary variations across a range of
1362
timescales from a single individual, yet currently, only 9% of studies measured multiple tissues within
1363
the same individual organism. Employing a multi-tissue approach by sampling faster turnover tissues
1364
e.g. organs and blood, in addition to slower turnover tissues e.g. muscle tissue, could better discriminate
1365
dietary consistency or the lack thereof (Wyatt et al. 2019).
1366
3. Employ an isoscape approach. To date, and likely to due to their inherent complexity, there are few
1367
studies employing an isoscape approach on coral reefs. However, isoscapes of coral reef ecosystems
1368
could offer valuable insight into the complex processes that might influence coral reef trophodynamics
1369
and SI values across various locations (Fig. 4; Table 3). For example, spatial variations in the relative
1370
importance of different production sources to reef food webs are evident. Given that environmental
1371
conditions are fluctuating due to climate change, there is a need for studies examining the drivers of
1372
these likely differences in coral reef ecosystems, which could be facilitated by a broadening of the
1373
spatial scale across which studies are conducted. Isoscapes provide a solid isotopic foundation of a
1374
region, providing the context upon which other avenues of research can be conducted (McMahon et al.
1375
2013).
1376
4. Explore trophodynamics across the wider community. SI data have helped to highlight that categorising
1377
reef consumers into simple trophic groups masks dietary specialisations and the complexities of this
1378
dynamic system. However, few studies compare niche dynamics and resource partitioning outside of
1379
species groups or guilds. For example, most studies that focus on reef fish do not consider how
1380
invertebrates occupying the same TP may share similar resources (but see Zapata-Hernández et al.
1381
2021). A better understanding of trophic interactions across the wider reef community is required. This
1382
55
will help understand the importance of potential competitiveness among disparate taxa on coral reefs.
1383
Given increased feasibility, this body of knowledge is expected to grow, and it will be important for
1384
greater resolution to be assimilated into whole ecosystems models.
1385
5. Quantify species-specific TDF and tissue turnover rates. To date, few studies have accurately quantified
1386
species-specific TDF and isotopic turnover rates (i.e. the time it takes for a given consumer tissue to
1387
reflect the isotopic composition of its diet) across different tissue types for coral reef consumers. While
1388
this will be challenging to determine for species with longer turnover times (e.g. hard coral; Tanaka et al.
1389
2008), it is highly recommended that more laboratory feeding studies be conducted. Where
1390
environment- and species-specific TDF are not available, analyses must rely on TDF across a range of
1391
values to account for uncertainty surrounding the inherent variability (e.g. likely ≥ 1 ‰ SD for Δ15N).
1392
6. Consider parasites. Parasites may be involved in >50% of food web links but they are rarely considered
1393
in food web science (Dunne et al. 2013). Indeed, on coral reefs, there are estimated to be ten times
1394
more parasite species than fish species (Justine et al. 2012). In addition to their huge biodiversity,
1395
parasites can affect food web structure indirectly by modifying their host’s behaviour and subsequent
1396
resource use (Welicky et al. 2017). Despite their important ecological role shaping trophic interactions in
1397
coral reef ecosystems, parasite trophodynamics on reefs are poorly understood. While several more
1398
recent studies are beginning to address this knowledge gap (e.g. Jenkins et al. 2018, Jenkins et al. 2020,
1399
Riekenberg et al. 2021), few have considered parasite life stage dietary shifts.
1400
7. Measure other isotopes. Despite limited application on coral reefs to date, bulk δ34S shows promise for
1401
distinguishing among reef production sources and revealing consumer habitat use (Connolly et al. 2004,
1402
Granek et al. 2009, Skinner et al. 2019a). Moreover, recent advances in SI technology mean that δ13C,
1403
δ15N, and δ34S can now be measured from the same sample aliquot with a high level of precision (Fourel
1404
et al. 2015). Similarly, to our knowledge, there are currently no coral reef SI studies that utilise mercury
1405
isotopes (δ202Hg), which can characterize resource partitioning and identify foraging depth separation
1406
among predators in marine environments (Besnard et al. 2021). Given the myriad of co-occurring
1407
predators on coral reefs, δ202Hg may offer a useful tool to understand resource partitioning across
1408
depths. Finally, copper (δ65Cu) and zinc (δ66Zn) isotopes have recently been suggested as a proxy for
1409
56
coral stress, as both increase in coral host and symbiont tissue at higher temperatures (Ferrier-Pagès et
1410
al. 2018). Given the difficulty in disentangling the production sources and consumer resource use on
1411
reefs, studies incorporating such additional isotopes are strongly encouraged.
1412
8. Apply compound-specific SIA (CSIA). The number of studies employing CSIA on coral reefs is rapidly
1413
increasing (of coral reef SI studies >5 in 2020 alone, compared to 15 between 1982 2019). This
1414
advance will certainly offer new insights into the complexities of the coral reef ecosystem. Although
1415
multiple isotope approaches show significant potential for better elucidation of organic matter fluxes
1416
and resource use in this complex ecosystem, few studies combine CSIA of both carbon and nitrogen in
1417
the compounds of interest. Given the depth of information they can provide on production sources and
1418
trophic interactions, studies combining both are likely to provide a more holistic understanding of the
1419
specific systems of interest.
1420
9. Incorporate and explore relationships with environmental drivers. Laboratory studies generally focus
1421
on manipulating one aspect of the environment, while field studies use SIs as indicators rather than
1422
testing what drives changes in their values. Few studies consider the effect of multiple environmental
1423
drivers, whether natural or anthropogenic, or the interactions occurring between them. Measuring
1424
multiple environmental variables (e.g. nutrients, pH, temperature, salinity etc) and linking those with SI
1425
values is a multi-method approach which could better explain spatial gradients (of human impacts) on
1426
reef ecosystems (Teichberg et al. 2018).
1427
10. Utilise archival samples. Analysis of archival samples, held in museums and other institutions across the
1428
globe, can provide additional information on prior isotopic baselines and how they have shifted over
1429
time. Soft and hard corals are common, slow growing, benthic organisms, which may comprise
1430
important baselines against which current pollution levels can be monitored. This gives the historical
1431
context against which the significance of contemporary isotope values can be compared. Archival
1432
specimen preservation is a considerable cause for concern, however improved mass spectrometer
1433
technology means that very little material is physically required to obtain SI data. Where possible, we
1434
encourage researchers to incorporate archival or historical baseline samples. Such approaches are
1435
particularly powerful when drawing conclusions about current anthropogenic impacts.
1436
57
11. Conduct multiple analyses on the same sample material. There are logistical efforts involved with
1437
collecting samples on coral reefs and conservation concerns involved with intensive sampling, especially
1438
of threatened species. As such, we advocate for multiple analyses on the same sample material. Not
1439
only would this facilitate more collaboration between different regions (which at present tend to
1440
specialise on particular fields of research), but it would maximize the information derived from each
1441
individual sample.
1442
12. Adopt a multi-regional, aggregative research approach. There are now many SI studies (and therefore
1443
data) conducted on coral reefs, but almost all are isolated to a single area. Further still, there are
1444
regional disparities in research focus. Here, it is worth acknowledging that collecting SI data often
1445
requires extensive sampling to which there may be substantial limitations, particularly in coral reef
1446
regions (i.e. lack of access, funding, infrastructure, requiring specialist vessels or equipment, sampling
1447
during extreme conditions to assess seasonality). Regardless, generality at the regional to global scale is
1448
currently lacking. Moreover, inferences from different studies can be contradictory (e.g. see 5.2 Trophic
1449
niches: Isotopic Niches). Given the number of studies now published, a more aggregative research
1450
approach (i.e. meta-analyses) may help discern ubiquity in patterns and drivers of these.
1451
Complementary to this, we advocate for larger, multi-region studies to detect, if any, general trends in
1452
SIs (e.g. variability in production sources cf. Fig. 4) on coral reefs to act as ecological baselines. This
1453
could help generate mechanistic theory of overall ‘typical’ coral reef SI functioning. Deviations from this
1454
may point to interesting but currently not-yet-realised avenues of further study.
1455
8.2 Caveats and limitations
1456
1. Provide enough detail for SI data to be useful to others. Researchers should strive to clearly report
1457
their sampling design and SI data so it can be used by others. Over 15% of identified articles did not
1458
report when or over what period samples were taken (Fig. 5A), despite the high spatiotemporal
1459
variability of SI data on coral reefs. Furthermore, when extracting SI baseline data (i.e. Table 3), we
1460
found that many articles only published data in figures and/or did not provide sample sizes that are
1461
58
required to understand data spread and error. More rigorous reporting will facilitate point 12 above,
1462
allowing researchers to employ an aggregative approach to future coral reef SI studies.
1463
2. Interpret SI data within the context of underlying SI baselines. Variation in SI baselines (values at the
1464
base of the food web) must be considered when assessing variation in resource- and nutrient-use using
1465
isotopes. There is a tendency to extrapolate from taxon-specific variations in tissue SI data to infer
1466
biogeochemical rates or anthropogenic impacts, which is especially concerning across small ranges in
1467
isotope values. These variations may be within ranges of a few permille or less, that alone can be
1468
explained by discrimination and metabolic variations or simply to fluctuations in isotopic baselines. This
1469
presents a high risk of overinterpretation, especially given the dynamic fluxes present on coral reefs.
1470
Understanding natural variation in SI values is also needed when analysing the influence of stochastic
1471
events (e.g. coral spawning, coral bleaching) with SIs, especially when SI perturbations are small relative
1472
to the magnitude of the natural variance in SI baselines (see Fig. 4). We urge researchers to take careful
1473
consideration when interpreting spatial and temporal heterogeneity in reef organism SI values and to
1474
measure isotopic baselines wherever possible. This might involve comprehensive source sampling for
1475
bulk SIA or the application of more source specific CSIA. Complementary data, e.g. acoustic telemetry or
1476
nutrient concentrations, and data simulations prior to SIA (to estimate the likely impact e.g. TDF
1477
variability could have on observed data) could also be hugely beneficial for considering SI variation in
1478
context. In some cases, it may not be possible to measure baseline source values in situ. As a first
1479
estimate, researchers should consider the source SI values presented herein (Fig. 4 and Table 3 for
1480
interstudy means; Table S6 for study-specific values). The presented values integrate differences
1481
attributable to spatiotemporal dynamics, biodiversity, and inter-laboratory differences, highlighting the
1482
variability that has been observed across studies. For example, macroalgae SI values show high variance
1483
(mean ± s.d. for δ13C = -15.68 ± 4.87 and δ15N = 3.93 ± 3.45; Table 3), especially between species, and, as
1484
such, authors should consider some form of abundance weighting by species.
1485
3. Isotopic niche ≠ trophic niche. While variability in SI data within a population may indicate there are
1486
individuals with consistent differences in trophic ecology, low variability in SI data in a population does
1487
not necessarily indicate a narrow trophic niche. In the latter case, because of the time integrating
1488
59
character of the SI signatures, there are two possibilities: 1) individuals could be constantly feeding on
1489
one production source or 2) they could be feeding on different sources in proportions or combinations
1490
that happen to integrate to the same SI value. Specifically, how equivalent is the isotopic niche to the
1491
trophic niche? What is the magnitude of isotopic variation expected from processes independent of
1492
feeding variation (e.g. growth rates, diet quality)? Given the diversity of production sources available on
1493
coral reefs, these questions and the discrepancy between the trophic niche and the isotopic niche may
1494
be more of a problem in this ecosystem compared to others with less diversity in available sources for
1495
consumers. Consequently, care must be taken when making inferences regarding trophic niches based
1496
on isotopic niches on coral reefs.
1497
9. Conclusions
1498
SIA are an important tool that have elucidated many of the diverse and complex processes and
1499
relationships occurring on coral reef ecosystems. By combining a traditional literature search of databases
1500
with topic modelling of article abstracts, we identified recurring patterns and themes in the SI coral reef
1501
studies published to date. Summarising how SIs have advanced our understanding of coral reefs is
1502
challenging due to the inherent cross-over between studies, but the topic modelling approach partitioned
1503
the article text data to generate non-biased categories, providing a clear guide on the most logical structure
1504
for the review.
1505
One of the fundamental components of SI advances for coral reef ecology, and also one of the challenges,
1506
involves identifying the available energy fluxes to reef food webs and how these vary across different
1507
scales. SIA have been used to successfully identify and quantify inputs to reefs, and releases of material
1508
from reef organisms, showing that both are closely linked to the spatial arrangement of organisms and
1509
hydrodynamics across reefs. Fluctuations in available resources, related to the structure and layout of the
1510
seascape, are reflected in reef primary producer and consumer SI values. This is highlighted by the global
1511
variability in reported SI baselines across the literature (Table 3, Fig. 4), emphasising the importance of
1512
considering variations in when studying trophodynamics across coral reef food webs.
1513
60
Corals represent one of the most studied organisms on coral reefs due to their instrumental role as
1514
ecosystem engineers. SI studies have revealed just how complex their nutrient uptake and feeding
1515
strategies are, often with stark contrasts between species and little consistency in their reliance on auto- or
1516
hetero-trophic resources spatially and temporally. SI studies have also highlighted the importance of
1517
previously underappreciated holobiont groups to the overall coral reef ecosystem, notably sponges that
1518
underpin diverse pathways of in situ DOM and POM recycling. The myriad of interactions occurring
1519
between the vast numbers of reef primary producers and consumers is challenging to explore, but SI data
1520
has begun to disentangle some of these. For example, many isotopic niche studies have shown that
1521
traditional dietary classifications mask individual-level variations in resource use. They suggest that many
1522
reef organisms have more varying and flexible functional roles than those inferred from traditional
1523
techniques. Such insight also points to the fact that a better understanding of wider feeding relationships
1524
across guilds, not only within guilds, is required, particularly among invertebrates.
1525
Coral reef ecosystems do not persist in isolation, and SI studies demonstrate energetic linkages with
1526
adjacent habitats. These linkages are formed not only from ontogenetic shifts in habitat use of reef fishes,
1527
from their nursery grounds in mangroves and seagrass beds to the reef, but also through the movement
1528
patterns of larger, mobile predators. SIs also reveal vertical movement patterns across depth gradients,
1529
highlighting the extent of connectivity between shallow and deeper reefs. Connectivity is not only
1530
determined by energy flows however, with SIs providing a methodological means of transgenerational
1531
tagging to follow larval distributions, highlighting the diversity of research questions that can be addressed
1532
using SI techniques.
1533
SIA have become an increasingly important tool for exploring anthropogenic impacts on coral reef food
1534
webs; they can highlight nutrient inputs, ocean acidification effects, and thermal stress, although care is
1535
needed not to conflate such effects with natural SI baseline variations. Climate change is increasing global
1536
water temperatures, resulting in catastrophic bleaching of coral reefs worldwide. Such a loss of live coral
1537
can impact associated food webs by reducing biodiversity and degrading habitats, resulting in lower trophic
1538
complexity across the community (Gabara et al. 2021), as well as reduced ecological stability due to loss of
1539
61
functional redundancy. SIA represents a crucial tool to increase our understanding of the complex trophic
1540
interactions occurring on coral reefs that are modulated by environmental drivers and their associated
1541
dynamics, including human induced climate change.
1542
The number of SI reef studies has been increasing rapidly in recent years. By objectively drawing on the
1543
published literature, we have synthesised the current knowledge and understanding acquired through the
1544
application of SIA on coral reefs into five broad bodies of research focus. In doing so, we have highlighted
1545
potential research avenues that warrant further exploration, including increasing the scale at which SI
1546
studies are conducted (both through time and across space) in the hope of identifying more general
1547
patterns and processes that underpin coral reef structure and functionality. While acknowledging the
1548
considerations that need to be made when utilising SIs, we hope this review acts as a useful synthesis, and
1549
will serve to bolster the expanding literature on ecological applications of SI approaches to coral reef
1550
ecosystems.
1551
10. Acknowledgements
1552
CS and ASJW were supported by funding from the Hong Kong Branch of the Southern Marine Science and
1553
Engineering Guangdong Laboratory (Guangzhou) (SMSEGL20SC01) and the Research Grants Council (RGC)
1554
of Hong Kong (RGC Project No. 26100120). MRDC was supported by the NERC-BMBF CAO Coldfish project
1555
(NE/R012520/1) at Newcastle University and was funded by Irish Research Council Laureate Award
1556
IRCLA/2017/186 to Andrew L Jackson, Trinity College Dublin. We thank Dr Veronica Radice, one other
1557
anonymous reviewer, and the editor, Prof Peter Mumby, for their constructive comments.
1558
11. References
1559
Abaya, L. M., T. N. Wiegner, S. L. Colbert, J. P. Beets, K. M. Carlson, K. L. Kramer, R. Most, and C. S. Couch.
1560
2018. A multi-indicator approach for identifying shoreline sewage pollution hotspots adjacent to
1561
coral reefs. Mar Pollut Bull 129:70-80.
1562
Achlatis, M., M. Pernice, K. Green, J. M. de Goeij, P. Guagliardo, M. R. Kilburn, O. Hoegh-Guldberg, and S.
1563
Dove. 2019. Single-cell visualization indicates direct role of sponge host in uptake of dissolved
1564
organic matter. Proc Biol Sci 286:20192153.
1565
62
Achlatis, M., M. Pernice, K. Green, P. Guagliardo, M. R. Kilburn, O. Hoegh-Guldberg, and S. Dove. 2018.
1566
Single-cell measurement of ammonium and bicarbonate uptake within a photosymbiotic
1567
bioeroding sponge. The ISME Journal 12:1308-1318.
1568
Adam, T. C., D. E. Burkepile, S. J. Holbrook, R. C. Carpenter, J. Claudet, C. Loiseau, L. Thiault, A. J. Brooks, L.
1569
Washburn, and R. J. Schmitt. 2021. Landscape-scale patterns of nutrient enrichment in a coral reef
1570
ecosystem: implications for coral to algae phase shifts. Ecol Appl 31:e2227.
1571
Akhand, A., K. Watanabe, A. Chanda, T. Tokoro, K. Chakraborty, H. Moki, T. Tanaya, J. Ghosh, and T. Kuwae.
1572
2021. Lateral carbon fluxes and CO2 evasion from a subtropical mangrove-seagrass-coral
1573
continuum. Sci Total Environ 752:142190.
1574
Alamaru, A., Y. Loya, E. Brokovich, R. Yam, and A. Shemesh. 2009a. Carbon and nitrogen utilization in two
1575
species of Red Sea corals along a depth gradient: Insights from stable isotope analysis of total
1576
organic material and lipids. Geochimica et Cosmochimica Acta 73:5333-5342.
1577
Alamaru, A., R. Yam, A. Shemesh, and Y. Loya. 2009b. Trophic biology of Stylophora pistillata larvae:
1578
evidence from stable isotope analysis. Marine Ecology Progress Series 383:85-94.
1579
Allgeier, J. E., M. A. Andskog, E. Hensel, R. Appaldo, C. Layman, and D. W. Kemp. 2020. Rewiring coral:
1580
Anthropogenic nutrients shift diverse coral-symbiont nutrient and carbon interactions toward
1581
symbiotic algal dominance. Glob Chang Biol 26:5588-5601.
1582
Allgeier, J. E., C. A. Layman, P. J. Mumby, and A. D. Rosemond. 2014. Consistent nutrient storage and supply
1583
mediated by diverse fish communities in coral reef ecosystems. Global Change Biology 20:2459-
1584
2472.
1585
Allgeier, J. E., A. Valdivia, C. Cox, and C. A. Layman. 2016. Fishing down nutrients on coral reefs. Nat
1586
Commun 7:12461.
1587
Andrews, J. C., and P. Gentien. 1982. Upwelling as a source of nutrients to the Great Barrier Reef
1588
ecosystem: A solution to Darwin's question? Marine Ecology Progress Series 8:257-269.
1589
Araújo, M. S., D. I. Bolnick, and C. A. Layman. 2011. The ecological causes of individual specialisation. Ecol
1590
Lett 14:948-958.
1591
Araújo, M. S., D. I. Bolnick, G. Machado, A. A. Giaretta, and S. F. dos Reis. 2007. Using δ13C stable isotopes to
1592
quantify individual-level diet variation. Oecologia 152:643-654.
1593
Arias-González, J. E., B. Delesalle, B. Salvat, and R. Galzin. 1997. Trophic functioning of the Tiahura reef
1594
sector, Moorea Island, French Polynesia. Coral Reefs 16:231-246.
1595
Ashworth, E. C., M. Depczynski, T. H. Holmes, and S. K. Wilson. 2014. Quantitative diet analysis of four
1596
mesopredators from a coral reef. J Fish Biol 84:1031-1045.
1597
Atkinson, M. J. 1992. Productivity of the Enewetak flats predicted from mass transfer relationships.
1598
Continental Shelf Research 12:799-807.
1599
Atkinson, M. J. 2011. Biogeochemistry of Nutrients. Pages 199-206 in Z. Dubinsky and N. Stambler, editors.
1600
Coral Reefs: An Ecosystem in Transition. Springer Netherlands, Dordrecht.
1601
Baker, D. M., J. P. Andras, A. G. Jordán-Garza, and M. L. Fogel. 2013a. Nitrate competition in a coral
1602
symbiosis varies with temperature among Symbiodinium clades. The ISME Journal 7:1248-1251.
1603
63
Baker, D. M., K. Kim, J. P. Andras, and J. P. Sparks. 2011. Light-mediated 15N fractionation in Caribbean
1604
gorgonian octocorals: implications for pollution monitoring. Coral Reefs 30:709-717.
1605
Baker, D. M., T. J. T. Murdoch, I. Conti-Jerpe, and M. Fogel. 2017. Investigating Bermuda's pollution history
1606
through stable isotope analyses of modern and museum-held gorgonian corals. Mar Pollut Bull
1607
114:169-175.
1608
Baker, D. M., R. E. Rodríguez-Martínez, and M. L. Fogel. 2013b. Tourism’s nitrogen footprint on a
1609
Mesoamerican coral reef. Coral Reefs 32:691-699.
1610
Baker, D. M., K. L. Webster, and K. Kim. 2010. Caribbean octocorals record changing carbon and nitrogen
1611
sources from 1862 to 2005. Global Change Biology 16:2701-2710.
1612
Barragan-Barrera, D. C., A. Luna-Acosta, L. J. May-Collado, C. J. Polo-Silva, F. G. Riet-Sapriza, P. Bustamante,
1613
M. P. Hernandez-Avila, N. Velez, N. Farias-Curtidor, and S. Caballero. 2019. Foraging habits and
1614
levels of mercury in a resident population of bottlenose dolphins (Tursiops truncatus) in Bocas del
1615
Toro Archipelago, Caribbean Sea, Panama. Mar Pollut Bull 145:343-356.
1616
Barrón, C., and C. M. Duarte. 2015. Dissolved organic carbon pools and export from the coastal ocean.
1617
Global Biogeochemical Cycles 29:1725-1738.
1618
Bascompte, J., C. J. Melian, and E. Sala. 2005. Interaction strength combinations and the overfishing of a
1619
marine food web. Proc Natl Acad Sci U S A 102:5443-5447.
1620
Baumann, J., A. G. Grottoli, A. D. Hughes, and Y. Matsui. 2014. Photoautotrophic and heterotrophic carbon
1621
in bleached and non-bleached coral lipid acquisition and storage. Journal of Experimental Marine
1622
Biology and Ecology 461:469-478.
1623
Bearhop, S., C. E. Adams, S. Waldron, R. A. Fuller, and H. Macleod. 2004. Determining trophic niche width: a
1624
novel approach using stable isotope analysis. Journal of Animal Ecology 73:1007-1012.
1625
Bednarz, V. N., R. Grover, J.-F. Maguer, M. Fine, C. Ferrier-Pagès, and G. Capone Douglas. 2017. The
1626
Assimilation of Diazotroph-Derived Nitrogen by Scleractinian Corals Depends on Their Metabolic
1627
Status. mBio 8:e02058-02016.
1628
Bellwood, D. R., R. P. Streit, S. J. Brandl, and S. B. Tebbett. 2019. The meaning of the term ‘function’ in
1629
ecology: A coral reef perspective. Functional Ecology 33:948-961.
1630
Benoît, H. P., and D. P. Swain. 2008. Impacts of environmental change and direct and indirect harvesting
1631
effects on the dynamics of a marine fish community. Canadian Journal of Fisheries & Aquatic
1632
Sciences 65:2088-2104.
1633
Berkström, C., T. L. Jörgensen, and M. Hellström. 2013. Ecological connectivity and niche differentiation
1634
between two closely related fish species in the mangrove-seagrass-coral reef continuum. Marine
1635
Ecology Progress Series 477:201-215.
1636
Besnard, L., G. Le Croizier, F. Galván-Magaña, D. Point, E. Kraffe, J. Ketchum, R. O. Martinez Rincon, and G.
1637
Schaal. 2021. Foraging depth depicts resource partitioning and contamination level in a pelagic