Available via license: CC BY 4.0
Content may be subject to copyright.
1
Differential mosquito attraction to humans is 1
associated with skin-derived carboxylic acid levels 2
3
Maria Elena De Obaldia1, Takeshi Morita1, Laura C. Dedmon1, Daniel 4
J. Boehmler2, Caroline S. Jiang3, Emely V. Zeledon1, Justin R. Cross2, 5
and Leslie B. Vosshall1,4,5
6
7
1Laboratory of Neurogenetics and Behavior, The Rockefeller University, New York, NY 10065 8
USA 9
2Donald B. and Catherine C. Marron Cancer Metabolism Center, Memorial Sloan Kettering 10
Cancer Center, New York, NY 10021 USA 11
3Center for Clinical and Translational Science, The Rockefeller University, New York, NY 12
10065 USA 13
4Kavli Neural Systems Institute, New York, NY 10065 USA 14
5Howard Hughes Medical Institute, New York, NY 10065 USA 15
†Current address: Kingdom Supercultures, Brooklyn, NY 11205 USA (M.E.D.) 16
Correspondence: medeobaldia@gmail.com (M.E.D) and leslie@rockefeller.edu (L.B.V.) 17
18
Key words: Aedes aegypti; mosquito; behavior; olfaction; chemosensory receptors; skin; 19
sebum, metabolomics 20
SUMMARY 21
Female Aedes aegypti mosquitoes feed on human blood, which they use to develop their 22
eggs. It has been widely noted that some people are more attractive to mosquitoes than 23
others, but the mechanistic basis of this phenomenon is poorly understood. Here we tested 24
mosquito attraction to skin odor collected from human subjects and identified people who are 25
exceptionally attractive or unattractive to mosquitoes. Notably, these preferences were stable 26
over several years, indicating consistent longitudinal differences in skin odor between subjects. 27
We carried out gas chromatography/quadrupole time of flight-mass spectrometry to analyze 28
the chemical composition of human skin odor in these subjects and discovered that highly 29
attractive people produce significantly increased levels of carboxylic acids. Mosquitoes could 30
reliably distinguish a highly attractive human from their weakly attractive counterparts unless 31
we substantially diluted the odor of the “mosquito magnet.” This is consistent with the 32
hypothesis that odor concentration is a major driver of differential attraction, rather than the 33
less-favored skin odor blend containing repellent odors, although these are not mutually-34
exclusive. Mosquitoes detect carboxylic acids with a large family of odor-gated ion channels 35
encoded by the Ionotropic Receptor gene superfamily. Mutant mosquitoes lacking any of the 36
Ionotropic Receptor (IR) co-receptors Ir8a, Ir25a, and Ir76b, were severely impaired in 37
attraction to human scent but retained the ability to differentiate highly and weakly attractive 38
people. The link between elevated carboxylic acids in “mosquito-magnet” human skin odor and 39
phenotypes of genetic mutations in carboxylic acid receptors suggests that such compounds 40
contribute to differential mosquito attraction. Understanding why some humans are more 41
attractive than others provides insights into what skin odorants are most important to the 42
mosquito and could inform the development of more effective repellents. 43
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
2
INTRODUCTION 44
The globally invasive mosquito Aedes aegypti is a highly efficient vector of viruses including 45
yellow fever, dengue, chikungunya, and Zika among human populations (1). A single female 46
mosquito will bite multiple humans during her 3 to 6-week lifetime to obtain sufficient protein to 47
produce a new batch of eggs as often as every four days (2). This repetitive human-directed 48
feeding behavior allows the mosquito to contract and transmit pathogens in successive bites. 49
An important factor in the effectiveness of Aedes aegypti as a disease vector is that it 50
specializes on human hosts (3-5), thereby focusing pathogen transmission among our species. 51
Female Aedes aegypti mosquitoes have a strong innate drive to hunt humans, using sensory 52
cues including exhaled CO2, body heat, and skin odor. While CO2 and heat are generic stimuli 53
that signify a living warm-blooded animal, skin odor provides information about whether the 54
target is a human or non-human animal (3-5). It is well-documented that mosquitoes are more 55
strongly attracted to some humans than others (6-20), but the underlying mechanisms for this 56
phenomenon remain unclear. The observation that some people are “mosquito magnets” is a 57
topic that captivates the general public and scientific community alike. There is much 58
speculation about possible mechanisms, but only some have a scientific basis. A common 59
explanation offered by non-experts is that differences in ABO blood type “explain” 60
attractiveness to mosquitoes, but experimental data that address this belief are contradictory 61
(21-25). The widely quoted efficacy of eating garlic (26) or B vitamins (27) as a home remedy 62
to repel mosquitoes is similarly unclear. Although a twin study documented a strong heritable 63
component (18), non-genetic factors also contribute to selective attractiveness to mosquitoes. 64
A given person can become more attractive to mosquitoes in contexts including pregnancy 65
(12, 28), malaria parasite infection(29-33), and beer consumption (34, 35). The most widely 66
accepted explanation for these differences is that variation in skin odors produced by different 67
humans, related in part to their unique skin microbiota (9, 17), governs their attractiveness to 68
mosquitoes (8, 13-16). However, the specific chemical mechanism for differential 69
attractiveness to mosquitoes remains unclear. 70
71
Human skin odor is a blend of many organic compounds (36-38), the composition of which has 72
not been exhaustively inventoried. It remains unclear how consistent human skin odor is over 73
time within an individual. Whereas much work has focused on characterizing human axillary 74
(armpit) malodor, there is relatively little information about the composition of the markedly less 75
intense skin odor that emanates from body sites commonly bitten by mosquitoes. Furthermore, 76
additional work is needed before we can fully appreciate the extent of interindividual variation 77
in human skin odor. Thus, it is not known which specific components are most relevant for 78
mosquito attraction to humans, nor do we understand which odorants cause mosquitoes to 79
choose to bite some people over others. Humans who are highly attractive to mosquitoes may 80
produce more attractant odors than other people (5). Alternatively, less attractive humans may 81
emit compounds that repel mosquitoes (39, 40). To date no single molecule obtained from 82
human skin can be said to be sufficient to explain how attractive a person is to mosquitoes. 83
Blends of odorants can be more or less attractive depending on the composition of the blend 84
and the concentration of a specific molecule. For example, the binary blend of ammonia and 85
lactic acid strongly synergizes to elicit mosquito attraction (41, 42). Although carboxylic acids 86
are neutral or repellent when presented individually or in combination with each other, they 87
strongly increase mosquito attraction when combined with ammonia and lactic acid (43, 44). 88
Mosquito attraction behavior is elicited much more reliably using live human hosts or natural 89
odor blends collected directly from humans, than it is by mixing pre-specified compounds, 90
despite improvements in synthetic odor blends as lures for attract-and-kill traps for use in the 91
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
3
field (45-47). Moreover, the current absence of a complete reference metabolome of chemical 92
compounds found on human skin, and the lack of commercially available standard molecules 93
for many skin compounds also limits the effectiveness of human odor blend reconstitution 94
approaches for studying mosquito attraction. 95
96
Mosquitoes use two large multigene families to detect olfactory cues that each encode odor-97
gated ion channels, the odorant receptors (ORs) and the ionotropic receptors (IRs) (48-54). 98
ORs and IRs are evolutionarily unrelated, but both assemble into multi-subunit complexes with 99
a ligand-selective subunit and a co-receptor subunit that does not respond to odorants (48-54). 100
There are hundreds of ligand-selective ORs and IRs in a given insect species, but only one OR 101
co-receptor (Orco) and three IR co-receptors (Ir8a, Ir76b, Ir25a). In Aedes aegypti there are 102
116 ligand-selective ORs and 132 ligand-selective IRs (55). Together these large gene families 103
of odor-gated ion channels sense a vast number of chemical ligands. Although there is some 104
overlap in ligand tuning, ORs generally respond to esters, alcohols, ketones, and aldehydes, 105
and IRs respond to carboxylic acids and amines (50, 56). Because of this co-receptor 106
organization, mutating a single co-receptor gene leads to profound deficits in the ability of an 107
insect to detect whole classes of odorants (49, 54, 57, 58). Nevertheless, mosquitoes are 108
remarkably resilient in the face of such genetic manipulations. Animals lacking the major 109
receptor for carbon dioxide, Gr3, continue to be attracted to humans in semi-field conditions 110
(59). Mosquitoes with a loss of function mutation in Orco lose strong preference for humans 111
over non-human animals but retain strong attraction to humans overall (60, 61). Finally, Ir8a 112
mutants show severe deficits in detecting lactic acid, a major human skin odor, but 113
nevertheless remain partially attracted to humans (61). The recent discovery of extensive co-114
expression of ORs and IRs in single olfactory sensory neurons may explain this functional 115
redundancy (62, 63). 116
117
In this paper we analyzed the skin-derived compounds that differentiate highly from weakly 118
attractive humans and asked which mosquito sensory pathways are required to distinguish 119
such people. We developed a new two-choice behavioral assay that allowed us to test 120
mosquito attraction with higher throughput, allowing for frequent, repeat sampling of human 121
subjects. We collected human skin odor samples on nylon stockings worn on the forearms and 122
profiled the attractiveness of 64 human subjects to mosquitoes. We identified a cohort of highly 123
and weakly attractive people and discovered that the Orco co-receptor is not required for 124
discriminating between them. Mutants lacking Ir8a, Ir76b, and Ir25a retained a preference for 125
“mosquito magnets,” but showed an overall reduced attraction to human skin odor (61). 126
Therefore although neither the OR nor the IR pathway is solely required for discriminating 127
among different people, mutating the IR pathway produced significantly stronger effects on 128
overall mosquito attraction to humans than the OR pathway. We used gas 129
chromatography/quadrupole time of flight-mass spectrometry (GC/QTOF-MS) to identify skin 130
odor molecules that are associated with attractiveness to mosquitoes. Because carboxylic 131
acids have been shown to be attractive to mosquitoes (44), we focused our chemical analysis 132
on detecting acids in the human skin odor blend by using specialized sample preparation to 133
enrich for highly polar acids. Since these are otherwise difficult to detect using standard 134
analytical chemistry approaches, their contribution to differential mosquito attraction to humans 135
is relatively understudied (38). We determined that highly attractive humans have higher levels 136
of several carboxylic acids on their skin than less attractive humans. When we substantially 137
diluted nylons from the most highly attractive subject, mosquitoes were no longer able to 138
distinguish this subject from the least attractive subjects. Our results strongly suggest a link 139
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
4
between the known function of IRs in acid-sensing and our observation that skin-derived 140
carboxylic acids are associated with a person being a “mosquito magnet.” 141
142
RESULTS 143
Mosquitoes show strong preferences for individual humans 144
Preferential mosquito attraction to different stimuli is typically measured in a two-choice 145
olfactometer. In previous studies, we used a Gouck olfactometer (64) to characterize female 146
Aedes aegypti preferences for a human or non-human animal (5, 60). This large apparatus 147
was not suitable for the higher-throughput analysis of mosquito preference for humans 148
required for this study. We therefore adapted a previously described single-stimulus 149
olfactometer (65) to reconfigure it as a two-choice olfactometer (Figure 1A), allowing us to test 150
mosquito preferences between the forearms of two different live human subjects or their 151
forearm skin odor collected on nylon sleeves. In this assay, a mixture of air and carbon dioxide 152
(CO2) was passed over each stimulus to convey volatile odors to mosquitoes downwind 153
(Figure 1A). Mosquitoes flew upwind and those that entered either of the two cylindrical traps 154
in front of each stimulus were scored as attracted. The two-choice assay tests mosquito 155
attraction to odor stimuli that are ~0.91 m away, meaning that to reach the attraction trap, 156
mosquitoes must travel ~610 times their body length, assuming an average female mosquito 157
thorax length of 1.5 mm (66). During the 3.1-year course of this study, we carried out more 158
than 2,330 behavior trials on 174 experimental days. In control experiments we verified that 159
there was no difference in attraction when mosquitoes were offered the left and right forearms 160
of the same subject (Figure 1B). In pilot experiments comparing mosquito attraction among all 161
possible pairings of 3 live human subjects, we identified one (Subject 33) that was significantly 162
more attractive than two others (Subjects 25 and 28) (Figure 1B). Moreover, when the two less 163
attractive subjects were competed against each other, Subject 25 attracted significantly more 164
mosquitoes than Subject 28 (Figure 1B). 165
166
The two-choice olfactometer as configured for live human subjects requires participants to be 167
physically present for competitions that take place in a warm, humid room. To make it feasible 168
to carry out hundreds of competitions between two subjects over many months, we turned to 169
using human forearm odor collected on nylon sleeves as previously described (5, 60). Empty 170
stimulus traps did not attract mosquitoes, and mosquitoes did not prefer unworn nylon over an 171
empty trap (Figure 1C). However, a 3” x 4” swatch cut from nylons worn by Subject 25 was 172
significantly more attractive than the same-sized swatch from an unworn nylon (Figure 1C). To 173
check for side bias in the assay, we used nylons from Subject 25 in both stimulus traps and 174
confirmed that mosquitoes showed no preference (Figure 1C). 175
176
To study interindividual differences in attractiveness to mosquitoes, we recruited an additional 177
5 human subjects who provided skin-scented nylon samples frequently over a period of 178
several months. When we tested nylons worn by each of these 8 subjects against unworn 179
nylons in the two-choice olfactometer assay, we found that each was significantly more 180
attractive than an unworn nylon (Figure 1D). However, there were remarkable differences in 181
the attractiveness of the 8 subjects. When we ranked median attraction, we found that nylons 182
from Subject 28 attracted few mosquitoes, Subject 25 showed intermediate attractiveness, 183
whereas Subject 33 was highly attractive (Figure 1D), mirroring the results of the live human 184
experiments in Figure 1B. This suggests that skin odor is the primary driver of differential 185
mosquito attraction to humans, since temperature and CO2 cues were held constant across all 186
of our nylon experiments. It also suggests that skin odor captured on forearm-worn nylon 187
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
5
sleeves is a good approximation of the odor emanating from a live human forearm. When we 188
tested the 5 additional subjects, we found one additional low attractor (Subject 19) and two 189
additional high attractors (Subjects 24 and 31). Two other subjects showed intermediate 190
attractiveness (Subjects 30 and 32) (Figure 1D). 191
192
We reasoned that in a real-world situation, mosquitoes would choose among multiple different 193
humans in a local area, such that the absolute attractiveness of a single human would not 194
necessarily predict their attractiveness relative to another person. To systematically determine 195
the relative attractiveness of these 8 humans to mosquitoes, we performed a round-robin style 196
“tournament”, competing nylons from all possible subject pairings from this group of 8 subjects, 197
for a total of 28 separate competitions using the two-choice olfactometer assay (Figure 1E). On 198
each experimental day, we verified that mosquitoes did not show a preference when both 199
stimulus ports were baited with nylons collected from the same subject. We sampled each pair 200
of humans on 6 separate days over a period of several months (558 trials, performed over 42 201
experimental days). Among 28 subject pairs tested, we found 13 pairs for which mosquitoes 202
significantly preferred one subject’s odor over the other (Figure 1E). In the other cases, 203
mosquitoes had no preference for one subject over the other. Subject 33 attracted significantly 204
more mosquitoes than every other subject in essentially every trial performed, usually by a 205
large margin. Subjects 19 and 28 were significantly less attractive than several other subjects. 206
When mosquitoes were presented with a choice between the two low attractors, Subjects 19 207
and 28, they were not able to distinguish between them (Figure 1E). To rank subjects from 208
most to least attractive, we devised a scoring system (“attraction score”) based on how many 209
more mosquitoes each subject attracted when competed against all 7 other subjects. By this 210
metric, Subject 33 was the most attractive, yielding an attractiveness score that was 4 times 211
the attractiveness score of the next most attractive subject, and over 100 times greater than 212
that of the two least attractive Subjects 19 and 28 (Figure 1F). These differences in attraction 213
to specific pairs of humans were remarkably stable over many months and were seen with two 214
different wild-type strains of Aedes aegypti (Figure 1G-I). Taken together these results provide 215
empirical evidence that mosquitoes strongly prefer some people over others, and that the 216
olfactory cues that make some people mosquito magnets are stable over many months. 217
218
Orco and Ir8a mutant mosquitoes retain individual human preferences 219
We have shown that small swatches of human-scented nylon provide enough information for 220
mosquitoes to distinguish between and prefer one person over another. What sensory 221
mechanisms do mosquitoes rely on to detect these interindividual differences in skin odor? We 222
first tested the preference of mosquitoes lacking the OR co-receptor Orco, which retain strong 223
attraction to humans, but show severe deficits in discriminating humans from non-human 224
animals (60). Orco mutant mosquitoes showed wild-type levels of activation and attraction in 225
control trials where nylons from Subject 25 were placed in both stimulus boxes (Figure 2A). In 226
the course of carrying out these control experiments we noted low participation in some trials. 227
Therefore, we put inclusion criteria in place such that trials in which 9 or fewer mosquitoes 228
entered either trap were excluded. We note that low levels of participation preclude the 229
accurate calculation of attraction preferences, such as an instance in which only 3 mosquitoes 230
entered either trap. A high percentage of excluded trials may reflect an overall deficit in 231
locomotor activity or general decreases in activation and attraction to human sensory cues or a 232
combination of both factors. 233
234
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
6
More than 90% of all such control trials using both wild type and Orco mutants resulted in at 235
least 10 total mosquitoes being attracted to either stimulus (Figure 2B). Among trials that met 236
the inclusion criteria, Orco mutants did not differ from wild-type controls, retaining the ability to 237
distinguish 5 pairs of highly and weakly attractive humans (Figure 2C-D). This is an important 238
result because in Orco mutants none of the 117 OR genes (55) are functional, suggesting that 239
either ligands that activate ORs do not contribute to interindividual differences in human 240
attractiveness to mosquitoes or that mosquitoes have redundant chemosensory abilities to 241
detect these differences. 242
243
We next tested mosquitoes lacking the co-receptor Ir8a, which is expressed in the antenna 244
and necessary for detection of several acids, including lactic acid, a component of human 245
sweat (61). Ir8a mutants showed decreased overall attraction to Subject 25 in the two-choice 246
olfactometer assay, despite normal levels of activation, as defined by entry into the flying 247
chamber (Figure 3A). In control trials using Subject 25 nylons in both stimulus boxes, 94% of 248
trials using wild-type mosquitoes resulted in 10 or more total mosquitoes being attracted to 249
either stimulus, but this was reduced to 73% for trials using Ir8a mutants (Figure 3B). Among 250
trials that met the inclusion criteria, Ir8a mutants largely retained the same preferences as wild-251
type controls, despite their diminished overall attraction to human odor (Figure 3C-D). This 252
result again suggests that mosquitoes impaired in detecting specific human body odors are still 253
able to tell the difference between different people. 254
255
Generation and behavioral characterization of Ir76b and Ir25a mutants 256
We next used CRISPR-Cas9 to generate mosquitoes that lack the two other IR co-receptors 257
Ir76b and Ir25a (Figure 4A-B, Supplementary Figure S1). We generated 2 mutant alleles of 258
each gene (Ir76b32, Ir76b61, Ir25aBamHI, and Ir25a19) and tested the behavior of heterozygous 259
animals of all 4 strains and the heteroallelic Ir76b32/61 and Ir25aBamHI/19 null mutants. We noted 260
that homozygous Ir76b and Ir25a mutants had difficulty blood-feeding and that Ir25a 261
homozygous mutants generally laid fewer eggs. A recent analysis of Anopheles coluzzii Ir76b 262
null mutants reported that they show normal attraction to human host cues but do not blood-263
feed or produce eggs (67). To characterize the response of these strains and wild-type control 264
mosquitoes to human cues, we first used a single-stimulus olfactometer (65) to examine 265
activation and attraction levels (Figure 4C), before conducting preference assays. Ir76b32/61
266
mutants displayed reduced general activity levels in response to an unworn nylon or a nylon 267
worn by Subject 33 (Figure 4D). However, this defect was readily overcome when Ir76b 268
mutants were presented with the forearm of Subject 33, indicating the absence of gross motor 269
defects (Figure 4D). Ir25aBamHI/19 mutants showed wild-type levels of activation across all 270
stimuli tested (Figure 4E). We additionally asked if the Ir76b and Ir25a mutants were attracted 271
to the same stimuli described above. Ir76b32/61 mutants showed normal levels of attraction to 272
nylons worn by Subject 33 and to the forearm of Subject 33 (Figure 4F). In contrast, 273
Ir25aBamHI/19 mutants displayed significant defects in their attraction to both Subject 33 nylons 274
and to the forearm of Subject 33 (Figure 4G). This suggests that the Ir25a co-receptor, along 275
with one or more of the ligand-selective IRs with which it assembles a functional receptor, 276
plays an important role in detecting human skin emanations. 277
278
Ir76b and Ir25a mutant mosquitoes retain individual human preferences 279
Given the dramatic decrease in attraction of Ir25aBamHI/19 mutant mosquitoes to Subject 33, we 280
next asked if these mutants along with Ir76b32/61 mutants could distinguish between nylons 281
worn by highly and weakly attractive human subjects, when these were presented 282
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
7
simultaneously in the two-choice olfactometer assay. The two-choice assay is substantially 283
larger than the single-choice assay, and thus may represent a more difficult behavioral task, so 284
we expected that it might reveal additional phenotypes not seen in Figure 4D-G. 285
286
In control two-choice assay trials in which Subject 25 nylons were placed in both stimulus 287
boxes, Ir76b32/61 mutants showed lower activation in response to human worn nylons, and both 288
Ir76b32/61 and Ir25aBamHI/19 mutants showed significantly decreased attraction compared to wild-289
type controls (Figure 5A). Furthermore, only 25% of Ir76b32/61 and 38% of Ir25aBamHI/19 mutant 290
trials had at least 10 mosquitoes attracted to either stimulus, compared to 100% of wild type 291
trials (Figure 5B). 292
293
Analysis of trials that met these inclusion criteria showed that despite substantial defects in 294
overall attraction to human odor, both Ir76b32/61 and Ir25aBamHI/19 mutants were able to 295
distinguish highly and weakly attractive human subjects (Figure 5C-E). These data indicate 296
that mosquitoes have evolved highly redundant sensory systems permitting them to retain 297
attraction to humans even with significant genetic disruption of their olfactory system (62). 298
Nevertheless, mutating the IR pathway produced significantly stronger effects on overall 299
mosquito attraction to humans than the OR pathway. 300
301
Carboxylic acids are elevated in skin odor of highly attractive humans 302
We have shown that human-worn nylon sleeves provide the sensory cues needed for 303
mosquitoes to reliably discriminate between humans. We reasoned that the chemistry of skin-304
derived odors in these nylons would provide insights into what distinguishes a highly attractive 305
person from a weakly attractive person. We used gas chromatography/quadrupole time of 306
flight-mass spectrometry (GC/QTOF-MS) to identify compounds on the nylon sleeves that 307
were associated with mosquito attractiveness (Figure 6A). Because IR co-receptor mutants 308
showed significantly reduced attraction to humans, we focused our chemical analysis on acidic 309
compounds, which are detected by the IR pathway (61, 68, 69). We therefore did not 310
investigate whether “mosquito-magnet” humans produce elevated levels of other compounds 311
not captured by our analytical strategy. We collected nylons from 7 of 8 human subjects in our 312
initial subject cohort on 4 days spaced at least 1 week apart, and then performed 4 313
independent analyses (Experiments 1.1-1.4; Figure 6B). Briefly, sections of human-exposed 314
nylon sleeves were extracted in 80:20 methanol:water. The resulting extract was derivatized 315
with pentafluorobenzyl bromide (PFB-Br), an aqueous-compatible derivatization reagent that 316
reacts with carboxylic acids, improving their chromatography performance and ionization 317
efficiency (Supplementary Figure S2A). Samples were then analyzed by GC-MS using 318
negative methane chemical ionization on a Q-TOF instrument, allowing formula prediction from 319
the mass of detected ions. This strategy has been widely used to measure volatile and semi-320
volatile carboxylic acids but to our knowledge it has not previously been applied to examine 321
skin metabolites in the context of mosquito preferences. 322
323
Using unbiased feature detection and data filtering we identified 204 molecular features that 324
were likely to be human-derived, since they were enriched on subject nylons versus unworn 325
nylons and method blanks, across all 4 experiments (Supplementary Table S1). We noted that 326
high attractor subjects appeared to have more of these putative “human-derived” peaks overall 327
than low attractor subjects (Figure 6C). Of these, ~50 features were differentially present in 328
samples from the two most attractive subjects, Subjects 33 and 31, versus the two least 329
attractive subjects, Subjects 19 and 28, in all 4 replicate experiments (Figure 6D,E, 330
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
8
Supplementary Table S1). Interestingly, nearly all differentially enriched features (49/51) were 331
more abundant in the 2 highly attractive subjects, although we did find 2 features that were 332
enriched in the 2 low attractors (Figure 6E). We were able to predict chemical formulas for 333
about 40 of these features and ultimately identified 9 as straight chain fatty acids by matching 334
their mass and retention time to that of authentic standards (Figure 6E). We then extracted the 335
signals for all the straight chain acids with acyl chain lengths between 3-20 carbons (Figure 336
6F). Consistent with the untargeted analysis described above, these compounds were all low 337
or absent in unworn nylons and solvent controls, but only fatty acids with >10 carbons 338
appeared to be enriched in the most highly attractive subjects (Figure 6F-G, Supplementary 339
Figure S2B). To exclude these changes being attributable to variations in sample loading, 340
derivatization or extraction efficiency we verified that several control compounds were similarly 341
abundant in the high and the low attractor samples, including 2 deuterated internal standards 342
present in the extraction solvent and an unknown “nylon-derived” entity that was present in all 343
nylon containing-samples, and absent in solvent-alone controls (Supplementary Figure 344
S2C,D). 345
346
Association of elevated skin-derived carboxylic acids and mosquito 347
attraction confirmed in a validation cohort 348
To confirm our finding that highly mosquito-attracting humans produced more abundant 349
carboxylic acids on their skin, we enrolled 56 new human subjects in a validation study 350
(Supplementary Figure S3A) and used the higher-throughput single-stimulus olfactometer 351
assay to screen the attractiveness of nylons from a given new subject compared to an unworn 352
nylon (Figure 7A). Alongside the 56 new subjects, we also tested nylons from 7 subjects from 353
the initial cohort in this assay (Figure 7B). Consistent with our initial behavioral studies, 354
performed over 1 year earlier, Subject 28 was the least attractive of all 64 human subjects that 355
we tested over the course of the whole study, and Subject 33 was among the most attractive 356
subjects (Figure 7B, Figure 1G). Based on initial behavioral results we moved 18 subjects (4 357
from the initial cohort and 14 from the validation cohort) forward to metabolite profiling with 358
GC/QTOF-MS. These comprised 11 subjects that were highly attractive and 7 that were 359
weakly attractive. Subjects provided 4 more odor samples, spaced 1 week apart, which were 360
used for additional behavioral testing to confirm their high/low attractor status, and for 361
GC/QTOF-MS analysis. In Figure 7B we plot all data collected from the low and high attractor 362
groups that were included in the GC/QTOF-MS analysis. Behavioral data from 45 additional 363
subjects, 42 subjects from the validation cohort and 3 subjects from the initial cohort, who were 364
not included in the GC/QTOF-MS validation study because of sample size limitations are 365
available on Zenodo (DOI: 10.5281/zenodo.5822539). 366
367
We again performed 4 replicate metabolomic experiments (Figure 7C; Experiments 2.1-2.4) 368
using PFB-Br derivatization to detect carboxylic acid-containing molecules. Since the validation 369
study included more subjects than our initial cohort, we reasoned that untargeted analysis may 370
reveal additional features of interest that were not found in the first study. Hence, we repeated 371
the untargeted analysis of GC/QTOF-MS data, as described above, and found 161 molecular 372
features that were likely to be human-derived, since they were more abundant in worn nylons 373
from at least 1 subject than unworn nylons and solvent controls in all 4 experiments 374
(Supplementary Table S1). We then filtered features that were differentially abundant in high 375
versus low attractor groups in all 4 replicate experiments, resulting in a list of 13 features, all of 376
which were enriched in the high attractor group (Figure 7D-E). We identified 3 of these 377
features as carboxylic acids: pentadecanoic acid, heptadecanoic acid, and nonadecanoic acid 378
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
9
(Figure 7E, Supplementary Figure S2G). We were not able to identify the remaining 10 379
features definitively, although in some cases we were able to predict a chemical formula. Of 380
note, many of these features had the same predicted formula as the identified straight-chain 381
fatty acids but eluted at different retention times, making it likely these are branched chain 382
isoforms of the identified carboxylic acids. As expected, deuterated internal standards and a 383
nylon-derived entity showed no difference in abundance between the high and low attractor 384
groups (Supplementary Figure S2E,F). 385
386
We next performed a targeted re-analysis of carboxylic acids with 10-20 carbons in individual 387
human subjects and control samples (Figure 7F-H). The abundance of carboxylic acids on 388
individual subjects from the initial cohort was remarkably consistent with results obtained about 389
1 year earlier: low attractors Subjects 19 and 28 had much lower levels of many carboxylic 390
acids than the more highly attractive Subjects 31 and 33 (Figure 6F, Figure 7F, Supplementary 391
Figure S3B). The carboxylic acid pattern was similarly consistent from week to week for 392
individual subjects in the larger cohort in Experiments 2.1-2.4 (Supplementary Figure S3C). 393
Overall, the high attractor group had significantly higher levels of 3 carboxylic acids 394
(pentadecanoic, heptadecanoic, nonadecanoic) than the low attractor subjects in this targeted 395
re-analysis of the data (Figure 7G-H). However, not all individual subjects fit this pattern. Low 396
attractor Subject 90 had high levels of all carboxylic acids examined, in contrast to the 6 other 397
low attractors (Figure 7F). In both the initial and validation cohorts, we documented an 398
association between high levels of skin carboxylic acids and attractiveness to mosquitoes. 399
400
Dilution of highly attractive human odor eliminates mosquito preferences 401
Our GC/QTOF-MS experiments revealed that 3 carboxylic acids and 10 unknown compounds 402
were consistently enriched in highly attractive subjects versus less attractive subjects, whereas 403
we did not find any compounds that were consistently enriched in the low attractor group. 404
Accordingly, we hypothesized that the high attractors have higher levels of mosquito attractant 405
compounds on their skin. To test this hypothesis, we performed a dose-response experiment, 406
in which we competed different sized swatches of high attractor Subject 33 nylon against a 407
standard 5.08 cm x 2.54 cm-sized swatch of low attractor worn nylon from Subject 19 or 408
Subject 28 (Figure 8). We found that mosquitoes preferred the odor blend of high attractor 409
Subject 33 to that of both low attractor subjects, even when presented with a substantially 410
smaller swatch of Subject 33 nylon than the less attractive subject’s nylon (Figure 8). Subject 411
33 nylon only became indistinguishable from low attractor nylons from Subject 28 and Subject 412
19 nylon, when competed against a swatch of Subject 33 nylon that was 32-fold or 8-fold 413
smaller, respectively (Figure 8). Therefore, the skin odor blend of highly attractive Subject 33 414
provided a much more potent attractive stimulus than that of weakly attractive humans, which 415
is consistent with our GC/QTOF-MS findings that highly attractive subject nylons contained 416
more human-derived compounds, including carboxylic acids, than those of less attractive 417
subjects. 418
419
DISCUSSION 420
Why are some people more attractive to mosquitoes than others? 421
In this work, we establish that the differential attractiveness of individual humans to mosquitoes 422
is a stable over many months and is associated with the abundance of skin-associated 423
carboxylic acids. Previous work demonstrated that mice at a specific stage of infection with 424
malaria parasites were more attractive to mosquitoes, and that these infected mice showed an 425
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
10
overall increase in concentration of many emitted volatile compounds (32). These results are 426
consistent with our work. Highly attractive subjects produced significantly higher levels of 3 427
carboxylic acids—pentadecanoic, heptadecanoic, and nonadecanoic acids—as well as 10 428
unidentified compounds in this same chemical class. The specific blend of these and other 429
carboxylic acids varied between different high attractive subjects. Therefore, there may be 430
more than one way for a person to be highly attractive to mosquitoes. We did not identify any 431
compounds that were reproducibly enriched on the skin of the least attractive humans, 432
consistent with the idea that these individuals lack mosquito attractants, rather than emitting a 433
shared set of repellent compounds. A limitation of our study is that the chemical derivatization 434
we used specifically focused on metabolites containing carboxylic acid groups. However, a 435
strength of this approach is that this aqueous-compatible derivatization allowed us to measure 436
metabolites with a wide range of volatilities using a single analytical method, from highly 437
volatile short chain fatty acids to medium volatility acids to longer chain fatty acids of lower 438
volatility. A previous study that analyzed interindividual differences in mosquito attractiveness 439
focused on different classes of compounds and identified five—6-methyl-5-hepten-2-one, 440
octanal, nonanal, decanal, and geranylacetone—that were enriched on the skin of weakly 441
attractive humans (39). Individually these compounds reduced mosquito flight activity and/or 442
attraction, suggesting that some people may release natural repellents that make them less 443
attractive to mosquitoes. One of our subjects, Subject 90, had high levels of carboxylic acids in 444
their skin emanations but was only weakly attractive to mosquitoes. It is plausible that Subject 445
90 produces higher levels of a natural repellent that would counteract the elevated levels of 446
carboxylic acids, but this was not tested in our study. 447
448
Mosquito attractiveness of a given person is stable over years 449
Our study focused on humans who were empirically determined to lie at extremes of the 450
mosquito-attractiveness spectrum. Several subjects were tested regularly over the entire 3-451
year duration of the study, allowing us to determine that extreme mosquito attraction is 452
remarkably stable. All humans tested in our study were at least somewhat attractive to 453
mosquitoes when their skin odor blend was presented as a single stimulus. However, we 454
estimate that skin odor from Subject 33 was over 30 times as potent that of the least attractive 455
human sampled, Subject 28. Our behavioral assay results corroborate anecdotal evidence that 456
context matters for how attractive a person is to mosquitoes in a real-world setting, since 457
mosquitoes feed opportunistically. If one human walks into a highly mosquito-infested 458
environment alone, they may receive many bites, regardless of their overall attractiveness 459
level because they are the only feeding option. Mosquito preferences matter more in group 460
settings. The “mosquito-magnet” in the group may receive the most bites, leaving the less 461
attractive humans largely untouched. We observed that mosquitoes show preferences 462
between individuals who are extremely highly attractive on their own. For example, nylons from 463
Subjects 31, 24, and 33 were all similarly highly attractive when competed against an unworn 464
nylon. However, Subject 33 was more attractive than the two other subjects when competed 465
head-to head in the two-choice assay. Furthermore, mosquito preferences were transitive: 466
Subject 33>Subject 25>Subject 28. This suggests that mosquitoes distinguish the scent of two 467
human samples using cues that exist along a continuum. Consistent with this idea, we were 468
able to dilute the dose of nylon from a highly attractive human to the point where it was equally 469
attractive as a nylon from a low attractor human. We propose that exceptionally high or low 470
attractiveness to mosquitoes is a “fixed” trait, caused by factors that remain constant over a 471
period of several years, even when environmental factors are not strictly controlled. We did not 472
require study participants to maintain a constant diet, exercise regimen, and we did not limit 473
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
11
changes to medication or personal care product usage over the 3 years of the study. It has 474
been shown that identical twins are more similarly attractive to mosquitoes than fraternal twins 475
(18), suggesting a genetic component to mosquito attractiveness. Moreover, the blend of 476
carboxylic acids that characterizes individual human body odor types is more similar in 477
monozygotic twins than unrelated subjects (70). We speculate that genetically determined skin 478
characteristics and/or other very stable inter-individual differences contribute to making 479
someone highly or weakly attractive to mosquitoes. 480
481
Sensory mechanisms of mosquito discrimination between humans 482
In this study, we discovered that mutating any one of the three carboxylic acid-sensing IR co-483
receptors, Ir8a, Ir76b, or Ir25a, strongly reduced overall attraction to human scent, but did not 484
abolish the ability of any of these mutants to discriminate between highly and weakly attractive 485
people. We also found that elevated levels of skin-derived carboxylic acids – ligands that 486
activate the IR pathway – are associated with high mosquito attractiveness. Although our data 487
do not allow us to conclude that skin carboxylic acid abundance causes an individual to be 488
highly attractive to mosquitoes, the association between these phenotypes is interesting. In 489
contrast, Orco mutants, which have an impaired ability to sense other classes of chemical 490
compounds, including esters, ketones, alcohols, and aldehydes, are highly attracted to 491
humans (5, 60, 61). In our behavioral assays using Orco mutants, there was no significant 492
difference in either activation or attraction, or in their ability to distinguish between any of the 493
pairs of human subjects. 494
495
Because no single mutation was able to disrupt the ability of mosquitoes to discriminate 496
between two subjects, we speculate that there is extensive redundancy in the detection of 497
human-derived skin odors. This redundancy may be due to central olfactory coding 498
mechanisms or to the recently described co-expression of ORs and IRs in the same olfactory 499
sensory neuron (63, 71). One possible mechanism of IR redundancy could lie with their use of 500
three and not one co-receptor as for the OR system. Removing any one IR co-receptor might 501
have only a partial effect on the ability of the mosquito to detect odorants sensed by IRs. One 502
possibility is that the three IR co-receptors together with ligand-selective IRs collaborate to tile 503
the chemical space of carboxylic acids, such that removing any single IR co-receptor reduces 504
the overall attraction to humans but allows the mutants to retain the ability to detect differences 505
in levels of carboxylic acids. 506
507
Our findings argue against the idea that mosquitoes distinguish between highly and weakly 508
attractive humans using a single odor, such as lactic acid, as has been suggested (13). If this 509
were the case, we would expect that Ir8a mutants, which cannot sense lactic acid (61), would 510
lose their preference for highly attractive humans. Instead, we propose that mosquito magnets 511
produce elevated levels of multiple mosquito-attractant compounds, and that this drives 512
mosquito preferences. Among the chemical entities we found to be elevated on highly 513
mosquito attractive human skin, we identified 3 compounds as straight chain, unsaturated 514
carboxylic acids with 15,17, and 19 carbons. The relative contribution of these 3 identified 515
compounds and the unidentified entities to mosquito preferences remains unknown. Notably, 516
fatty acids with greater than 10 carbons are not very volatile, so it is unclear whether the 517
compounds we identified are directly sensed by the mosquito, or whether they may give rise to 518
more volatile components that are also enriched on the skin of mosquito magnet subjects. 519
520
Microbiota influence over human skin acid production 521
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
12
Free fatty acids are more abundant on the skin surface of humans than non-human animals 522
(72), and therefore they may be an important indicator to mosquitoes that there is a human 523
nearby. Human skin is unique among mammals because it has relatively little hair and 524
numerous eccrine sweat glands across most of its surface. Some animals, including humans, 525
produce a specialized waxy substance from sebaceous glands called sebum. In humans, 526
sebum is triglyceride-rich, thereby producing a characteristic surface lipid composition, which 527
contains about 25% free fatty acids (72). The unique skin lipid composition of humans is 528
thought to have protective effects, such as limiting sun damage in the absence of protective 529
hair, and emulsifying eccrine sweat, preventing its overly rapid evaporation to allow for 530
appropriate body temperature regulation (73, 74). Human skin acids are astonishingly diverse, 531
with branched, odd-chain, and esterified fatty acids reported, along with skin-specific patterns 532
of desaturation (73). This complexity is also likely reflected in the 10 unidentified features 533
upregulated in highly attractive subjects, as many appeared to be structural isomers of the 534
identified straight-chain acids. Given the vast array of acid types found on the skin, it is unlikely 535
that two individual humans will possess the same exact complement of acids in the exact 536
same ratios, potentially giving each human a unique chemical signature. Skin bacteria 537
contribute to the pool of free fatty acids found on human skin by producing several types of 538
fatty acid synthetase enzymes that allow them to produce diverse types of acids themselves 539
(75, 76) and by cleaving free fatty acids from human sebum triglycerides using lipase enzymes 540
(72). Additionally, recent work has shown that skin microbiota composition is remarkably stable 541
within an individual over time, even though skin is exposed to a constantly fluctuating 542
environment (77). Most viable skin bacteria reside in the pores, where they are protected from 543
external factors, such as hygiene habits and seasonal weather changes (78). It is reasonable 544
to think that an individual’s skin microbiota contributes to their skin acid composition, which we 545
have shown to be remarkably stable over time. 546
547
Closing remarks 548
The attraction preferences of disease-vectoring mosquitoes have important public health 549
implications, since it is estimated that in disease endemic areas a small fraction of humans is 550
more frequently targeted, and these individuals serve as a reservoir of pathogens (11, 79). 551
Understanding the mechanistic basis for mosquito biting preferences will suggest new ways to 552
reduce mosquito attraction to humans and curb the spread of dangerous arboviruses. Studies 553
in humans (29-31, 33) and mice (32) have demonstrated that infection with malaria parasites 554
alters their attractiveness to mosquitoes by altering the odors they produce, leading to greater 555
pathogen transmission. Understanding what makes someone a “mosquito magnet” will 556
suggest ways to rationally design interventions such as skin microbiota manipulation to make 557
people less attractive to mosquitoes. We propose that the ability to predict which individuals in 558
a community are high attractors would allow for more effective deployment of resources to 559
combat the spread of mosquito-borne pathogens. 560
561
ACKNOWLEDGMENTS 562
We thank Lindy McBride, Jessica Zung, and members of the Vosshall Lab for comments on 563
the manuscript; all study participants for their generous participation in this research; Gloria 564
Gordon and Libby Mejia for expert mosquito rearing; Ben Matthews and Veronica Jové for 565
discussion and guidance on Ir76b and Ir25a mutagenesis strategies; Maya Davidov and 566
Naquia Unwala for help with pilot behavioral studies to establish the feasibility of this work; 567
Arka Rao, Sara Nunes Violante, Michelle Saio, Ruben Jose Jesus Faustino Ramos, and 568
members of the Donald B. and Catherine C. Marron Cancer Metabolism Center for guidance 569
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
13
with metabolomic experiments and GC/QTOF-MS data analysis; and the PIT Crew: Jim 570
Petrillo, Peer Strogies, and Dan Gross of the Rockefeller University Precision Instrumentation 571
Technologies resource center for advice on assay fabrication. 572
573
AUTHOR CONTRIBUTIONS 574
M.E.D. carried out or supervised all experiments and data analysis in the paper with additional 575
contributions from co-authors. T.M. generated and carried out initial characterization of Ir76b 576
and Ir25a mutants. L.C.D. helped conceive the preference study in Figure 1E and carried out 577
all behavioral experiments in the paper except the validation cohort experiments in Figure 7B, 578
which were carried out by E.V.Z. M.E.D. conducted the initial GC/QTOF-MS experiments in 579
Figure 6. D.J.B. together with M.E.D. carried out the validation GC/QTOF-MS experiments in 580
Figure 7. C.S.J. contributed statistical expertise and code for analysis of both mosquito 581
behavior data and untargeted analysis of GC/QTOF-MS data. J.R.C. supervised D.J.B. and 582
provided technical and conceptual guidance for the GC/QTOF-MS experiments. M.E.D. and 583
L.B.V. together conceived the study, designed the figures, and wrote the paper with input from 584
all authors. 585
586
FUNDING 587
This work was supported in part by grant # UL1 TR001866 from the National Center for 588
Advancing Translational Sciences (NCATS, National Institutes of Health (NIH) Clinical and 589
Translational Science Award (CTSA) program, which also provided a pilot award to M.E.D. 590
Additional funding for this study was provided by a Harvey L. Karp Discovery Award and a 591
Helen Hay Whitney Postdoctoral Fellowship to M.E.D.; a Harvey L. Karp Discovery Award and 592
a Japan Society for Promotion of Science Overseas Research Fellowship to T.M.; and an NIH-593
NIAID R25 AI140472 Tri-Institutional Metabolomics Training Program grant to J.R.C. L.B.V. is 594
an investigator of the Howard Hughes Medical Institute. 595
596
REFERENCES 597
598
1. S. Bhatt et al., The global distribution and burden of dengue. Nature 496, 504-507 599
(2013). 600
2. A. Ponlawat, L. C. Harrington, Blood feeding patterns of Aedes aegypti and Aedes 601
albopictus in Thailand. J Med Entomol 42, 844-849 (2005). 602
3. Z. Zhao et al., Chemical signatures of human odour generate a unique neural code in 603
the brain of Aedes aegypti mosquitoes. bioRxiv, 10.1101/2020.1111.1101.363861 604
(2020). 605
4. N. H. Rose et al., Climate and urbanization drive mosquito preference for humans. Curr 606
Biol 30, 3570-3579 e3576 (2020). 607
5. C. S. McBride et al., Evolution of mosquito preference for humans linked to an odorant 608
receptor. Nature 515, 222-227 (2014). 609
6. I. H. Gilbert, H. K. Gouck, N. Smith, Attractiveness of men and women to Aedes aegypti 610
and relative protection time obtained with DEET. Fla Entomol 49, 53-66 (1966). 611
7. W. R. Mukabana, W. Takken, R. Coe, B. G. Knols, Host-specific cues cause differential 612
attractiveness of Kenyan men to the African malaria vector Anopheles gambiae. Malar J 613
1, 17 (2002). 614
8. Y. T. Qiu, R. C. Smallegange, V. A. N. L. JJ, C. J. Ter Braak, W. Takken, Interindividual 615
variation in the attractiveness of human odours to the malaria mosquito Anopheles 616
gambiae s. s. Med Vet Entomol 20, 280-287 (2006). 617
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
14
9. N. O. Verhulst et al., Composition of human skin microbiota affects attractiveness to 618
malaria mosquitoes. PLoS One 6, e28991 (2011). 619
10. B. G. Knols, R. de Jong, W. Takken, Differential attractiveness of isolated humans to 620
mosquitoes in Tanzania. Trans R Soc Trop Med Hyg 89, 604-606 (1995). 621
11. L. C. Harrington et al., Heterogeneous feeding patterns of the dengue vector, Aedes 622
aegypti, on individual human hosts in rural Thailand. PLoS Negl Trop Dis 8, e3048 623
(2014). 624
12. J. Ansell, K. A. Hamilton, M. Pinder, G. E. Walraven, S. W. Lindsay, Short-range 625
attractiveness of pregnant women to Anopheles gambiae mosquitoes. Trans R Soc 626
Trop Med Hyg 96, 113-116 (2002). 627
13. F. Acree, Jr., R. B. Turner, H. K. Gouck, M. Beroza, N. Smith, L-Lactic acid: a mosquito 628
attractant isolated from humans. Science 161, 1346-1347 (1968). 629
14. U. R. Bernier, D. L. Kline, C. E. Schreck, R. A. Yost, D. R. Barnard, Chemical analysis 630
of human skin emanations: comparison of volatiles from humans that differ in attraction 631
of Aedes aegypti (Diptera: Culicidae). J Am Mosq Control Assoc 18, 186-195 (2002). 632
15. J. Brady, C. Costantini, N. Sagnon, G. Gibson, M. Coluzzi, The role of body odours in 633
the relative attractiveness of different men to malarial vectors in Burkina Faso. Ann Trop 634
Med Parasitol 91, S121-S122 (1997). 635
16. R. Brouwer, Variations in human body odour as a cause of individual differences of 636
attraction for malaria mosquitoes. Trop Geogr Med, 186-192 (1960). 637
17. M. A. H. Braks, W. Takken, Incubated human sweat but not fresh sweat attracts the 638
malaria mosquito Anopheles gambiae sensu stricto. J Chem Ecol 25, 663-672 (1999). 639
18. G. M. Fernandez-Grandon, S. A. Gezan, J. A. Armour, J. A. Pickett, J. G. Logan, 640
Heritability of attractiveness to mosquitoes. PLoS One 10, e0122716 (2015). 641
19. Y. E. Himeidan, M. I. Elbashir, I. Adam, Attractiveness of pregnant women to the 642
malaria vector, Anopheles arabiensis, in Sudan. Ann Trop Med Parasitol 98, 631-633 643
(2004). 644
20. J. K. Konopka et al., Olfaction in Anopheles mosquitoes. Chem Senses 46, (2021). 645
21. Y. Shirai, H. Funada, T. Seki, M. Morohashi, K. Kamimura, Landing preference of 646
Aedes albopictus (Diptera: Culicidae) on human skin among ABO blood groups, 647
secretors or nonsecretors, and ABH antigens. J Med Entomol 41, 796-799 (2004). 648
22. S. A. Khan, A. Ombugadu, S. Ahmad, Host-seeking behavior and fecundity of the 649
female Aedes aegypti to human blood types. Pest Manag Sci, (2021). 650
23. C. Thornton, C. J. Doré, J. O. C. Willson, J. L. Hubbard, Effects of human blood group, 651
sweating and other factors on individual host selection by species A of the Anopheles 652
gambiae complex (Diptera, Culicidae). Bull Entomol Res 66, 651-663 (1976). 653
24. C. S. Wood, G. A. Harrison, C. Doré, J. S. Weiner, Selective feeding of Anopheles 654
gambiae according to ABO blood group status. Nature 239, 165 (1972). 655
25. M. Anjomruz et al., ABO blood groups of residents and the ABO host choice of malaria 656
vectors in southern Iran. Exp Parasitol 136, 63-67 (2014). 657
26. T. V. Rajan, M. Hein, P. Porte, S. Wikel, A double-blinded, placebo-controlled trial of 658
garlic as a mosquito repellant: a preliminary study. Med Vet Entomol 19, 84-89 (2005). 659
27. A. R. Ives, S. M. Paskewitz, Testing vitamin B as a home remedy against mosquitoes. J 660
Am Mosq Control Assoc 21, 213-217 (2005). 661
28. S. Lindsay et al., Effect of pregnancy on exposure to malaria mosquitoes. Lancet 355, 662
1972 (2000). 663
29. R. Lacroix, W. R. Mukabana, L. C. Gouagna, J. C. Koella, Malaria infection increases 664
attractiveness of humans to mosquitoes. PLoS Biol 3, e298 (2005). 665
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
15
30. A. O. Busula et al., Gametocytemia and attractiveness of Plasmodium falciparum-666
Infected Kenyan children to Anopheles gambiae mosquitoes. J Infect Dis 216, 291-295 667
(2017). 668
31. A. Robinson et al., Plasmodium-associated changes in human odor attract mosquitoes. 669
Proc Natl Acad Sci U S A 115, E4209-e4218 (2018). 670
32. C. M. De Moraes et al., Malaria-induced changes in host odors enhance mosquito 671
attraction. Proc Natl Acad Sci U S A 111, 11079-11084 (2014). 672
33. C. M. De Moraes et al., Volatile biomarkers of symptomatic and asymptomatic malaria 673
infection in humans. Proc Natl Acad Sci U S A 115, 5780-5785 (2018). 674
34. T. Lefevre et al., Beer consumption increases human attractiveness to malaria 675
mosquitoes. PLoS One 5, e9546 (2010). 676
35. O. Shirai et al., Alcohol ingestion stimulates mosquito attraction. J Am Mosq Control 677
Assoc 18, 91-96 (2002). 678
36. M. Gallagher et al., Analyses of volatile organic compounds from human skin. Br J 679
Dermatol 159, 780-791 (2008). 680
37. U. R. Bernier, M. M. Booth, R. A. Yost, Analysis of human skin emanations by gas 681
chromatography/mass spectrometry. 1. Thermal desorption of attractants for the yellow 682
fever mosquito (Aedes aegypti) from handled glass beads. Anal Chem 71, 1-7 (1999). 683
38. U. R. Bernier, D. L. Kline, D. R. Barnard, C. E. Schreck, R. A. Yost, Analysis of human 684
skin emanations by gas chromatography/mass spectrometry. 2. Identification of volatile 685
compounds that are candidate attractants for the yellow fever mosquito (Aedes aegypti). 686
Anal Chem 72, 747-756 (2000). 687
39. J. G. Logan et al., Identification of human-derived volatile chemicals that interfere with 688
attraction of Aedes aegypti mosquitoes. J Chem Ecol 34, 308-322 (2008). 689
40. J. G. Logan et al., Arm-in-cage testing of natural human-derived mosquito repellents. 690
Malar J 9, 239 (2010). 691
41. M. Geier, O. J. Bosch, J. Boeckh, Ammonia as an attractive component of host odour 692
for the yellow fever mosquito, Aedes aegypti. Chem Senses 24, 647-653 (1999). 693
42. B. M. Steib, M. Geier, J. Boeckh, The effect of lactic acid on odour-related host 694
preference of yellow fever mosquitoes. Chem Senses 26, 523-528 (2001). 695
43. R. C. Smallegange, Y. T. Qiu, J. J. van Loon, W. Takken, Synergism between 696
ammonia, lactic acid and carboxylic acids as kairomones in the host-seeking behaviour 697
of the malaria mosquito Anopheles gambiae sensu stricto (Diptera: Culicidae). Chem 698
Senses 30, 145-152 (2005). 699
44. R. C. Smallegange, Y. T. Qiu, G. Bukovinszkine-Kiss, J. J. Van Loon, W. Takken, The 700
effect of aliphatic carboxylic acids on olfaction-based host-seeking of the malaria 701
mosquito Anopheles gambiae sensu stricto. J Chem Ecol 35, 933-943 (2009). 702
45. U. R. Bernier, D. L. Kline, S. A. Allan, D. R. Barnard, Laboratory comparison of Aedes 703
aegypti attraction to human odors and to synthetic human odor compounds and blends. 704
J Am Mosq Control Assoc 23, 288-293 (2007). 705
46. N. O. Verhulst et al., Improvement of a synthetic lure for Anopheles gambiae using 706
compounds produced by human skin microbiota. Malar J 10, 28 (2011). 707
47. B. N. Njiru, W. R. Mukabana, W. Takken, B. G. Knols, Trapping of the malaria vector 708
Anopheles gambiae with odour-baited MM-X traps in semi-field conditions in western 709
Kenya. Malar J 5, 39 (2006). 710
48. K. Sato et al., Insect olfactory receptors are heteromeric ligand-gated ion channels. 711
Nature 452, 1002-1006 (2008). 712
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
16
49. R. Benton, K. S. Vannice, C. Gomez-Diaz, L. B. Vosshall, Variant ionotropic glutamate 713
receptors as chemosensory receptors in Drosophila. Cell 136, 149-162 (2009). 714
50. A. F. Silbering et al., Complementary function and integrated wiring of the evolutionarily 715
distinct Drosophila olfactory subsystems. J Neurosci 31, 13357-13375 (2011). 716
51. R. Rytz, V. Croset, R. Benton, Ionotropic receptors (IRs): chemosensory ionotropic 717
glutamate receptors in Drosophila and beyond. Insect Biochem Mol Biol 43, 888-897 718
(2013). 719
52. J. A. Butterwick et al., Cryo-EM structure of the insect olfactory receptor Orco. Nature 720
560, 447-452 (2018). 721
53. J. Del Mármol, M. A. Yedlin, V. Ruta, The structural basis of odorant recognition in 722
insect olfactory receptors. Nature 597, 126-131 (2021). 723
54. L. Abuin et al., Functional architecture of olfactory ionotropic glutamate receptors. 724
Neuron 69, 44-60 (2011). 725
55. B. J. Matthews et al., Improved reference genome of Aedes aegypti informs arbovirus 726
vector control. Nature 563, 501-507 (2018). 727
56. E. A. Hallem, J. R. Carlson, Coding of odors by a receptor repertoire. Cell 125, 143-160 728
(2006). 729
57. M. C. Larsson et al., Or83b encodes a broadly expressed odorant receptor essential for 730
Drosophila olfaction. Neuron 43, 703-714 (2004). 731
58. E. M. Neuhaus et al., Odorant receptor heterodimerization in the olfactory system of 732
Drosophila melanogaster. Nat Neurosci 8, 15-17 (2005). 733
59. C. J. McMeniman, R. A. Corfas, B. J. Matthews, S. A. Ritchie, L. B. Vosshall, 734
Multimodal integration of carbon dioxide and other sensory cues drives mosquito 735
attraction to humans. Cell 156, 1060-1071 (2014). 736
60. M. DeGennaro et al., orco mutant mosquitoes lose strong preference for humans and 737
are not repelled by volatile DEET. Nature 498, 487-491 (2013). 738
61. J. I. Raji et al., Aedes aegypti mosquitoes detect acidic volatiles found in human odor 739
using the IR8a pathway. Curr Biol 29, 1253-1262 e1257 (2019). 740
62. M. A. Younger et al., Non-canonical odor coding in the mosquito. BioRxiv 741
DOI_10.1101/2020.11.07.368720v2, (2022). 742
63. D. Task et al., Chemoreceptor co-expression in Drosophila olfactory neurons. BioRxiv, 743
DOI 10.1101/2020.1111.1107.355651 (2021). 744
64. H. K. Gouck, Host preferences of various strains of Aedes aegypti and Aedes simpsoni 745
as determined by an olfactometer. Bull. World Health Organ. 47, 680–683 (1972). 746
65. N. S. Basrur et al., Fruitless mutant male mosquitoes gain attraction to human odor. 747
Elife 9, e63982 (2020). 748
66. H. L. Yeap, N. M. Endersby, P. H. Johnson, S. A. Ritchie, A. A. Hoffmann, Body size 749
and wing shape measurements as quality indicators of Aedes aegypti mosquitoes 750
destined for field release. Am J Trop Med Hyg 89, 78-92 (2013). 751
67. Y. Ze, F. Liu, H. Sun, A. Baker, L. J. Zwiebel, Discrete roles of the Ir76b Ionotropic Co-752
Receptor impact olfaction, blood feeding, and mating in the malaria vector mosquito 753
Anopheles coluzzii. BioRxiv, DOI 10.1101/2021.1107.1105.451160 (2021). 754
68. L. L. Prieto-Godino et al., Evolution of acid-sensing olfactory circuits in Drosophilids. 755
Neuron 93, 661-676.e666 (2017). 756
69. M. Ai et al., Acid sensing by the Drosophila olfactory system. Nature 468, 691-695 757
(2010). 758
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
17
70. F. Kuhn, A. Natsch, Body odour of monozygotic human twins: a common pattern of 759
odorant carboxylic acids released by a bacterial aminoacylase from axilla secretions 760
contributing to an inherited body odour type. J R Soc Interface 6, 377-392 (2009). 761
71. M. A. Younger et al., Non-canonical odor coding ensures unbreakable mosquito 762
attraction to humans. BioRxiv DOI_10.1101/2020.11.07.368720v1, (2020). 763
72. N. Nicolaides, H. C. Fu, G. R. Rice, The skin surface lipids of man compared with those 764
of eighteen species of animals. J Invest Dermatol 51, 83-89 (1968). 765
73. N. Nicolaides, Skin lipids: their biochemical uniqueness. Science 186, 19-26 (1974). 766
74. Z. R. Huang, Y. K. Lin, J. Y. Fang, Biological and pharmacological activities of squalene 767
and related compounds: potential uses in cosmetic dermatology. Molecules 14, 540-554 768
(2009). 769
75. T. Kaneda, Iso- and anteiso-fatty acids in bacteria: biosynthesis, function, and 770
taxonomic significance. Microbiol Rev 55, 288-302 (1991). 771
76. Y. J. Lu, Y. M. Zhang, C. O. Rock, Product diversity and regulation of type II fatty acid 772
synthases. Biochem Cell Biol 82, 145-155 (2004). 773
77. J. Oh et al., Temporal stability of the human skin microbiome. Cell 165, 854-866 (2016). 774
78. E. M. Acosta et al., Bacterial DNA on the skin surface overrepresents the viable skin 775
microbiome. BioRxiv, DOI 10.1101/2021.1108.1116.455933 (2021). 776
79. R. C. Muirhead-Thomson, The distribution of anopheline mosquito bites among different 777
age groups; a new factor in malaria epidemiology. Br Med J 1, 1114-1117 (1951). 778
80. K. E. Kistler, L. B. Vosshall, B. J. Matthews, Genome engineering with CRISPR-Cas9 in 779
the mosquito Aedes aegypti. Cell Rep 11, 51-60 (2015). 780
81. Y. Ye, F. Liu, H. Sun, A. Baker, L. J. Zwiebel, Discrete roles of the Ir76b Ionotropic Co-781
Receptor impact olfaction, blood feeding, and mating in the malaria vector mosquito 782
Anopheles coluzzii. BioRxiv, DOI 10.1101/2021.1107.1105.451160 (2021). 783
82. C. O. Fritz, P. E. Morris, J. J. Richler, Effect size estimates: current use, calculations, 784
and interpretation. J Exp Psychol Gen 141, 2-18 (2012). 785
83. Z. Pang et al., MetaboAnalyst 5.0: narrowing the gap between raw spectra and 786
functional insights. Nucleic Acids Res 49, W388-W396 (2021). 787
84. U. Omasits, C. H. Ahrens, S. Muller, B. Wollscheid, Protter: interactive protein feature 788
visualization and integration with experimental proteomic data. Bioinformatics 30, 884-789
886 (2014). 790
791
Materials and Methods 792
Human and animal ethics statement 793
Blood-feeding procedures and mosquito behavior with live human hosts were approved and 794
monitored by The Rockefeller University Institutional Review Board (IRB protocol LVO-0652) 795
and the Rockefeller University Institutional Animal Care and Use Committee (IACUC protocol 796
20068-H). Human subjects gave their written informed consent to participate in this study. 797
798
Human subject information 799
64 healthy human subjects participated in this study. Age at inception of study on 12/7/2017: 800
mean 29.8, median 29, range 19-57 years. Self-identified gender: 37 female, 26 male, 1 non-801
binary. Due to sample size limitations, we intentionally did not subdivide these groups further 802
to investigate the contribution of demographic factors such as sex, age, and ethnicity, or 803
behavioral factors such as diet, personal care, or activity levels on mosquito attractiveness, 804
which we determined empirically. The behavior experiments in Figures 1-5, and Figure 8 805
included only subjects from the initial cohort (Subjects 19, 24, 25, 28, 30, 31, 32, 33). Seven of 806
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
18
these subjects were tested in GC/QTOF-MS Experiments 1.1-1.4 in Figure 6. Subject 24 had 807
moved away before we performed the experiments in Figures 5-7. These 7 subjects from the 808
initial cohort (Subjects 19, 25, 28, 30, 31, 32, 33) were also analyzed about a year later 809
alongside 56 newly recruited subjects (Figure 7B). The full single stimulus olfactometer 810
behavior dataset for all 63 subjects is available on Zenodo (DOI: 10.5281/zenodo.5822539), 811
including data for the 18 subjects whose data are plotted in Figure 7B, comprising 4 subjects 812
from the initial cohort (Subjects 19, 28, 31, and 33) and 14 subjects from the validation cohort. 813
These 18 subjects were selected for inclusion in GC/QTOF-MS Experiments 2.1-2.4, based on 814
their overall level of attractiveness to mosquitoes in initial screening experiments (a subset of 815
the data presented in (Figure 7B), and their availability to participate in additional behavior and 816
GC/QTOF-MS experiments. 817
818
Human skin odor collection 819
Human subjects washed their forearms with Dove unscented soap and water, dried them with 820
clean laboratory paper towels, and then wore nylons sleeves on both forearms between the 821
wrist and elbow for 6 hours. Nylon sleeves were prepared by using scissors to remove 2” of 822
fabric from the tip of the stocking foot of knee highs (L’eggs brand Everyday, Amazon), so that 823
the modified stocking was open on both ends. Subjects wore 2 nylon sleeves on each arm: a 824
brown experimental nylon was worn next to the skin, and a black outer nylon was placed over 825
the experimental nylon to minimize contamination of the inner nylon. Subjects wore nylons 826
during the day and were allowed to perform typical daytime activities but were asked not to 827
exercise or drink alcohol while wearing the nylons. After the 6-hour wearing period, nylons 828
were deposited in Whirl-Pak bags and kept at -20oC for 1-10 days before behavioral and 829
chemical analysis. For the “round-robin” two-choice assay experiment in Figure 1E, we 830
competed 2 nylon sleeves that had been worn on the same day by 2 different human subjects, 831
to strictly control the “age” of the nylons being used to determine if mosquitoes preferred one 832
subject over the other. We relaxed the requirement that subjects wear the nylons on the same 833
day for subsequent two-choice assay experiments, since small differences in nylon age did not 834
change the preference of mosquitoes for specific human subjects. In Figures 2,3,5, and 8, we 835
compared nylons that had been worn within 2 days of each other by two different subjects. 836
This allowed us to better accommodate the schedules of human subjects, who were not 837
always available on the same day over the extended duration of the study. 838
839
Mosquito rearing and maintenance 840
Aedes aegypti wild-type laboratory strains (Orlando and Liverpool) were reared in an 841
environmental room maintained at 70-80% relative humidity and 25-28°C, as previously 842
described (60). All animals were maintained with a photoperiod of 14 hours light: 10 hours dark 843
throughout larval, pupal, and adult life stages. Adult mosquitoes were provided constant 844
access to 10% sucrose. Female mosquitoes were fasted for 14-24 hours without sucrose in 845
the presence of water prior to behavioral experiments. For stock maintenance, females were 846
blood fed on live mice. 847
848
Ir25a and Ir76b mutant strain generation 849
Ir25a and Ir76b mutants were generated using methods described previously (80). sgRNA 850
sequences were designed with the CRISPOR v4.3 sgRNA design tool (http://crispor.tefor.net/) 851
(Concordet, 2018) using the following parameters: Genome, Aedes aegypti – yellow fever 852
mosquito – NCBI GCF_002204515.2 (AaegL5.0); Protospacer Adjacent Motif (PAM), 20bp-853
NGG - Sp Cas9, SpCas9-HF1, eSpCas9 1.1. For each gene, two pairs of sgRNA with 854
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
19
predicted MIT Specificity Scores ≥95 were selected for targeted double stranded break-855
induced mutagenesis, with each pair flanking roughly 250 base pairs within exon 2 for Ir76b 856
and exons 2 and 3 for Ir25a. sgRNA DNA templates were prepared by annealing 857
oligonucleotides as previously described using the following target sequences: 858
Ir25a-sgRNA1: GTTGAGCTACTAACCGTCGA 859
Ir25a-sgRNA2: TACTGACAGCAAAGGGCTGT 860
Ir25a-sgRNA3: CCTACGGTTTCCGCATCAAC 861
Ir25a-sgRNA4: AAGAAGGCGACTTGAGGCAA 862
Ir76b-sgRNA1: GTTACACCGAACGTCAGAA 863
Ir76b-sgRNA2: TACTCTGGTCGGACGCGGTG 864
Ir76b-sgRNA3: CTCCTTTCAATCGGGACGTG 865
Ir76b-sgRNA4: CAACGGCCAGCAGCGATACC 866
In vitro transcription was performed using HiScribe Quick T7 kit (NEB E2050S) following the 867
manufacturer’s protocol. Following in vitro transcription and DNAse treatment for 15 minutes at 868
37oC, sgRNA was purified using RNAse-free SPRI beads (Ampure RNAclean, Beckman-869
Coulter A63987), and eluted in Ultrapure water (Invitrogen, 10977–015). To further facilitate 870
isolation of loss-of-function mutants, 200 bp single-stranded DNA oligodeoxynucleotide 871
(ssODN) donors were designed as a template for homology-directed repair (80). The ssODN 872
donor had homology arms of 88–90 bases on either side of the inner most sgRNA target sites 873
(sgRNA2 and sgRNA3 for both Ir25a and Ir76b), flanking an insert with stop codons in all three 874
frames of translation and a BamHI restriction site (IDT). Since homology arms contained 875
sgRNA target sites for the outer most sgRNA target sites (sgRNA1 and sgRNA4 for both Ir25a 876
and Ir76b), PAM motifs for the respective sgRNAs were mutated to avoid ssODN donor 877
cleavage. The sequences for each ssODN follow below with left and right homology arms 878
italicized, stop codons underlined, and BamHI restriction site highlighted in bold: 879
Ir25a-ssODN: 880
TTGCGCTGAACTATATAAGAAAGAACCCAAGCCTCGGACTTTCAGTTGAGCTACTAACCG881
TCGAATGAAACCGTACTGACAGCAAAGGGCTAATAAGGATCCATAACTAAGGAACCGGCC882
AAGAAGGCGACTTGAGGCAATAGCGATCTCTATCAAACGTAAAAAGCAACTATCTGTTGC883
AGGTTTGCTACAAAACTTTA 884
Ir76b-ssODN: 885
TCTAATTGCATCGAACTCTCTTTCCCACGTTCAACAGGACTGGCCGCTGAGTTACACCGA886
ACGTCAGAATAGTACTCTGGTCGGACGCGTAATAAGGATCCATAACTAAGGGTGTGGATT887
TTGATTCTGGTATCGCTGCTGGCCGTTGGTCCAATCATCTACGGAATGCTGATTGTGCGG888
TACAAAATGACCAAAGACAA 889
For each target gene, approximately 500 wild-type Aedes aegypti (Liverpool LVP-IB12 strain) 890
embryos were injected with a mixture containing recombinant Cas9 protein (PNA Bio, CP01) at 891
300 ng/µl, 4 sgRNAs at 40 ng/µl each, and donor ssODN at 125 ng/µl. Embryos were injected 892
by the Insect Transformation Facility at the University of Maryland Institute for Bioscience & 893
Biotechnology Research. Embryos were hatched and G0 females were crossed to wild-type 894
Liverpool males, and their G1 offspring were screened for germline mutation by PCR 895
amplification and Sanger sequencing the regions flanking the sgRNA target sites. Two unique 896
stable mutant lines, each resulting in an early stop codon due to a frameshift mutation, were 897
isolated from each injection. For Ir25a, one isolated mutant line had a 209-bp deletion with an 898
ssODN integration (Ir25aBamHI), and the other had a 19-bp deletion (Ir25a19). For Ir76b, one 899
isolated mutant line had a 61-bp deletion (Ir76b61), and the other had a 32-bp deletion 900
(Ir76b32). Virgin females from each mutant line were backcrossed to wild-type Liverpool males 901
for 8 generations prior to establishment of stable homozygous lines. To control for potential 902
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
20
CRISPR-Cas9 off-target effects on behavior, homozygous mutant lines were intercrossed to 903
generate heteroallelic mutants that were tested in all behavior experiments alongside 904
appropriate genetic controls. It was previously shown that although Ir76b mutant Anopheles 905
coluzzii females show normal attraction to human host cues, they fail to blood feed and 906
therefore produce no offspring (81). We also observed deficits in blood-feeding and egg-laying 907
in homozygous Aedes aegypti Ir25a and Ir76b strains, albeit far less severe than the 908
Anopheles coluzzii Ir76b mutants. All homozygous Ir25a and Ir76b mutant strains were 909
maintained by blood-feeding on a human arm. We determined empirically that these mutant 910
mosquitoes fed more avidly when they were 14 days post eclosion, rather than 7 days, which 911
is the standard timepoint at which we blood feed to propagate other strains. These strains did 912
not feed effectively on live mice or on an artificial membrane blood feeder. A volunteer inserted 913
an arm into a standard BugDorm rearing cage (30 cm3) cage for 30 minutes at ambient 914
temp/humidity. These cages contained a mixed population of males and females, which had 915
been allowed to mate freely since eclosion. Mosquitoes had access to the entire hand and 916
forearm of the subject. For context, normal strains feed to repletion on a human arm after 5-10 917
minutes, even when the arm is placed outside of the cage netting and not inside the cage. 918
Feeding of these mutants was unsuccessful when the human arm was placed against the 919
netting on the outside of the cage. About 50% of females were engorged at the end of a given 920
30-minute feeding session, while the rest were partially fed or unfed. Notably, a given female 921
took much longer to probe the arm repeatedly before successfully feeding. No attempt was 922
made to refeed females that did not feed during a given 30-minute feeding session. Four days 923
after feeding, an oviposition cup was placed into the cage and as many eggs as possible were 924
collected over a 4-day period. After this stage the entire process was repeated once or twice 925
with the same cage to obtain enough eggs to propagate the strains and carry out experiments. 926
It is our impression that Ir25a females but not Ir76b females laid fewer eggs than wild type, but 927
we did not investigate this further in the course of this study. Unlike the observed difficulties 928
with the homozygous mutants, heterozygous mutants fed normally and laid normal numbers of 929
eggs. 930
931
Ir25a and Ir76b mutant genotyping 932
Genotypes were confirmed using Phire Tissue Direct PCR Master Mix (Thermo Fisher, F170L) 933
followed by gel electrophoresis and Sanger sequencing (Genewiz) using the following primer 934
combination for each mutant allele: 935
Ir25aBamHI, Ir25a19 936
Forward: AATACTTGAGGAGTCGTTGAAT 937
Reverse: GAAGCAATGCCTTGTACTTATG 938
Ir76b61 939
Forward: AGCCGAATATGAAGGTCAAGC 940
Reverse: CAGCACCTGTTCCTTGTCTT 941
Ir76b32 942
Forward: TGCATCGAACTCTCTTTCCC 943
Reverse: CGATAGCTAAGATGCCAGTACAT 944
The Ir25aBamHI allele was detected by either a 159 bp deletion or presence of an exogenous 945
BamHI restriction site from the donor ssODN. BamHI restriction digest of the PCR product 946
generated a single 764 bp fragment in wild-type animals and two fragments (275 bp and 329 947
bp) in the Ir25aBamHI mutant. For mutants with small deletions, the presence or absence of 948
endogenous restriction enzyme target sites was used to distinguish between mutant and wild-949
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
21
type alleles. PCR products were generated and digested with the indicated enzyme, producing 950
the indicated bands in mutant and wild type: 951
Ir25a19 with MspI 952
Wild type: 502 and 262 bp; Ir25a19: 764 bp 953
Ir76b61 with BstUI 954
Wild type: 367 and 397 bp; Ir76b61: 764 bp 955
Ir76b32 with BstNI 956
Wild type: 392 and 365bp; Ir76b61: 757 bp 957
All genotyping experiments were performed with a no DNA control as well as fragment size 958
validation using 1Kb Plus DNA ladder (ThermoFisher Scientific, 10787026). See 959
Supplementary Figure S1. 960
Given the severe host-seeking and blood-feeding deficits displayed by Ir25a and Ir76b female 961
homozygous mutants, it is difficult to maintain these as homozygous strains. For the benefit of 962
scientists wishing to work with these new strains, we have devised a crossing scheme that 963
uses heterozygous mutant females to propagate the mutant alleles. An important aspect of this 964
approach is that Ir25a and Ir76b mutants do not carry a fluorescent marker at their respective 965
gene loci. This strategy consists of first crossing homozygous Ir25a or Ir76b mutant males to 966
heterozygous females from the corresponding Ir25a-QF2 and Ir76b-QF2 gene-sparing knock-967
in driver lines (71). These strains contain the 3xP3-dsRed marker integrated at the Ir25a or 968
Ir76b genetic locus. This initial cross will generate both fluorescent and non-fluorescent 969
heterozygous mutants at a 1:1 ratio. Then, taking the fluorescent heterozygous mutant 970
females, these animals will again be crossed to the homozygous mutant males. This cross will 971
generate non-fluorescent homozygous mutants and fluorescent heterozygous mutants at a 1:1 972
ratio. At this point, mutant alleles can easily be maintained by collecting eggs from the cross 973
between non-fluorescent homozygous mutant males with fluorescent heterozygous females 974
isolated at each generation. Since the mutations and the locations of the inserted fluorescent 975
markers are tightly linked, we expect the recombination rates between the mutation sites and 976
the markers to be extremely rare. However, occasional genotyping is recommended to ensure 977
proper propagation of each of the mutant alleles. 978
979
Behavioral assays 980
The single choice olfactometer assay referred to in this work (Figure 4C, Figure 7B) is the 981
same assay previously referred to as the “Quattroport” in an earlier publication (65), because 982
the assay allows 4 independent single stimulus olfactometer trials to be run in parallel. In this 983
work, we refer to the Quattroport assay as the “single stimulus olfactometer assay” to avoid 984
confusion about the number of stimuli being presented to each group of mosquitoes in a single 985
trial. We repurposed components of the Quattroport assay to create the two-choice 986
olfactometer assay, which allows us to compete 2 different stimuli against each other in the 987
same trial. We performed two separate preference trials in parallel, increasing throughput over 988
the Gouck olfactometer assay (64). Details of fabrication and operation of the two-choice 989
assay are available on Zenodo (DOI: 10.5281/zenodo.5822539). Air flow and CO2 conditions 990
for the two-choice olfactometer assay were carried as described for the Quattroport (65). 991
Human forearm-worn nylon sleeves were used as the stimulus in all behavior figures with the 992
two-choice olfactometer assay, except Figure 1B in which human subjects placed their forearm 993
over a hole in the stimulus box lid, exposing 12.9 cm2 of skin to mosquitoes (demonstrated 994
with a mannequin arm in Figure 1A). We shuffled the order in which different stimuli were 995
assessed over the course of the day, and we randomized the position of different stimuli 996
across all stimulus boxes of the assays, to reduce time of day and position effects. All 997
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
22
behavioral experiments were carried out in an environmental room set to 25oC, 70-80% 998
relative humidity. The day before behavior was measured, 20 female mosquitoes (aged 7-14 999
days post-eclosion, mated) were sorted under cold anesthesia into each start canister and 1000
given access only to water for 18-22 hours. The same canisters were used for the single 1001
stimulus and two-choice assays. However, twice as many females were tested in each two-1002
choice assay trial (40 females per trial, 20 in each of 2 canisters) as in each single stimulus 1003
trial (20 females per trial). In both assays, mosquitoes were acclimated to a carbon-filtered air 1004
stream for 10 minutes (Donaldson Ultrac-A). CO2 was then introduced into the air stream for 1005
30 seconds, at which point mosquitoes were released, and given 5 minutes to assess the 1006
stimulus or stimuli. Mosquitoes were prevented from contacting stimuli by a mesh divider. 1007
Sliding doors between assay compartments allow the experimenter to count the number of 1008
mosquitoes that were not activated, or activated but not attracted, or both activated and 1009
attracted to the stimulus. In the single stimulus assay, mosquitoes that entered an attraction 1010
trap were scored as attracted to the stimulus. In the two-choice assay, mosquitoes that entered 1011
a cylindrical flying tube or the adjacent trap, both downwind of the stimulus, were scored as 1012
attracted to that stimulus. In both assays, all mosquitoes that left the start canister were scored 1013
as activated. We are confident that mosquitoes use olfactory information to detect nylon 1014
stimuli, since mosquitoes cannot contact or taste the nylon in our assays due to the presence 1015
of a mesh barrier, and there are no visual cues that differentiate nylons worn by different 1016
subjects. 1017
1018
Cleaning behavioral assay apparatus 1019
Between trials, the assay apparatus was vacuumed to remove live and dead mosquitoes, and 1020
air was flowed through the assay for 5-10 minutes to flush out residual CO2 and odor. Two-1021
choice assay parts were cleaned before use in every experiment, as described below. In some 1022
cases, assay parts needed to be cleaned between trials, so that they could be re-used during 1023
the same behavior experiment. Whenever possible, we constructed enough replicate assay 1024
parts to reduce the need to wash parts between trials, which slows down assay throughput. 1025
Experimenters always wore gloves when cleaning and handling clean assay parts. Detailed 1026
procedures are described here for how and when we cleaned each part of the two-choice 1027
assay (going from left to right in the schematic in Figure 1A: 1) start canisters, 3D-printed 1028
connector joints, and accompanying acrylic sliding doors were washed in a dishwasher (Miele 1029
Optimal Series dishwasher, Cascade Original “Actionpacs” detergent pods) at least 2 days 1030
before behavior and allowed to air dry completely before female mosquitoes were loaded into 1031
the canisters on the day before behavior; 2) the flying box was too big to be washed in the sink 1032
or dishwasher, so the inside of the box was sprayed down with 70% ethanol from a laboratory 1033
spray bottle and this was wiped down with laboratory paper towels, and allowed to air dry; 3) 1034
two cylindrical flying tubes were washed in the dishwasher as described above and allowed to 1035
air dry; 4) Acrylic stands that support the cylindrical flying tubes were washed in the 1036
dishwasher as described above; 5) the complete supply of attraction traps and accompanying 1037
3D printed joins and acrylic sliding doors were washed in the dishwasher as described above 1038
at least one day before the experiment and allowed to air dry. When we performed more trials 1039
than we had traps available, we hand washed the traps used in the first few trials using hot 1040
water and soap (Bac Down Handsoap, Decon Labs, Inc., Catalog #: 7001), and allowed them 1041
to air dry before re-using them in trials later in the day. 6) For stimulus boxes and lids, the 1042
cleaning procedure was the same as that for the traps. The complete supply of stimulus boxes 1043
and lids was washed in the dishwasher as described above at least one day before the 1044
experiment and allowed to air dry. When we performed more trials than we had stimulus boxes 1045
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
23
available, we hand washed the stimulus boxes used in the first few trials using hot water and 1046
soap and allowed them to air dry before re-using them in trials later in the day. 7) The 3D 1047
printed stop piece which connects the air/CO2 supply to the stimulus box was wiped down with 1048
70% ethanol that had been sprayed onto a laboratory paper towel before the first trial, and 1049
between every trial. Similarly, the single stimulus olfactometer assay parts were cleaned 1050
before every experiment. This means that, going from left to right in the schematic in Figure 1051
4C: 1) start canisters, 3D-printed connector joints, and accompanying acrylic sliding doors 1052
were washed in a dishwasher at least 2 days before behavior and allowed to air dry completely 1053
before female mosquitoes were loaded into the canisters on the day before behavior; 2) the 1054
two cylindrical flying tubes were hand washed in the sink with hot water and soap and a bottle 1055
brush (Dr. Brown’s, Amazon, ASIN: B01NCUKCC0), and allowed to air dry ; 3) Acrylic stands 1056
that support the cylindrical flying tubes were washed in the dishwasher as described above; 4) 1057
the complete supply of attraction traps and accompanying 3D printed joins and acrylic sliding 1058
doors were washed in the dishwasher as described above at least one day before the 1059
experiment and allowed to air dry. When we performed more trials than we had traps available, 1060
we hand washed the traps used in the first few trials using hot water and soap and allowed 1061
them to air dry before re-using them in trials later in the day. 5) For stimulus boxes and lids, the 1062
cleaning procedure was the same as that for the traps. The complete supply of stimulus boxes 1063
and lids were washed in the dishwasher as described above at least one day before the 1064
experiment and allowed to air dry. When we performed more trials than we had stimulus boxes 1065
available, we hand washed the stimulus boxes used in the first few trials using hot water and 1066
soap and allowed them to air dry before re-using them in trials later in the day. 6) The 3D 1067
printed stop piece which connects the air/CO2 supply to the stimulus box was wiped down with 1068
70% ethanol that had been sprayed onto a laboratory paper towel before the first trial, and 1069
between every trial. 1070
1071
Behavior inclusion criteria 1072
Two-choice olfactometer assay data indicating the overall percent mosquito activation and 1073
attraction in response to a single human subject stimulus, are presented in Figure 2A, Figure 1074
3A, and Figure 5A for all control trials performed (i.e. all trials examining Subject 25 vs Subject 1075
25, before the application of any inclusion criteria). In Figure 2B, Figure 3B, and Figure 5B, we 1076
report the percent of control trials that met the inclusion criteria used in later parts of these 1077
Figures, which required that there were: 1) at least 30 live mosquitoes at the end of the assay 1078
and 2) at least 10 mosquitoes attracted to either stimulus. We chose these inclusion criteria 1079
because IR mutants displayed large defects in overall attraction to human subjects, and we 1080
wished to examine preferences specifically using trials which passed a minimal threshold of 1081
overall attraction to human odor. Trials with very low overall attraction to either stimulus, could 1082
give misleading results, because they are subject to “jackpotting” effects. For example, if 2 1083
mosquitoes were attracted to Subject A, and 1 mosquito was attracted to Subject B, this is not 1084
a meaningful difference in preference, so we would exclude this trial for having <10 animals 1085
attracted to either subject. To assay mosquito preferences between genotypes in Figure 2C-D, 1086
Figure 3C-D, and Figure 5C-E, we present only trials which passed the inclusion criteria. We 1087
discarded trials in which substantially fewer mosquitoes were loaded into the assay 1088
(unintentionally), and which very few mosquitoes (<10) were attracted to either stimulus. There 1089
were some exceptions to the application of inclusion criteria, described here: the experiment in 1090
Figure 1C compared mosquito preferences between several stimulus pairs that were expected 1091
to (and did) result in very low levels of attraction. These stimuli were: 1) unworn nylons versus 1092
no stimulus, and 2) no stimulus versus no stimulus. For this experiment, we did not require that 1093
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
24
more than 10 mosquitoes were attracted to either stimulus. We only required that there were 1094
>30 live animals at the end of the trial. For single stimulus olfactometer experiments (Figure 1095
4D-G, Figure 7B), we included trials with >14 live mosquitoes at the end of the trial. 1096
Nylon sleeve behavioral assay stimuli 1097
For human odor host-seeking assays, a 7.62 cm x 10.16 cm piece of the brown experimental 1098
nylon was cut and used as a stimulus in either the two-choice olfactometer assay or the single 1099
stimulus olfactometer assay. This was the largest size of material that could be laid flat on the 1100
bottom of the stimulus boxes used in both assays. By using this size of nylon stimulus, we 1101
could cut 3 pieces of nylon from each nylon sleeve, yielding a total of 6 pieces of nylons from 1102
each day of nylon wearing, or enough for 6 trials of that subject. Experimenters always 1103
handled nylon sleeves with gloves and cleaned scissors with 70% ethanol between samples. 1104
Subjects wore 2 nylon sleeves on each of their 2 forearms (one brown experimental nylon next 1105
to their skin, covered by one black protective nylon, as described above). We asked subjects 1106
to remove both nylon sleeves they wore on each arm as a single “unit” (keeping the black 1107
protective nylon outside the brown experimental nylon) and to place these in a Whirl-Pak bag 1108
(Nasco, Catalog #: B01062) in the freezer. This allowed us to determine which side of the 1109
brown experimental nylon had touched the subject’s skin (the side that was not touching the 1110
black outer nylon), and we marked this side with a fine point permanent marker. The side of 1111
the nylon that touched the subject’s skin was placed facing upwards in the stimulus box, to 1112
ensure consistency between trials. In all behavior experiments, nylon samples were de-1113
identified before being cut and presented to mosquitoes in the behavioral assay (either the 1114
two-choice assay or the single stimulus olfactometer assay), such that experimenters were 1115
blinded to the Subject ID. Each piece of nylon was attached to a flexible plastic rectangle of the 1116
same size (cut with scissors from plastic file folders: Letter size, Office Depot, Catalog #: 1117
700259), using small binder clips (3/4” size, Office Depot Catalog #: 808857). The color of the 1118
plastic rectangle indicated the de-identified label given to that nylon, corresponding to the 1119
source of the nylon. For instance, in a given experiment, Subject 33 was designated as 1120
Subject “A” by someone other than the experimenter, and then the experimenter cut the de-1121
identified nylon pieces and attached all Subject “A” nylons to red plastic cards. The color of the 1122
card was chosen randomly for each experiment, and colors rotated between subjects. Each 1123
nylon piece was used for only one trial and then discarded. Binder clips and plastic cards were 1124
washed with soap and water as described above and set out to air dry at the end of each 1125
experiment. 1126
1127
Statistical analysis of behavior 1128
Statistical analyses were performed using R. Two-choice assay data were analyzed by 1129
comparing the percent of mosquitoes attracted to each of two stimuli, for several pairs of 1130
stimuli, using Wilcoxon rank-sum tests with Bonferroni correction (Figure 1B-E, Figure 2C-D, 1131
Figure 3C-D, Figure 5C-E). Nonparametric effect size (ES) was calculated as ES= Z/sqrt(N) 1132
where N is the number of observations (Figure 1E) (82). A power analysis was used to pre-1133
determine the approximate sample size needed for Figure 1E (G*Power software). Wilcoxon 1134
rank-sum tests with Bonferroni correction were used to compare the percent of mosquitoes in 1135
each of three categories (attracted, activated but not attracted, or not activated) between wild 1136
type and each mutant genotype in Figure 2A, Figure 3A, Figure 5A. Single stimulus 1137
olfactometer assay data in Figure 4D-G were analyzed by using a Kruskal-Wallis test to 1138
compare the percentage of mosquitoes activated or attracted across 4 genotypes: including 1139
wild-type controls, 2 heterozygous mutants, and the heteroallelic null mutants. When indicated, 1140
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
25
post hoc analysis used pairwise Wilcoxon rank-sum tests with Bonferroni correction to 1141
compare wild-type mosquitoes to each of the other 3 genotypes for a given stimulus (Figure 1142
4D-G). 1143
1144
Calculation of the “Attraction Score” 1145
The attraction score reported in Figure 1F was calculated as follows. For each of the 28 1146
subject pairs in the “round-robin tournament”, we calculated the total number of mosquitoes 1147
attracted to each subject in the pair, by adding the number that were attracted across every 1148
trial performed for that specific pair in Figure 1E. For example, for the Subject 33 vs Subject 28 1149
pair, we summed the total number of mosquitoes attracted to Subject 33 or to Subject 28 1150
across the 18 trials in which they were compared to each other: Subject 33 attracted 468 1151
mosquitoes, whereas Subject 28 attracted only 35 mosquitoes. We took the difference 1152
between these values (468-35=433) and divided it by 18 trials to get the average “margin of 1153
victory” per trial for this subject pair (433/18=24). Since Subject 33 attracted 24 more 1154
mosquitoes than Subject 28 per trial, on average, Subject 33 was “awarded” 24 points, and 1155
Subject 28 was awarded 0 points. Subject 33 also attracted more mosquitoes than the 1156
remaining 6 other subjects, with average “margin of victory scores” of 15, 18, 20, 21, 23, and 1157
23. The “attraction score” for each subject, reported in Figure 1F, represents the sum of 7 1158
average “margin of victory” values. In the example above, the “attraction score” is calculated 1159
by summing the average “margin of victory scores” for Subject 33, compared to the 7 other 1160
subjects: 15+18+20+21+23+23+24=144. 1161
1162
Sample preparation for gas chromatography/quadrupole time of flight-mass 1163
spectrometry (GC/QTOF-MS) 1164
Nylon sleeves that had been stored at -20°C for 1-10 days were thawed and cut into 5.08 cm x 1165
2.54 cm pieces. Experimenters always handled nylon sleeves with gloves and cleaned 1166
scissors with 70% ethanol between samples. For each subject, 6 separate pieces (5.08 cm x 1167
2.54 cm each) of the same nylon sleeve were analyzed in each of the 4 experiments shown in 1168
Figure 6 (Experiments 1.1-1.4), and 5 pieces of the same nylon sleeve were analyzed in each 1169
of the 4 experiments shown in Figure 7 (Experiments 2.1-2.4). Extraction solvent was 80% 1170
methanol (CID: 67-56-1, Methanol, Fisher Chemical, Catalog #:A456-4) prepared with Millipore 1171
water and spiked with deuterated isotope-labeled internal standards sourced from Cambridge 1172
Isotope Laboratories: nonanoic acid-D17 (CID: 130348-94-6, Catalog #: DLM-9501-0.5), 1173
propanoic acid-D5 (CID: 60153-92-6, Catalog #: DLM-1919-5), phenol-D5 (CID:4165-62-2, 1174
Catalog #:DLM-695-1), acetate-D4 (CID:1186-52-3, Catalog #: DLM-12-10 ), butyrate-D7 (CID: 1175
73607-83-7, Catalog #: DLM-1508-5) and valerate-D9 (CID: 115871-50-6, Catalog #:DLM-572-1176
1). Each piece of nylon was extracted in 1 mL of extraction solvent in a 5 mL Eppendorf tube 1177
(Fisher Scientific Catalog #: 14-282-301). 5 mL Eppendorf tubes containing extraction solvent 1178
and nylon were kept overnight at -20°C. Then, each tube was vortexed for 20 seconds to 1179
ensure thorough extraction of the nylon. A P1000 pipet with standard 1000 µL pipet tip was 1180
used to press the nylon against the side of the 5 mL Eppendorf tube, so that the extraction 1181
solvent was squeezed out of the nylon. Before doing this, the plunger of the pipet had been 1182
pressed down, so that it could be released after squeezing the nylon, to pick up the extraction 1183
solvent, before it could be reabsorbed into the nylon sleeve. Typically, 700-900 µL of nylon 1184
“extract” (a cloudy mixture of extraction solvent plus nylon compounds) was recovered from 1185
each sample, and this was immediately transferred to a 2 mL glass GC vial (Wheaton µL 1186
MicroLiter autosampler vials; 12x32mm, Catalog #: 11-1200) and secured with a septum cap. 1187
1188
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
26
Pentafluorobenzyl bromide (PFB-Br) derivatization 1189
100 µL of the liquid nylon extract was manually transferred using a pipet to a new glass 1190
autosampler vial, containing 100 µL of borate buffer (100mM boric acid CID: 10043-35-3, 1191
Sigma Aldrich, Catalog #: 339067, adjusted to pH 10 using 10 M NaOH (CID: 1310-73-2, 1192
Fisher Chemical, Catalog #: S318500. Subsequent steps were performed using a single 1193
needle liquid handling system (Gerstel MPS). First 400 µL of 100 mM pentafluorobenzyl 1194
bromide (PFB-Br, CID: 1765-40-8, Sigma-Aldrich, Catalog number: 101052), prepared in 1195
acetonitrile (Fisher Chemical; Catalog #: A955-4) and 400 µL cyclohexane (Sigma-Aldrich, 1196
Catalog #: 650455-4L) were added sequentially to the reaction vials. Samples were then 1197
sealed using crimp caps (Wheaton MicroLiter Catalog #: 11-0040A) and heated to 65°C with 1198
shaking for 1 hour to esterify carboxyl and hydroxyl groups (Supplementary Figure S3A). 1199
Samples were allowed to cool to room temperature and centrifuged for 2 minutes at 2,000 rpm, 1200
20°C to promote phase separation, then returned to the robotic sample preparation instrument. 1201
Finally, the upper cyclohexane layer was transferred to clean autosampler vials (Wheaton 1202
MicroLiter, Catalog #: 11-1200) and 50 µL further transferred into a second clean autosampler 1203
vial pre-filled with 450 µL cyclohexane, creating a 10-fold dilution of the derivatized nylon 1204
extracts that was subsequently used for GC/QTOF-MS analysis. 1205
1206
GC/QTOF-MS data collection 1207
GC/QTOF-MS data was collected using an Agilent 7890B gas chromatograph (GC) and an 1208
Agilent 7200 quadrupole time-of-flight (QTOF) mass spectrometer (MS), fitted with a Gerstel 1209
MPS Robotic Autosampler. 1 µL of the sample volume was injected in splitless mode onto a 1210
DB5ms column (30 m × 250 μm, 0.25 μm film thickness; Agilent Technologies, Catalog 1211
#:19091S-433). GC conditions were as follows: initial oven temperature 60°C (hold 1 minute), 1212
ramp at 25°C/minute to 300°C (hold 2.5 minutes), ramp at 120°C/minute to 60°C (hold 1 1213
minute); the total run time was 16.1 minutes. The inlet, transfer line, chemical ionization (CI) 1214
source, and quadrupole temperatures were set to 280°C, 300°C, 150°C, and 150°C 1215
respectively. The emission current was set to 4.2 µA and data collection was in 2 GHz mode, 1216
mass range m/z 50-650. For all GC/QTOF-MS experiments, the sample injection order was 1217
randomized using the “RAND” function in Microsoft Excel. The same GC column was used for 1218
data acquisition shown in Figure 6 and Figure 7, column trimming as part of routine 1219
maintenance was responsible for slightly shorter retention times shown in Figure 7. The Q-1220
TOF mass spectrometer was operated in negative chemical ionization (nCI) mode with 1221
methane as the reagent gas (1 mL/minute); in nCI mode, derivatized molecules undergo 1222
electron capture dissociation (ECD) and are detected as deprotonated (M-H) ions. 1223
1224
GC/QTOF-MS data analysis 1225
The workflow for GC/QTOF-MS data analysis is summarized in Supplementary Table S1. Raw 1226
data (Agilent “.D” files) were analyzed using Agilent MassHunter software. Untargeted feature 1227
finding was performed using Agilent Unknowns Analysis software (v10.1) (Supplementary 1228
Table S1, step 1). Specifically, data was first converted using SureMass peak detection 1229
(absolute height filter >1000 counts, RT window size factor=50, Extraction window +-50ppm, 1230
use base peak shape=yes, sharpness threshold 25%, Ion peaks minimum=1, maximum=1). 1231
Subsequently, feature finding was performed on all worn nylon samples across 4 replicate 1232
experiments for Experiments 1.1-1.4 (Figure 6) and, again for Experiments 2.1-2.4 (Figure 7). 1233
Putative features, denoted as accurate mass @ retention time, were exported from Unknowns 1234
Analysis as a .csv file, and imported into R. A custom deduplication script was used to remove 1235
most “fuzzy” duplicate features with similar mass and retention time values, and features 1236
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
27
detected in <10 data files, resulting in a list of 1494 features for the experiments in Figure 6, 1237
and 1925 features for the experiments in Figure 7 (Supplementary Table S1, step 2). This 1238
feature list was imported into Agilent Quantitative Analysis software (v10.1) and a targeted 1239
data re-extraction performed (settings: +Gaussian smoothing, +area filter ≥ 0, RT window: ± 1240
0.05 minutes extraction window ± 100ppm) (Supplementary Table S1, step 3), resulting in a 1241
fully aligned dataset across 4 replicate experiments. Raw peak areas were then exported as a 1242
.csv file, and a custom R script was again used to remove features present in <10% of 1243
samples, impute missing values with 1/2 the lowest value for that feature, and log2 transform 1244
data (Supplementary Table S1, step 4). The resulting .csv file was imported into Agilent Mass 1245
Profiler Professional software (v15.1). We considered high quality features to be those found in 1246
at least 50% of samples in at least 1 subject group, with a coefficient of variation <40% in at 1247
least 1 subject group. Putative human skin-derived features were considered those at least 2-1248
fold enriched (false discovery rate, FDR<0.05) in any Subject’s worn nylons versus unworn 1249
nylons and solvent blank controls (Supplementary Table S1, step 5). Volcano plots were used 1250
to filter on features differentially abundant in the high versus low attractor subjects in each 1251
individual experiment of 4 replicate experiments (Supplementary Table S1, step 6; Figure 6D, 1252
Figure 7D). Venn diagrams were used to identify hits that met the differential abundance 1253
criteria in high versus low attractors across all 4 replicate experiments (Supplementary Table 1254
S1, step 7): Experiments 1.1-1.4 (Figure 6E) and Experiments 2.1-2.4 (Figure 7E). Redundant 1255
features with closely matching mass and retention times, that were missed by the earlier 1256
deduplication step were removed manually, after inspecting the raw data in Agilent Qualitative 1257
Analysis software (Supplementary Table S1, step 8). Data File S1 contains lists of mass and 1258
retention times for unknown features that were found in 4 replicate experiments, either 1259
Experiments 1.1-1.4 or Experiments 2.1-2.4. For all targeted data re-extractions, 100 ppm 1260
mass accuracy was used to extract the detected pseudomolecular [M-H]- ion, unless the peak 1261
was considered saturated, in which case the [(M+1)-H]- isotope was used consistently across 1262
all data files. For Experiments 1.1-1.4, we calculated the median abundance of each 1263
compound in each of 4 experiments for each of 2 high attractor subjects and 2 low attractor 1264
subjects (Supplementary Figure S2B). Then we compared these values for each compound 1265
between the high and low attractor groups using a nonparametric linear mixed effects model 1266
for repeated measures based on ranks, with group as a fixed effect and subject as a random 1267
effect. Benjamini-Hochberg correction was applied for multiple comparisons (FDR<0.1). For 1268
Experiments 2.1-2.4, we calculated the median abundance value for each compound in each 1269
subject across 4 replicate experiments, and then we compared these values between the high 1270
(n=11) vs low attractor (n=7) groups (Figure 7H) using a Wilcoxon rank-sum test. Benjamini-1271
Hochberg correction was applied for multiple comparisons (FDR<0.1). Statistical analysis was 1272
performed in R version 4.0.5. Heatmaps were generated using Metaboanalyst (v5.0; 1273
https://www.metaboanalyst.ca/) (83) using autoscaled data, clustering was Ward’s linkage and 1274
Euclidean distance measure. 1275
1276
GC/QTOF-MS compound identification 1277
To identify compounds of interest, we used the accurate mass of the molecular ion (M-H) to 1278
predict chemical formulas using Agilent Qualitative Analysis software (v10.0). Formula 1279
prediction was constrained to C, H, O, N atoms, and formulas within 100 ppm mass error of 1280
the measured mass were considered (Supplementary Table S1, step 9). Predicted formulas for 1281
unknown compounds are included in Data File S1. To achieve positive identification, we spiked 1282
nylon samples with authentic standards (details below) and asked whether the area of the 1283
unknown peak increased symmetrically (Supplementary Table S1, step 10). Overlaid extracted 1284
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
28
ion chromatograms generated in Agilent Qualitative Analysis software for 3 hit compounds are 1285
shown in Supplementary Figure S3G. Commercially standards used were: propanoic acid 1286
(CID: 5818-15-5, Millipore Sigma; Catalog #: 94425-1ML-F), butyric acid(CID: 107-92-6; 1287
Millipore Sigma; Catalog #: 19215-5ML), pentanoic acid (CID: 109-52-4; Millipore Sigma; 1288
Catalog #: 75054-1ML), hexanoic acid(CID: 142-62-1; Millipore Sigma; 21529-5ML), heptanoic 1289
acid (CID: 111-14-8; Millipore Sigma; Catalog #: 43858), octanoic acid (CID: 124-07-2; 1290
Millipore Sigma; Catalog #:21639), nonanoic acid (CID: 112-05-0; Millipore Sigma; Catalog 1291
#:73982), decanoic acid (CID: 334-48-5; Millipore Sigma; Catalog #:21409), undecanoic acid 1292
(112-37-8; Millipore Sigma; Catalog #:89764), dodecanoic acid (143-07-7; Millipore Sigma; 1293
Catalog #:61609), tridecanoic acid (CID: 638-53-9; Millipore Sigma; Catalog #: 91988), 1294
tetradecanoic acid (CID: 62217-71-4; Millipore Sigma; Catalog #:70079), pentadecanoic acid 1295
(CID: 1002-84-2; Millipore Sigma; Catalog #:91446), hexadecanoic acid (CID: 57-10-3; 1296
Millipore Sigma; Catalog #:76119), heptadecanoic acid (CID: 506-12-7; Millipore Sigma; 1297
Catalog #:H3500), octadecanoic acid (CID: 38003-60-0; Millipore Sigma; Catalog #:85679), 1298
nonadecanoic acid (CID: 646-30-0; Millipore Sigma; Catalog #:72332), icosanoic acid (CID: 1299
506-30-9); Millipore Sigma; Catalog #: 39383). 1300
1301
DATA AVAILABILITY 1302
Supplementary Figures S1-S3 and Supplementary Table S1 accompany the paper. All raw 1303
data in the paper and instructions for the behavioral assays are available on Zenodo (DOI: 1304
10.5281/zenodo.5822539) at this link: https://zenodo.org/record/5822539#.YdXST2jMIuU. 1305
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
29
Figure 1 1306
1307
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
30
Figure 1: Mosquitoes show strong preferences among individual humans 1309
(A) Schematic of two-choice olfactometer assay (top). Photographs (bottom) of a mannequin 1310
arm modeling the position of a live human forearm on top of the stimulus box in the two-choice 1311
olfactometer (left) and the opening in the stimulus box lid used to expose an area of human 1312
skin (5.08 cm x 2.54 cm) in the assay (right). 1313
(B-E) Wild-type mosquitoes attracted to live human forearms (B) or a 5.08 cm x 2.54 cm piece 1314
of human-worn nylons and controls (C-E) of the indicated subjects in the two-choice 1315
olfactometer assay. Subject pairs in E are ordered by nonparametric effect size. 1316
(F) Attractiveness scores for human subjects derived from data in E, with subject ID font size 1317
scaled to the attractiveness score: Subject 33 (score=144), Subject 24 (score=34), Subject 31 1318
(score=32), Subject 32 (score=26), Subject 30 (score=23), Subject 25 (score=18), Subject 19 1319
(score=1), Subject 28 (score=0). 1320
(G-I) Longitudinal two-choice olfactometer data for two wild-type Aedes aegypti mosquito 1321
strains, Orlando (circles) and Liverpool (triangles), showing attraction to the indicated subject 1322
pairs. The total time elapsed between the first and last experiment shown is indicated, and 1323
corresponded to July 12, 2018 to July 3, 2019 (G), February 1, 2018 to August 6, 2019 (H), 1324
and July 30, 2018 to March 21, 2019 (I). 1325
In B-E, data are displayed as violin plots with median indicated by horizontal black lines and 1326
the bounds of the violin corresponding to the range (30-40 mosquitoes/trial). B: n=10-16 trials, 1327
C: n=7-8 trials, D: n=12-21 trials, E: n=14-20 trials (except n=81 trials for the Subject 25 vs 25 1328
comparison). Data corresponding to adjacent violin plots labeled with different letters are 1329
significantly different (p<0.05, Wilcoxon rank-sum tests with Bonferroni correction). 1330
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
31
Figure 2 1331
1332
1334
Figure 2: Mosquitoes lacking Orco retain individual human preferences 1335
(A) Schematic of two-choice olfactometer assay indicating the location of mosquitoes that were 1336
not activated, activated but not attracted, or attracted in response to a control trial in which 1337
Subject 25 nylons were placed in both stimulus boxes. Stacked bar plots indicate the mean 1338
total percent of mosquitoes that were in each category in all trials (30-40 mosquitoes/trial, 1339
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
32
n=30-34 trials, *p<0.01, Wilcoxon rank-sum tests with Bonferroni correction comparing each 1340
category across the two genotypes). 1341
(B) Top: schematic of two-choice olfactometer assay, indicating the location (purple shading) 1342
of all mosquitoes attracted to either stimulus in a control trial in which Subject 25 nylons were 1343
placed in both stimulus boxes. Bottom: trials in which 9 or fewer animals entered either trap 1344
were excluded. 1345
(C,D) Left: schematic of Orco and ligand-specific subunit (ORx). Right: percent of mosquitoes 1346
of the indicated genotype attracted to the indicated stimuli in the two-choice olfactometer 1347
assay. Data from trials that met the inclusion criteria are displayed as violin plots with median 1348
indicated by horizontal black lines and the bounds of the violin corresponding to the range (30-1349
40 mosquitoes/trial, n=11-18 trials, except n=29-31 for the Subject 25 vs 25 comparison). Data 1350
corresponding to adjacent violin plots labeled with different letters are significantly different 1351
(p<0.05, Wilcoxon rank-sum tests with Bonferroni correction). 1352
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
33
Figure 3 1353
1354
1356
Figure 3: Mosquitoes lacking Ir8a retain individual human preferences 1357
(A) Schematic of two-choice olfactometer assay indicating the location of mosquitoes that were 1358
not activated, activated but not attracted, or attracted in response to a control trial in which 1359
Subject 25 nylons were placed in both stimulus boxes. Stacked bar plots indicate the mean 1360
total percent of mosquitoes that were in each category in all trials (30-40 mosquitoes/trial, 1361
n=31-33 trials, *p<0.0001, Wilcoxon rank-sum tests with Bonferroni correction comparing each 1362
category across the two genotypes). 1363
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
34
(B) Top: schematic of two-choice olfactometer assay, indicating the location (purple shading) 1364
of all mosquitoes attracted to either stimulus in a control trial in which Subject 25 nylons were 1365
placed in both stimulus boxes. Bottom: trials in which 9 or fewer animals entered either trap 1366
were excluded. 1367
(C, D) Left: Schematic of Ir8a and ligand-specific subunit (IRx). Right: percent of mosquitoes of 1368
the indicated genotype attracted to the indicated stimuli in the two-choice olfactometer assay. 1369
Data from trials that met the inclusion criteria are displayed as violin plots with median 1370
indicated by horizontal black lines and the bounds of the violin corresponding to the range (30-1371
40 mosquitoes/trial, n=11-22 trials, except n=24-29 for the Subject 25 vs 25 comparison). Data 1372
corresponding to adjacent violin plots labeled with different letters are significantly different 1373
(p<0.05, Wilcoxon rank-sum tests with Bonferroni correction). 1374
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
35
Figure 4 1375
1376
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
36
Figure 4: Generation and characterization of Ir25a and Ir76b mutants 1378
(A-B) Schematic of the Aedes aegypti Ir76b (A) and Ir25a (B) genomic loci, detailing sgRNA 1379
sites and modified protein products of the indicated mutant alleles superimposed on Ir76b and 1380
Ir25a protein snake plots, which were generated using Protter v1.0 (84). 1381
(C) Schematic of single stimulus olfactometer assay. 1382
(D-G) Percent of mosquitoes of the indicated genotypes activated to leave the start canister 1383
(D-E) or attracted to the indicated stimuli (F-G) in the single stimulus olfactometer assay. Data 1384
are displayed as violin plots with median indicated by horizontal black lines and the bounds of 1385
the violin corresponding to the range (10-20 mosquitoes/trial, n=6-16 trials). Kruskal-Wallis test 1386
was used to compare each mutant allele to wild-type controls (ns, not significant; *p<0.05). 1387
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
37
Figure 5 1388
1389
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
38
Figure 5: Mosquitoes lacking Ir76b or Ir25a show reduced attraction to humans but 1391
retain individual human preferences 1392
(A) Schematic of two-choice olfactometer assay indicating the location of mosquitoes that were 1393
not activated, activated but not attracted, or attracted in response to a control trial in which 1394
Subject 25 nylons were placed in both stimulus boxes. Stacked bar plots indicate the mean 1395
total percent of mosquitoes that were in each category (30-40 mosquitoes/trial, n=13-16 trials, 1396
*p<0.05, Wilcoxon rank-sum tests with Bonferroni correction comparing each category across 1397
the two genotypes). 1398
(B) Top: schematic of two-choice olfactometer assay, indicating the location (purple shading) 1399
of all mosquitoes attracted to either stimulus in a control trial in which Subject 25 nylons were 1400
placed in both stimulus boxes. Bottom: trials in which 9 or fewer animals entered either trap 1401
were excluded. 1402
(C-E) Left: Schematic of Ir76b and Ir25a and ligand-specific subunit (IRx). Right: percent of 1403
mosquitoes of the indicated genotype attracted to the indicated stimuli in the two-choice 1404
olfactometer assay. Data from trials that met the inclusion criteria are displayed as violin plots 1405
with median indicated by horizontal black lines and the bounds of the violin corresponding to 1406
the range (30-40 mosquitoes/trial, n=8-13 trials, except n=4-13 for the Subject 25 vs 25 1407
comparison). Data corresponding to adjacent violin plots labeled with different letters are 1408
significantly different (p<0.05, Wilcoxon rank-sum tests with Bonferroni correction). 1409
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
39
Figure 6 1410
1411
1413
1414
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
40
Figure 6: Carboxylic acids are enriched on the skin of humans who are highly attractive 1415
to mosquitoes 1416
(A) Overview of experimental procedure for gas chromatography/quantitative time of flight 1417
mass spectrometry (GC/QTOF-MS) experiments. 1418
(B) Timeline of 4 replicate GC/QTOF-MS experiments in initial human subject cohort. 1419
(C) Representative chromatograms from the indicated sample groups, including merged 1420
extracted ion chromatograms from a set of ~200 features enriched on worn nylons versus 1421
unworn nylons and solvent controls in Experiments 1.1-1.4 (Supplementary Table S1). 1422
(D) Volcano plot of features enriched on worn nylons versus unworn nylons and solvent 1423
controls in Experiment 1.1. Nine identified compounds that were differentially abundant 1424
between high and low attractor groups in Experiments 1.1-1.4 are indicated. 1425
(E) Table of differential features in Experiments 1.1-1.4. 1426
(F) Heatmap quantifying abundance of carboxylic acids with 3-20 carbons in the indicated 1427
human subjects, averaged across 4 experiments. 1428
(G) Representative extracted ion chromatograms of several carboxylic acids in the two most 1429
and least attractive subjects from the initial cohort. 1430
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
41
Figure 7 1431
1432
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
42
Figure 7: Carboxylic acids are enriched in a validation cohort of highly mosquito 1434
attractive humans. 1435
(A) Schematic of single stimulus olfactometer assay. 1436
(B) Mosquitoes attracted to nylons from 18 subjects at the extremes of low and high attraction 1437
in the single stimulus olfactometer assay, comprising 14 subjects from the GC/QTOF-MS 1438
validation study and 4 subjects from the initial cohort. Single stimulus olfactometer assay data 1439
from 45 additional subjects, comprising 42 subjects from the GC/QTOF-MS validation study 1440
and 3 subjects from the initial cohort are available on Zenodo (DOI: 10.5281/zenodo.5822539). 1441
Data are displayed as violin plots with median indicated by horizontal black lines and the 1442
bounds of the violin corresponding to the range (14-24 mosquitoes/trial, n=13-28 trials) 1443
(C) Timeline of 4 replicate GC/QTOF-MS experiments (Experiments 2.1-2.4), performed 1444
approximately one year after Experiments 1.1-1.4. 1445
(D) Volcano plot of features enriched on worn nylons versus unworn nylons and solvent 1446
controls in Experiment 2.3. Identified compounds that were differentially abundant between 1447
high and low attractors in all Experiments 2.1-2.4 are indicated with an arrow, and labeled with 1448
an uppercase letter, corresponding to the table in E. 1449
(E) Table describing features that were consistently differentially abundant in high versus low 1450
attractors in Experiments 2.1-2.4. 1451
(F) Heatmap quantifying abundance of carboxylic acids with 10-20 carbons, averaged across 4 1452
experiments, in the 18 subjects in B. 1453
(G) Representative extracted ion chromatograms of three carboxylic acids in the 3 most and 3 1454
least attractive subjects of the 18 subject validation cohort in B. 1455
(H) Quantified abundance (median peak areas) of three carboxylic acids in high attractors 1456
(n=11) versus low attractors (n=7) across Experiments 2.1-2.4. Data are displayed as violin 1457
plots with median indicated by horizontal black lines and the bounds of the violin 1458
corresponding to the range. Each plotted point represents the overall median abundance of the 1459
compound in one subject across Experiments 2.1-2.4. 1460
Data corresponding to adjacent violin plots labeled with different letters are significantly 1461
different (Wilcoxon rank-sum test followed by FDR correction p≤0.1). 1462
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
43
Figure 8 1463
1464
1466
Figure 8: Dilution of highly attractive human odor eliminates mosquito preferences 1467
(A,B) Percent of mosquitoes attracted to the indicated stimuli in the two-choice olfactometer 1468
assay. Mosquitoes were presented with a constant size of nylon worn by a low attractor, either 1469
Subject 28 (A) or Subject 19 (B), and decreasing amounts of nylon worn by high attractor 1470
Subject 33, corresponding to the indicated fraction of the low attractor nylon size. The total 1471
amount of nylon was balanced by adding unworn nylon. Data are displayed as violin plots with 1472
median indicated by horizontal black lines and the bounds of the violin corresponding to the 1473
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
44
range (30-40 mosquitoes/trial, n=11-20 trials). Data corresponding to adjacent violin plots 1474
labeled with different letters are significantly different (p<0.05, Wilcoxon rank-sum tests with 1475
Bonferroni correction). 1476
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
45
Supplementary Figure S1 1477
1478
1479
1481
Supplementary Figure S1 - Related to Figure 4 1482
PCR genotyping Ir76b and Ir25a mutant strains 1483
(A-B) Agarose gel electrophoresis images of PCR fragments before (left) and after (right) 1484
restriction with the indicated restriction enzyme to genotype the indicated Ir76b (A) and Ir25a 1485
(B) mutants. 1486
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
46
Supplementary Figure S2 1487
1488
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
47
Supplementary Figure S2 - Related to Figure 6 and Figure 7 1490
Additional description and validation of GC/QTOF-MS data 1491
(A) Mechanism of pentafluorobenzyl-bromide (PFB-Br) derivatization reaction. 1492
(B) Abundance of several carboxylic acids in low attractors (Subjects 19, 28) vs. high attractors 1493
(Subjects 31,33). Each dot represents abundance of the indicated compound for one subject in 1494
one experiment (median of 4 replicate samples). Nonparametric linear mixed-effects model 1495
followed by Benjamini Hochberg FDR correction (p<0.1) was used. Violins labeled by different 1496
lowercase letters were significantly different. In addition to the 5 compounds plotted here, 3 1497
additional compounds were also found to significantly more abundant in the high attractors 1498
than low attractors: tetradecanoic acid, hexadecenoic acid, and icosanoic acid. Tridecanoic 1499
acid was not significantly different. 1500
(C) Representative extracted ion chromatograms (EICs) for 3 control compounds: 2 deuterated 1501
internal standards and 1 nylon-derived compound, from the indicated human subjects in 1502
Experiment 1.1. 1503
(D) Abundance of control compounds shown in (C). Each dot represents the abundance 1504
(median of 4 replicate samples) of the indicated compound for one subject in 1 of 4 1505
experiments: Experiments 1.1-1.4). 1506
(E) Representative extracted ion chromatograms (EICs) for 3 control compounds: 2 deuterated 1507
internal standards and 1 nylon-derived compound, from the indicated human subjects in 1508
Experiment 2.1. 1509
(F) Abundance of control compounds shown in E. Each dot represents the abundance (median 1510
of 4 replicate samples) of the indicated compound for one subject in 1 of 4 experiments: 1511
Experiments 2.1-2.4). 1512
(G) Extracted ion chromatograms of three identified hit compounds, overlaid with extracted ion 1513
chromatograms of the same sample spiked with a known standard. 1514
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
48
Supplementary Figure S3 1515
1516
1518
Supplementary Figure S3- Related to Figure 6 and Figure 7 1519
Intra-individual stability of skin chemistry 1520
A) Timeline of behavior (green blocks) and GC/QTOF-MS experiments (purple blocks), relative 1521
to one another. 64 human volunteers participated in this study, and nylons from 18 of these 1522
were analyzed by GC/QTOF-MS. Nylons from four of these subjects (Subjects 19, 28, 31, 33) 1523
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
49
were repeatedly tested behaviorally over a 3-year period and also analyzed using GC/QTOF-1524
MS in 2 sets of 4 replicate experiments (Experiments 1.1-1.4, Experiments 2.1-2.4) that were 1525
conducted 1 year apart. 1526
(B) Heatmaps depicting quantified abundance of carboxylic acids with 10-20 carbons, 1527
averaged across 4 replicate samples per experiment, in 4 subjects. Each heatmap represents 1528
one of 8 independent experiments. Experiments 1.1-1.4 were conducted about a year before 1529
Experiments 2.1-2.4. 1530
(C) Heatmaps depicting quantified abundance of carboxylic acids with 10-20 carbons, 1531
averaged across 5 replicate samples per experiment, in 18 subjects from the validation cohort. 1532
Each heatmap represents one of 4 independent experiments, conducted about a week apart. 1533
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
50
*Zenodo (DOI: 10.5281/zenodo.5822539) at this link: 1534
https://zenodo.org/record/5822539#.YdXST2jMIuU 1535
Supplementary Table S1
Untargeted workflow for GC/QTOF-MS # features at each step
Description of analysis step Experiments
1.1-1.4
Experiments 2.1-
2.4
1. Feature finding performed on all worn nylon samples (from 4
experiments) – using Agilent Unknowns Analysis software –
Sure Mass deconvolution
~133,000 ~300,000
2. Initial feature deduplication (R code “step 2”) 1,494 1,925
3. Targeted analysis of deduplicated feature list in all samples
(Agilent Qualitative Analysis 10) 1,494 1,925
4. Using R code “step 4”: replaced zero values with NA, remove
features present in <10% of all samples, impute missing values
with 1/2 the lowest value for that feature, perform Log2
transformation
Not determined Not determined
5. Imported data back into Agilent Mass Profiler Professional
software: baselining: “none”, scaling: “none”, and then use a
fold change filter to select high quality features in each
experiment
a. found in ≥50% of samples in ≥1 subject group
b. coefficient of variation ≤40% in ≥1 subject group
c. 2-fold upregulated (FDR< 0.05) in ≥1 subject group vs.
unworn and vs. solvent control group
Exp. 1.1: 404
Exp: 1.2: 376
Exp: 1.3: 345
Exp. 1.4: 378
(Note: 204 features
were found in 4 of 4
experiments)
Exp. 2.1: 633
Exp: 2.2: 619
Exp: 2.3: 604
Exp. 2.4: 597
(Note:161
features were
found in 4 of 4
experiments)
6. Filter on “hits”: features differentially abundant in high
attractor vs low attractor subject: ≥1.5-fold change (FDR<0.1)
(Agilent Mass Profiler Professional – Volcano plot)
• Note: In Exp. 1.1-1.4: 2 high attractor subjects: 31 & 33
were compared 2 low attractor Subjects 19 & 28.
(Subject 24 was not available for GC/MS analysis)
• Note: In Exp. 2.1-2.4: 7 low attractor subjects were
compared to 11 high attractor subjects
Exp. 1.1: 210
Exp. 1.2: 219
Exp: 1.3: 147
Exp. 1.4: 240
(Figure 6D shows
Experiment 1.1)
Exp. 2.1: 130
Exp. 2.2: 106
Exp: 2.3: 100
Exp. 2.4:105
(Figure 7D shows
Experiment 2.3)
7. Filter on hits found in 4 of 4 experiments (Agilent Mass
Profiler Professional – Venn diagram) 93 23
8. Manually remove redundant features from CSV file in
Microsoft Excel
51 (Figure 6E table,
more info: Zenodo*)
13 (Figure 7E
table, more info:
Zenodo*)
9. Formula prediction (Agilent Qualitative Analysis 10)
42 (more info:
Zenodo*)
5 (more info:
Zenodo*)
10. Positive identification of hit compounds (matched to known
standard) 9 3
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted January 5, 2022. ; https://doi.org/10.1101/2022.01.05.475088doi: bioRxiv preprint
