Origin of microbial biomineralization and magnetotaxis
during the Archean
, Greig A. Paterson
, Qiyun Zhu
, Yinzhao Wang
, Evguenia Kopylova
, Ying Li
, Rob Knight
Dennis A. Bazylinski
, Rixiang Zhu
, Joseph L. Kirschvink
, and Yongxin Pan
Key Laboratory of Earth and Planetary Physics, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China;
Bio-Mineralization and Nano-Structures Laboratory, Chinese Academy of Sciences, Beijing 100029, China;
Genomic Medicine, J. Craig Venter Institute, La
Jolla, CA 92037;
Department of Pediatrics, University of California, San Diego, La Jolla, CA 92037;
College of Biological Sciences, China Agricultural
University, Beijing 100193, China;
Department of Computer Science and Engineering, University of California, San Diego, La Jolla, CA 92037;
School of Life
Sciences, University of Nevada, Las Vegas, NV 89154-4004;
State Key Laboratory of Lithospheric Evolution, Institute of Geology and Geophysics, Chinese
Academy of Sciences, Beijing 100029, China;
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125;
Earth–Life Science Institute, Tokyo Institute of Technology, Meguro, Tokyo 152-8551, Japan
Edited by Donald E. Canfield, Institute of Biology and Nordic Center for Earth Evolution, University of Southern Denmark, Odense M, Denmark, and approved
January 10, 2017 (received for review September 3, 2016)
Microbes that synthesize minerals, a process known as microbial
biomineralization, contributed substantially to the evolution of
current planetary environments through numerous important
geochemical processes. Despite its geological significance, the
origin and evolution of microbial biomineralization remain poorly
understood. Through combined metagenomic and phylogenetic
analyses of deep-branching magnetotactic bacteria from the Nitro-
spirae phylum, and using a Bayesian molecular clock-dating
method, we show here that the gene cluster responsible for bio-
mineralization of magnetosomes, and the arrangement of magne-
tosome chain(s) within cells, both originated before or near the
Archean divergence between the Nitrospirae and Proteobacteria.
This phylogenetic divergence occurred well before the Great Oxy-
genation Event. Magnetotaxis likely evolved due to environmen-
tal pressures conferring an evolutionary advantage to navigation
via the geomagnetic field. Earth’s dynamo must therefore have
been sufficiently strong to sustain microbial magnetotaxis in the
Archean, suggesting that magnetotaxis coevolved with the geo-
dynamo over geological time.
Magnetotactic bacteria (MTB) are a polyphyletic group of
microorganisms that biomineralize intracellular nano-sized
magnetosomes of magnetite (Fe
) and/or greigite (Fe
Magnetosomes are normally organized into chain-like structures to
facilitate the navigation of MTB using Earth’s magnetic field, a
behavior known as magnetotaxis (2). Because of their ubiquitous
distribution, MTB play key roles in global iron, nitrogen, sulfur,
and carbon cycling (3). Magnetosome crystals can be preserved
as magnetofossils after MTB die and lyse, and these crystals can
be used as reliable biomarkers and are major contributors to
sedimentary paleomagnetic records (4–7). Evidence suggests
that magnetofossils as old as ∼1.9 Ga may be preserved (4).
Molecular and genetic studies have identified a group of
clustered genes that control magnetosome biomineralization in
MTB (8–10). However, the origin and evolution of these mag-
netosome gene clusters (MGCs) remain controversial, and several
scenarios have been posited to explain the patchy distribution of
magnetosome formation over a broad phylogenetic range. Such
scenarios include multiple evolutionary origins (11), extensive hor-
izontal gene transfers (12), and vertical gene transfer (13, 14). MTB
in the Nitrospirae phylum represent one of the deep-branching
MTB groups (15, 16). Therefore, analysis of MGCs from Nitrospirae
MTB could yield insights into the origin and evolution of magne-
tosome biomineralization and magnetotaxis. However, no Nitro-
spirae MTB has been successfully cultured, and only three draft
genomes from this phylum are currently available (17, 18). Recent
advances in high-throughput sequencing combined with improving
computational methods are enabling the successful recovery of
high-quality population genomes directly from metagenomic data
(19, 20). Here, we assess whether the phylogenies of key magne-
tosome genes from MTB of the Nitrospirae and Proteobacteria phyla
are consistent with their taxonomic phylogeny using a metagenomic
approach to acquire the population genome and MGC-containing
contigs from environmental Nitrospirae MTB.
We sampled MTB from two freshwater locations in China: the
city moat of Xi’an in Shaanxi province (HCH) and Lake Miyun
near Beijing (MY). MTB belonging to the Nitrospirae phylum were
identified in both samples (Fig. S1). Recovered metagenomic DNA
∼13 Mb of 3,042 contigs ≥1 kb for HCH and ∼32 Mb of 7,893
A wide range of organisms sense Earth’s magnetic field for
navigation. For some organisms, like magnetotactic bacteria,
magnetic particles form inside cells and act like a compass.
However, the origin of magnetotactic behavior remains a mys-
tery. We report that magnetotaxis evolved in bacteria during the
Archean, before or near the divergence between the Nitrospirae
and Proteobacteria phyla, suggesting that magnetotactic bacteria
are one of the earliest magnetic-sensing and biomineralizing or-
ganisms on Earth. The early origin for magnetotaxis would have
provided evolutionary advantages in coping with environmental
challenges faced by microorganisms on early Earth. The persis-
tence of magnetotaxis in separate lineages implies the temporal
continuity of geomagnetic field, and this biological evidence
provides a constraint on the evolution of the geodynamo.
Author contributions: W.L., J.L.K., and Y.P. designed research; W.L. and Y.W. performed
research; Q.Z. contributed new reagents/analytic tools; W.L. and Y.W. collected samples;
W.L., G.A.P., and Q.Z. analyzed data; andW.L., G.A.P., Q.Z., E.K.,Y.L., R.K., D.A.B., R.Z.,J.L.K.,
and Y.P. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
Data deposition: The draft genome of Candidatus Magnetominusculus xianensis strain
HCH-1 reported in this paper has been deposited in the DNA Data Bank of Japan (DDBJ)/
European Molecular Biology Laboratory (EMBL)/GenBank database (accessio n no.
LNQR00000000; the version described in this paper is LNQR01000000). The magnetosome
gene cluster-containing contigs reported in this paper have been deposited in the DDBJ/
EMBL/GenBank database (accession nos. KU221504–KU221507).
To whom correspondence may be addressed. Email: firstname.lastname@example.org, kirschvink@
caltech.edu, or email@example.com.
W.L. and G.A.P. contributed equally to this work.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
www.pnas.org/cgi/doi/10.1073/pnas.1614654114 PNAS Early Edition
AND PLANETARY SCIENCES
contigs ≥1kbforMY(Table S1). In HCH, only a single population
belonging to the phylum Nitrospirae was found and comprised
∼19% of the sampled MTB community (Fig. S1). The pop-
ulation genome of this Nitrospirae was recovered using a com-
bination of similarity- and composition-based approaches (Table
1). Only a single copy of an rRNA operon was identified, which
had a 16S rRNA gene sequence with <92% similarity to the
genera Candidatus Magnetobacterium (17, 18) and Candidatus
Magnetoovum (18). This MTB population was named as “Can-
didatus Magnetominusculus xianensis”strain HCH-1 (HCH-1
hereafter). One contig (18,138 bp) of HCH-1 contained a nearly
complete MGC including homologs to known magnetosome
genes (Fig. 1Aand Table S2). Despite their phylogenetic dis-
tance, the gene content and gene order of the MGC from HCH-1
and available Nitrospirae MGCs were highly conserved (>76%,
To explore the evolutionary history of the MGCs between the
Nitrospirae and Proteobacteria phyla, we performed a phyloge-
netic analysis of five core magnetosome proteins (MamABEKP)
from HCH-1 and published MTB strains. These proteins were
selected because they were the only bidirectional best hits between
the complete MGC of the Nitrospirae MTB strain “Candidatus
Magnetobacterium casensis”(Mcas) and that of representative
Proteobacteria MTB. These five proteins from the Nitrospirae phy-
lum form a monophyletic group (Fig. S2). Branches with >75%
bootstrap values have similar topology and agree with those in the
tree based on concatenated magnetosome proteins (Fig. 1B). This is
consistent with the pattern in the phylogenomic tree that is gener-
ated based on a concatenated alignment of up to 400 highly
conserved proteins (Fig. 1B), suggesting that these magnetosome
genes coevolved via vertical transmission with the genome. Fur-
thermore, comparison of magnetosome genes to three housekeep-
ing genes (recA,gyrB, and pyrH) revealed that the codon use bias
was not significantly different (P>0.05 by ttest) between
magnetosome genes and the vertically transmitted housekeeping
genes. Similarly, the number of synonymous substitutions per
synonymous site was also not significantly different (P>0.05 by t
test; Fig. 1C). The results of the codon use test are consistent
with that of the phylogenetic-based analysis, which rules out
uncertainty in the reliability of the codon analysis alone (21), and
strongly suggest that the magnetosome genes were inherited
through vertical transfer (22). Together, these results indicate
that magnetosome biomineralization and magnetotaxis is an
ancient metabolic process and was present before the separation
of the Nitrospirae and Proteobacteria phyla, or transferred unde-
tectably early between the base of Nitrospirae and Proteobacteria
soon after divergence.
Considering the deep-branching lineage of Nitrospirae MTB
and their tightly packed magnetosome genes (Fig. 1A), the
content and order of magnetosome genes in Nitrospirae are likely
conserved and represent an ancestral MGC. To test this, we
compared the implicit phylogenetic pattern of the core magne-
tosome genes to all other genes in HCH-1 and available Nitro-
spirae MTB genomes. The magnetosome genes are clustered
toward the majority of other genes (Fig. 2A), suggesting that the
evolutionary history of magnetosome genes in Nitrospirae is not
distinct from the genomic background. Analysis of metagenomic
data from MY yielded four contigs containing nearly complete
Nitrospirae MGCs (Fig. 2B). Of these, contig MY-22 (44,819 bp)
was >99.99% identical to the MGC of previously identified Mcas
isolated from the same lake (17), indicating the high-quality as-
sembly of metagenomic contigs in this study. All four MGCs
have a high level of conservation with HCH-1, Mcas, “Candi-
datus Magnetobacterium bavaricum”(Mbav), and “Candidatus
Magnetoovum chiemensis”strain CS-04 (Mchi) (Fig. 2B). Phy-
logenetic trees of concatenated magnetosome proteins (Fig. 2B)
and each of MamABEKP (Fig. S3) from the Nitrospirae MGCs
form a monophyletic group with consistent topology, which
further indicates the coevolution of magnetosome genes and
their antiquity and suggests the origin of magnetotaxis is earlier
than previously thought (13). Considering that all known Nitro-
spirae MTB biomineralize bullet-shaped magnetite magneto-
somes, a crystal morphology that has not yet been observed in
magnetite as a result of abiotic processes (3), our results indicate
that bullet-shaped magnetite may be the first type of magneto-
some (13) and may represent reliable microbial fossils.
There are very few geological constraints on the timing of the
separation of the major clades within the Bacterial domain.
Earlier evidence, based on organic biomarkers and the following
phylogenetic analysis, suggested an Archean origin of the Pro-
teobacteria (23) and the divergence of the Nitrospirae and Pro-
teobacteria before ∼2.7 Ga (24). However, the biomarkers upon
which those inferences were based are now known to be con-
taminants (25). A global phylogenomic reconstruction of the
evolution of gene families suggests that the Proteobacteria phy-
lum diverged during the Archean Eon (26). We calculate the age
of divergence for the Nitrospirae and Proteobacteria by phyloge-
nomic and Bayesian relaxed molecular clock analyses (Figs. S4
and S5). Two different relaxed clock models (log normal auto-
correlated clock and uncorrelated gamma multipliers clock) with
two different combinations of time calibrations were imple-
mented (SI Materials and Methods). Results are consistent across
clock modes and time calibrations, and all analyses suggest that
the Nitrospirae and Proteobacteria phyla diverged before 3.0 Ga
(∼3.38–3.21 Ga) (Fig. S4). Hence, MTB likely evolved in the
mid-Archean, and, as previously suggested (4), may be one of the
earliest biomineralizing and magnetotactic organisms on Earth.
is abundant in the atmosphere (∼21%) and oceans
on the present Earth. The early Earth, however, was character-
ized by a reducing system, and the ocean chemistry in the
Archean was different from today [e.g., with a scarcity of molecular
and abundant dissolved Fe(II) (∼40–120 μmol/L) (27, 28)].
Multiple lines of evidence have suggested that, through volcanic
processes, Archean oceans likely had sufficient nutrients (e.g.,
S, NO, NO−
4, etc.) to sustain mi-
croorganisms with anaerobic or microaerobic metabolisms (29, 30).
All known present-day MTB are microaerophilic and anaerobic,
and, according to their metabolic and genomic analyses, it seems
that the nutrients available in Archean oceans could support the
survival and growth of MTB. For example, many extant MTB are
chemolithoautotrophic and have the ability to fix CO
either via the
Calvin–Benson–Bassham cycle, the reverse tricarboxylic acid cycle,
or the reductive acetyl-CoA (Wood–Ljungdahl) pathway (1, 17, 18).
Table 1. General genomic features of the genome of
Candidatus Magnetominusculus xianensis strain HCH-1
xianensis strain HCH-1
Total genome size, Mb 3.593273
GC content, % 45.40
No. of contigs 152
N50, kb 45.767
Maximum contig length, kb 150.216
No. of coding sequences (CDS) 3,415
No. of RNAs 47
No. of copies of 5S rRNA 1
No. of copies of 16S rRNA 1
No. of copies of 23S rRNA 1
Estimated completeness, % 98.18
Estimated contamination, % 0.91
www.pnas.org/cgi/doi/10.1073/pnas.1614654114 Lin et al.
They are also capable of reducing NO−
2, and NO through
the denitrification pathway and capable of oxidizing H
the sulfur oxidation pathway (e.g., refs. 17, 18, and 31–33). The
temperature of Archean ocean water is still debated, with esti-
mates ranging from values comparable to modern tropical waters
(26–35 °C) (34) to as warm as 55–85 °C (35). Thermophilic MTB,
living at temperatures of 32–63 °C, have been discovered in
present-day hot springs (36), supporting the possibility of these
bacteria surviving in a hot Archean ocean.
Extant MTB respond to vertical redox gradients in the water
or sediment columns, particularly that of O
exploiting these gradients to maintain their optimal positions within
their microenvironments (1, 37). Archean oceans are traditionally
thought to have been uniformly anoxic and therefore devoid of the
redox gradients similar to today. Under such conditions, abundant
ferrous iron would have been prevalent throughout the entire Ar-
chean water column, except in the euphotic zone near the surface
where iron-oxidizing photosynthetic bacteria would presumably
rapidly remove it from solution (38). The resultant ferric precipi-
tates resulted in the production of the banded iron formations,
yielding an “upside-down”biosphere (39), in which the sediments
were rich in ferric iron but overlain by water rich in dissolved ferrous
iron. This might have produced enough of a vertical redox
gradient to provide MTB an environmental niche. Recently,
however, there is growing evidence that early oceans might
not be uniformly anoxic but were redox stratified, likely in
4 5 6 O 23 2425
28 1 K 210
4 5 6 O 24 25
45 6 O24 25
28 1 K 10
PM28 1 K29
mam gene mad gene
Phylogenomic treeMamABEKP concatenated tree
Number of substitutions per site dS
Effective number of codons Nc
Fig. 1. (A) Maximum-likelihood phylogenetic tree of 16S rRNA gene sequences showing the relationship between Candidatus Magnetominusculus xianensis
strain HCH-1 (HCH-1) and related magnetotactic bacteria. Bootstrap values at nodes are percentages of 100 replicates. On the right-hand side, comparison of
magnetosome gene clusters between HCH-1 and other reported Nitrospirae MTB. (B) Bootstrap consensus trees based on concatenated protein alignment of
five magnetosome proteins (MamABEKP) and phylogenomic tree based on concatenated ubiquitous amino acid sequences. Only bootstrap values of more
than 75% are shown, and branches of the trees are proportionally transformed. MSR-1, Magnetospirillum gryphiswaldense MSR-1; AMB-1, Magnetospirillum
magneticum AMB-1; QH-2, Magnetospira sp. QH-2; MC-1, Magnetococcus marinus MC-1; BW-1, “Candidatus Desulfamplus magnetomortis BW-1”; MMP,
“Candidatus Magnetoglobus multicellularis”; RS-1, Desulfovibrio magneticus RS-1; ML-1, alkaliphilic magnetotactic strain ML-1; MV-1, Magnetovibrio bla-
kemorei MV-1; IT-1, “Candidatus Magnetofaba australis strain IT-1”; SS-5, Gammaproteobacteria magnetotactic strain SS-5. (C) Comparison of sequence
divergence of five magnetosome genes and three housekeeping genes (recA,gyrB, and pyrH). Nc, the codon use bias; dS, the number of synonymous
substitutions per synonymous site.
Lin et al. PNAS Early Edition
AND PLANETARY SCIENCES
shallow water, and that “oxygen oases”may have existed as far
back as 2.8–3.2 Ga (40–42). These redox-stratified waters
could also provide potential habitable environments for the
origin and evolution of ancient MTB.
Typically, bacteria use a 3D “run-and-tumble”search strategy
for finding their preferred microenvironments. For MTB, mag-
netotaxis reduces this 3D search to an optimized 1D search along
geomagnetic field lines in chemically stratified water or sediment
mam gene mad gene
signal transduction, response regulator,
man gene iron metabolism
FeoB 28 29 1 K
4 5 6 O 23 24 25
28 29 1 K
4 5 6 O 23 2425
28 1 K 210
4 5 6 O 23 24 25
28 1 K 210
4 5 6 O 2425
28 1 K 210
4 5 6 O 2425
2328 1 K 10
PM28 1 K29
Candidatus Magnetominusculus xianensis Candidatus Magnetobacterium bavaricum
Candidatus Magnetobacterium casensis Candidatus Magnetoovum chiemensis
Distal weightDistal weight
Close weight Close weight
Close weight Close weight
Fig. 2. (A) Implicit phylogenetic pattern of magnetosome core genes (mamABEIKMOPQ) of four Nitrospirae MTB in their genomic background. Each point
represents a protein-coding gene in the genome, with magnetosome core genes highlighted in red. The xaxis represents the sum of the relative bit scores of
all hits within the phylum Nitrospirae, and the yaxis represents that outside the phylum Nitrospirae. Typically, genes that were vertically transmitted have a
moderate-to-high xvalue, representing its homologs in closely related taxonomic groups. Genes with a horizontal origin outside this phylum have a zero-to-
low xvalue along with a moderate-to-high yvalue, representing homologs in distant taxonomic groups instead of close ones. Genes without traceable
evolutionary history are located close to the origin, suggesting a lack of homologs in the database. (B) Maximum-likelihood tree based on a concatenation of
MamABEP and schematic comparison of the Nitrospirae MGCs recovered from metagenomic data of MY with those of HCH-1, Mcas, Mbav, and Mchi.
www.pnas.org/cgi/doi/10.1073/pnas.1614654114 Lin et al.
columns (2). Previous observations that magnetofossil concen-
trations in marine sediments plummet during weak-field intervals
surrounding geomagnetic reversals implies that magnetotaxis con-
fers a selective advantage in field strengths of ∼6μT or greater (43).
Recent analysis of reliable Archean paleointensity data suggests
that field strengths of 20–50 μT are observed (44), indicating that
the Archean geodynamo was sufficient to support magnetotaxis. An
Archean origin of magnetotaxis, and its persistence in multiple
lineages since their divergence during Archean time, implies both
temporal continuity of Earth’s dynamo and persistent environ-
The present study suggests an ancient origin of MTB in the
Archean Eon that is much earlier than previously reported. MTB
therefore represent one of the earliest magnetic-sensing and
biomineralizing organisms on Earth. The Archean origin of
MTB indicates that magnetotaxis is an evolutionarily ancient
characteristic and that evolved in response to gradients in the
Materials and Methods
Site Description and MTB Sample Preparation. Surface sediments (depths of
5–10 cm) were sampled from the following two freshwater locations in China:
the city moat of Xi’an in Shaanxi province (HCH) (34.25287, 108.92187), and
Lake Miyun near Beijing (MY) (40.48874, 117.00714). The collected sedi-
ments were transferred to 600-mL plastic flasks, transported to the labora-
tory (in Beijing), and incubated at room temperature without disturbance.
MTB cells from HCH were magnetically enriched using a double-ended open
magnetic separation apparatus known as the “MTB trap”(45). The collected
cells were washed twice and resuspended in sterile distilled water. MTB
samples from MY used in this work were collected by Lin et al. (46) and had
been stored at −80 °C. Genomic DNA was extracted from the enriched MTB
by using the TIANamp Bacteria DNA Kit (Tiangen) following the manufac-
16S rRNA Gene Sequences Generation and Analysis. 16S rRNA gene sequences
were amplified using the bacteria/archaeal universal primers 515F/806R that
target the V4 region (47). The purified PCR products were cloned into the
pMD19-T vector (TaKaRa) and chemically DH5a competent cells (Tiangen),
following the manufacturers’instructions. Clones were randomly selected
and were sequenced using an ABI 3730 genetic analyzer (Beijing Genomics
Institute, Beijing, China). Sequences were processed, aligned, and clustered
into operational taxonomic units (OTUs) using the QIIME pipeline (48).
Shotgun Metagenomic Sequencing and Data Analysis. The metagenomic DNA
was amplified using multiple displacement amplification. Shotgun sequencing of
metagenomic DNA was performed using Illumina HiSeq 2000 using the pair-end
125 ×125 library with a 600-bp inset size (Beijing Genomics Institute, Beijing,
China). Metagenomic reads were assembled into contigs using Velvet, version
1.2.10, assembler (49). Resulting contigs were filtered by a minimal length cutoff
of 1 kb. For details, see SI Materials and Methods.
Population Genome Binning of a Magnetotactic Nitrospirae from HCH. Because
only a single population of Nitrospirae MTB with high relative abundance
was identified in the metagenome of HCH, its population genome was re-
covered from the assemblies using a combination of similarity- [MEGAN,
version 5 (50)] and composition-based [CLARK, version 1.1.2 (51)] ap-
proaches. The quality and accuracy of the acquired population genome were
assessed using CheckM (52), using the lineage-specific workflow. For details,
see SI Materials and Methods.
Implicit Phylogenomic Analysis of Nitrospirae MTB Genes. The global implicit
phylogenetic pattern of the magnetotactic Nitrospirae genomes of HCH-1,
Mcas, Mbav, and Mchi was inferred using HGTector 0.2.0 (53), a sequence
similarity-based HGT prediction pipeline. For details, see SI Materials
Identification of Nitrospirae MGC-Containing Contigs. To detect Nitrospirae
MGC-containing contigs, magnetosome protein sequences from Mcas (17),
Mchi (18), and Mbav (18) were used as queries in tBLASTn analyses against
the assembled contigs of each sample. All matches (Evalue ≤1e-5) were
then manually verified.
Phylogenetic Analysis of Magnetosome Proteins. Magnetosome gene ortho-
logs between the complete MGC of Mcas and MGCs of representative MTB
populations including HCH-1, MSR-1, AMB-1, QH-2, MV-1, MC-1, IT-1, BW-1,
MMP, and RS-1 were calculated by bidirectional best-hit analysis through the
SEED viewer (54). The amino acid sequences of magnetosome proteins were
aligned by MUSCLE algorithms (55) using MEGA, version 6.06 (56), and
poorly conserved regions were trimmed by using the Gblocks method (57).
Appropriate protein models of substitution were selected using the Find
Best DNA/Protein Models implemented in MEGA, version 6.06 (56), and
maximum-likelihood phylogenetic trees were constructed using MEGA,
version 6.06, with a GTR and Gamma model (56) for 16S rRNA genes and
RAxML, version 8.0.19, with a LG and Gamma model (58) for magnetosome
proteins. Phylogenomic tree construction was performed using PhyloPhlAn
(59), which uses USEARCH (60) and MUSCLE (55) to extract up to 400 con-
served ubiquitous proteins (Table S3) coded in genomes and perform indi-
vidual protein alignments. The universally conserved and phylogenetically
discriminative positions in each protein alignment were then concatenated
into a single long sequence through PhyloPhlAn. PhyML, version 3.0 (initial
tree: BioNJ; tree topology search: NNIs) (61), was used to generate a maxi-
mum-likelihood tree. Confidence in phylogenetic results was assessed using
the 100 bootstrap resampling approaches.
Sequence Divergence. The average number of synonymous substitutions per
synonymous site (dS) of the five magnetosome genes and three house-
keeping genes (recA,gyrB, and pyrH) were calculated using MEGA, version
6.06 (56). The effective number of codons (N
) was calculated with CodonW
tool (version 1.4.4 at codonw.sourceforge.net/) (62). N
varies between 21 for
maximum codon bias and 61 for minimum codon bias. MTB involved in this
analysis were AMB-1, MSR-1, MC-1, QH-2, RS-1, MMP, Mbav, Mcas, Mchi,
Divergence Time Estimation. Sequence data from 64 genomes were used to
estimate the divergence time between phyla Nitrospirae and Proteobacteria.
A subset of up to 400 core proteins was extracted and aligned using Phy-
loPhlAn (59). Maximum likelihood tree and bootstrap values were per-
formed using RAxML, version 8.0.19, with a VT and Gamma model (58).
Divergence times were estimated using PhyloBayes, version 4.1c (63). For
details, see SI Materials and Methods.
ACKNOWLEDGMENTS. We thank Longfei Wu for valuable comments and
suggestions. W.L. and Y.P. acknowledge financial support from National
Natural Science Foundation of China (NSFC) Grants 41330104, 41621004, and
41374074. W.L. acknowledges support from the Youth Innovation Pro-
motion Association of the Chinese Academy of Sciences. G.A.P. acknowl-
edges funding from NSFC Grants 41374072 and 41574063. D.A.B. is
supported by US National Science Foundation Grant EAR-1423939. J.L.K. is
supported by US National Aeronautics and Space Administration Exobiology
1. Bazylinski D, Lefèvre C, Schüler D (2013) Magnetotactic bacteria. The Prokaryotes, eds
Rosenberg E, DeLong E, Lory S, Stackebrandt E, Thompson F (Springer, Berlin), pp
2. Frankel RB, Bazylinski DA, Johnson MS, Taylor BL (1997) Magneto-aerotaxis in marine
coccoid bacteria. Biophys J 73(2):994–1000.
3. Lin W, Bazylinski DA, Xiao T, Wu L-F, Pan Y (2014) Life with compass: Diversity and
biogeography of magnetotactic bacteria. Environ Microbiol 16(9):2646–2658.
4. Chang SBR, Kirschvink JL (1989) Magnetofossils, the magnetization of sediments, and
the evolution of magnetite biomineralization. Annu Rev Earth Planet Sci 17:169–195.
5. Roberts AP, et al. (2013) Magnetic properties of pelagic marine carbonates. Earth Sci
6. Kopp RE, Kirschvink JL (2008) The identification and biogeochemical interpretation of
fossil magnetotactic bacteria. Earth Sci Rev 86(1-4):42–61.
7. Pan Y, et al. (2005) The detection of bacterial magnetite in recent sediments of Lake
Chiemsee (southern Germany). Earth Planet Sci Lett 232(1-2):109–123.
8. Grünberg K, Wawer C, Tebo BM, Schüler D (2001) A large gene cluster encoding
several magnetosome proteins is conserved in different species of magnetotactic
bacteria. Appl Environ Microbiol 67(10):4573–4582.
9. Murat D, Quinlan A, Vali H, Komeili A (2010) Comprehensive genetic dissection of the
magnetosome gene island reveals the step-wise assembly of a prokaryotic organelle.
Proc Natl Acad Sci USA 107(12):5593–5598.
10. Lohsse A, et al. (2011) Functional analysis of the magnetosome island in Magneto-
spirillum gryphiswaldense: The mamAB operon is sufficient for magnetite bio-
mineralization. PLoS One 6(10):e25561.
11. Delong EF, Frankel RB, Bazylinski DA (1993) Multiple evolutionary origins of mag-
netotaxis in bacteria. Science 259(5096):803–806.
Lin et al. PNAS Early Edition
AND PLANETARY SCIENCES
12. Jogler C, Schüler D (2009) Genomics, genetics, and cell biology of magnetosome
formation. Annu Rev Microbiol 63(1):501–521.
13. Lefèvre CT, et al. (2013) Monophyletic origin of magnetotaxis and the first magne-
tosomes. Environ Microbiol 15(8):2267–2274.
14. Zeytuni N, et al. (2015) MamA as a model protein for structure-based insight into the
evolutionary origins of magnetotactic bacteria. PLoS One 10(6):e0130394.
15. Jogler C, et al. (2011) Conservation of proteobacterial magnetosome genes and
structures in an uncultivated member of the deep-branching Nitrospira phylum. Proc
Natl Acad Sci USA 108(3):1134–1139.
16. Spring S, et al. (1993) Do minating role of an unusual magnetotactic bacterium in the
microaerobic zone of a fr eshwater sediment. Appl Environ Microbiol 59(8):2397–2403.
17. Lin W, et al. (2014) Genomic insights into the uncultured genus “Candidatus Mag-
netobacterium”in the phylum Nitrospirae.ISME J 8(12):2463–2477.
18. Kolinko S, Richter M, Glöckner FO, Brachmann A, Schüler D (2016) Single-cell geno-
mics of uncultivated deep-branching magnetotactic bacteria reveals a conserved set
of magnetosome genes. Environ Microbiol 18(1):21–37.
19. Evans PN, et al. (2015) Methane metabolism in the archaeal phylum Bathyarchaeota
revealed by genome-centric metagenomics. Science 350(6259):434–438.
20. Brown CT, et al. (2015) Unusual biology across a group comprising more than 15% of
domain Bacteria. Nature 523(7559):208–211.
21. Friedman R, Ely B (2012) Codon usage methods for horizontal gene transfer detection
generate an abundance of false positive and false negative results. Curr Microbiol
22. Wallau GL, Ortiz MF, Loreto EL (2012) Horizontal transposon transfer in Eukarya:
Detection, bias, and perspectives. Genome Biol Evol 4(8):689–699.
23. Sheridan PP, Freeman KH, Brenchley JE (2003) Estimated minimal divergence times of
the major bacterial and archaeal phyla. Geomicrobiol J 20(1):1–14.
24. Brocks JJ, Summons RE (2003) Sedimentary hydrocarbons, biomarkers for early life.
Treatise on Geochemistry, eds Holland HD, Turekian KK (Pergamon, Oxford), pp
25. French KL, et al. (2015) Reappraisal of hydrocarbon biomarkers in Archean rocks. Proc
Natl Acad Sci USA 112(19):5915–5920.
26. David LA, Alm EJ (2011) Rapid evolutionary innovation during an Archaean genetic
expansion. Nature 469(7328):93–96.
27. Canfield DE (2005) The early history of atmospheric oxygen: Homage to Robert M.
Garrels. Annu Rev Earth Planet Sci 33(1):1–36.
28. Habicht KS, Gade M, Thamdrup B, Berg P, Canfield DE (2002) Calibration of sulfate
levels in the archean ocean. Science 298(5602):2372–2374.
29. Canfield DE, Rosing MT, Bjerrum C (2006) Early anaerobic metabolisms. Philos Trans R
Soc Lond B Biol Sci 361(1474):1819–1834, discussion 1835–1836.
30. Buick R (2012) Geobiology of the Archean Eon. Fundamentals of Geobiology, eds
Knoll AH, Canfield DE, Konhauser KO (Wiley, Chichester, UK), pp 351–370.
31. Williams TJ, Zhang CL, Scott JH, Bazylinski DA (2006) Evidence for autotrophy via the
reverse tricarboxylic acid cycle in the marine magnetotactic coccus strain MC-1. Appl
Environ Microbiol 72(2):1322–1329.
32. Ji B, et al. (2014) Comparative genomic analysis provides insights into the evolution
and niche adaptation of marine Magnetospira sp. QH-2 strain. Environ Microbiol
33. Schübbe S, et al. (2009) Complete genome sequence of the chemolithoautotrophic
marine magnetotactic coccus strain MC-1. Appl Environ Microbiol 75(14):4835–4852.
34. Blake RE, Chang SJ, Lepland A (2010) Phosphate oxygen isotopic evidence for a
temperate and biologically active Archaean ocean. Nature 464(7291):1029–1032.
35. Knauth LP (2005) Temperature and salinity history of the Precambrian ocean: Impli-
cations for the course of microbial evolution. Palaeogeogr Palaeoclimatol Palaeoecol
36. Bazylinski DA, Lefèvre CT (2013) Magnetotactic bacteria from extreme environments.
Life (Basel) 3(2):295–307.
37. Spring S, Bazylinski DA (2006) Magnetotactic bacteria. The Prokaryotes: An Evolving
Electronic Resource for the Microbiological Community, ed Dworkin M (Springer,
New York), pp 842–862.
38. Kappler A, Pasquero C, Konhauser KO, Newman DK (2005) Deposition of banded iron
formations by anoxygenic phototrophic Fe(II)-oxidizing bacteria. Geology 33(11):
39. Walker JCG (1987) Was the Archaean biosphere upside down? Nature 329(6141):
40. Planavsky NJ, et al. (2014) Evidence for oxygenic photosynthesis half a billion years
before the Great Oxidation Event. Nat Geosci 7(4):283–286.
41. Riding R, Fralick P, Liang L (2014) Identification of an Archean marine oxygen oasis.
Precambrian Res 251(9):232–237.
42. Satkoski AM, Beukes NJ, Li W, Beard BL, Johnson CM (2015) A redox-stratified ocean
3.2 billion years ago. Earth Planet Sci Lett 430:43–53.
43. Kirschvink JL (1982) Paleomagnetic evidence for fossil biogenic magnetite in western
Crete. Earth Planet Sci Lett 59(2):388–392.
44. Biggin AJ, et al. (2015) Palaeomagnetic field intensity variations suggest Meso-
proterozoic inner-core nucleation. Nature 526(7572):245–248.
45. Jogler C, et al. (2009) Toward cloning of the magnetotactic metagenome: Identifi-
cation of magnetosome island gene clusters in uncultivated magnetotactic bacteria
from different aquatic sediments. Appl Environ Microbiol 75(12):3972–3979.
46. Lin W, Jogler C, Schüler D, Pan Y (2011) Metagenomic analysis reveals unexpected
subgenomic diversity of magnetotactic bacteria within the phylum Nitrospirae.Appl
Environ Microbiol 77(1):323–326.
47. Caporaso JG, et al. (2012) Ultra-high-throughput microbial community analysis on the
Illumina HiSeq and MiSeq platforms. ISME J 6(8):1621–1624.
48. Caporaso JG, et al. (2010) QIIME allows analysis of high-throughput community se-
quencing data. Nat Methods 7(5):335–336.
49. Zerbino DR, Birney E (2008) Velvet: Algorithms for de novo short read assembly using
de Bruijn graphs. Genome Res 18(5):821–829.
50. Huson DH, Auch AF, Qi J, Schuster SC (2007) MEGAN analysis of metagenomic data.
Genome Res 17(3):377–386.
51. Ounit R, Wanamaker S, Close TJ, Lonardi S (2015) CLARK: Fast and accurate classifi-
cation of metagenomic and genomic sequences using discriminative k-mers. BMC
52. Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW (2015) CheckM: As-
sessing the quality of microbial genomes recovered from isolates, single cells, and
metagenomes. Genome Res 25(7):1043–1055.
53. Zhu Q, Kosoy M, Dittmar K (2014) HGTector: An automated method facilitating ge-
nome-wide discovery of putative horizontal gene transfers. BMC Genomics 15(1):717.
54. Overbeek R, et al. (2014) The SEED and the rapid annotation of microbial genomes
using subsystems technology (RAST). Nucleic Acids Res 42(Database issue, D1):
55. Edgar RC (2004) MUSCLE: Multiple sequence alignment with high accuracy and high
throughput. Nucleic Acids Res 32(5):1792–1797.
56. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S (2013) MEGA6: Molecular evo-
lutionary genetics analysis version 6.0. Mol Biol Evol 30(12):2725–2729.
57. Castresana J (2000) Selection of conserved blocks from multiple alignments for their
use in phylogenetic analysis. Mol Biol Evol 17(4):540–552.
58. Stamatakis A (2014) RAxML version 8: A tool for phylogenetic analysis and post-
analysis of large phylogenies. Bioinformatics 30(9):1312–1313.
59. Segata N, Börnigen D, Morgan XC, Huttenhower C (2013) PhyloPhlAn is a new
method for improved phylogenetic and taxonomic placement of microbes. Nat
60. Edgar RC (2010) Search and clustering orders of magnitude faster than BLAST.
61. Guindon S, et al. (2010) New algorithms and methods to estimate maximum-likeli-
hood phylogenies: Assessing the performance of PhyML 3.0. Syst Biol 59(3):307–321.
62. Peden JF (1999) Analysis of codon usage. PhD dissertation (University of Nottingham,
63. Lartillot N, Lepage T, Blanquart S (2009) PhyloBayes 3: A Bayesian software package
for phylogenetic reconstruction and molecular dating. Bioinformatics 25(17):
64. Surget-Groba Y, Montoya-Burgos JI (2010) Optimization of de novo transcriptome
assembly from next-generation sequencing data. Genome Res 20(10):1432–1440.
65. Li W, Godzik A (2006) Cd-hit: A fast program for clustering and comparing large sets
of protein or nucleotide sequences. Bioinformatics 22(13):1658–1659.
66. Buchfink B, Xie C, Huson DH (2015) Fast and sensitive protein alignment using DI-
AMOND. Nat Methods 12(1):59–60.
67. Bekker A, et al. (2004) Dating the rise of atmospheric oxygen. Nature 427(6970):
68. Crowe SA, et al. (2013) Atmospheric oxygenation three billion years ago. Nature
69. Hofmann HJ (1976) Precambrian microflora, Belcher Islands, Canada; significance and
systematics. J Paleontol 50(6):1040–1073.
70. Shih PM, Hemp J, Ward LM, Matzke NJ, Fischer WW (2017) Crown group Oxy-
photobacteria postdate the rise of oxygen. Geobiology 15(1):19–29.
71. Blank CE, Sánchez-Baracaldo P (2010) Timing of morphological and ecological inno-
vations in the cyanobacteria—a key to understanding the rise in atmospheric oxygen.
www.pnas.org/cgi/doi/10.1073/pnas.1614654114 Lin et al.
Lin et al. 10.1073/pnas.1614654114
SI Materials and Methods
Shotgun Metagenomic Sequencing and Data Analysis. To obtain
sufficient DNA for shotgun metagenomic sequencing, multiple
displacement amplification was performed using the GenomiPhi
V2 DNA Amplification Kit (GE Healthcare) following the
manufacturer’s instructions. Briefly, 1 μL of DNA was used as
the template and was mixed with 9 μL of sample buffer. The
mixed DNA was heated at 95 °C for 3 min and cooled to 4 °C,
before incubation at 30 °C for 90 min with 1 μL of enzyme
mixture and 9 μL of reaction buffer. To terminate the reaction,
the sample was heated at 65 °C for 10 min. For each sample, nine
amplifications were pooled to reduce potential bias. These were
purified using TIANquik Maxi Purification Kit (Tiangen).
Shotgun sequencing of metagenomic DNA was performed
using Illumina HiSeq 2000 using the pair-end 125 ×125 library
with a 600-bp inset size (Beijing Genomics Institute, Beijing,
China). The entire dataset of two samples is ∼5.55 Gb. Illumina
reads were trimmed to remove the adapter sequences and low-
quality bases, after which 86–88% of paired reads were retained
for each sample. Trimmed, paired-end reads were assembled
using a multiple k-mer–based assemblies (64). Briefly, metagenomic
reads of each sample were individually assembled into contigs using
the Velvet, version 1.2.10, assembler (49) with a range of k-mers
(41, 51, 61, 71, 81, and 91). The different assembles were sub-
sequently merged, and the duplicated and suboptimal contigs
were removed through CD-HIT-EST (65) using a sequence
identity threshold of 0.95 and a word length of 8 to get the final
assembly for each sample. Resulting contigs were filtered by a
minimal length cutoff of 1 kb.
Population Genome Binning of a Magnetotactic Nitrospirae from HCH.
Contigs of sample HCH were sorted using BLASTn alignment
against the NCBI genomes database (version May 2015) together
with previously sequenced MTB draft genomes of Mcas (17),
Mchi (18), and Mbav (18). BLASTn alignment hits with Evalues
larger than 1 ×10
were filtered, and the taxonomical level of
each contig was determined by the lowest common ancestor al-
gorithm implemented in MEGAN, version 5 (50). All contigs
binned to known Nitrospirae MTB species of Mcas, Mbav, and
Mchi were selected. Due to the incomplete nature of available
magnetotactic Nitrospirae draft genomes, the remaining contigs
were further classified using CLARK, version 1.1.2 (51), based
on reduced sets of k-mers by comparison with available genomes
or draft genomes of MTB strains. The measure of conservation
of gene content and gene order of MGCs between HCH-1 and
three available Nitrospirae MTB (Mcas, Mbav, and Mchi) is the
ratio between the number of genes located in conserved content
and order and the total number of bidirectional best-hits genes
between MGCs of HCH-1 and three Nitrospirae MTB.
Implicit Phylogenomic Analysis of Nitrospirae MTB Genes. The global
implicit phylogenetic pattern of the magnetotactic Nitrospirae
genomes of HCH-1, Mcas, Mbav, and Mchi was inferred using
HGTector 0.2.0 (53). Protein sequence similarity search was
performed using DIAMOND 0.9.7 (66) against a database (gen-
erated by HGTector) that contains one representative per species
from all available nonredundant RefSeq prokaryotic proteomes
(October 2015), plus the MTB proteomes reconstructed in this
study. Quality cutoffs for valid hits were Evalue ≤1e-20, percent-
age identity ≥30%, and query coverage ≥50%. For each protein-
coding gene, the top 250 highest-scoring hits from different species
were retained. For each hit, a “relative bit score”was calculated
as the original bit score of the hit divided by the bit score of the
query sequence aligned against itself. The overall distribution
pattern of all genes in a genome was visualized by plotting the
sum of the bit scores of hits within phylum Nitrospirae against
that outside this phylum per gene.
Divergence Time Estimation. Molecular-dating analyses were per-
formed using PhyloBayes, version 4.1c (63). The CAT-GTR model
was implemented for amino acid replacement, and analyses were run
under either the log-normal autocorrelated relaxed clock (-ln) or the
uncorrelated gamma multipliers (-ugam). For each condition, two
replicate chains with 20,000 generations were run. Dates were
assessed by running the readdiv with the first 20% of generations
removed as burn-in for each analysis. Two different combinations of
age constraints were used for the divergence time estimation. For the
first combination of age constraints, the minimum age of the root of
Oxyphotobacteria (oxygenic Cyanobacteria) was set at 2.32 Ga (the
rise in atmospheric oxygen) (67), and the maximum age was set at
3.0 Ga (40, 68). For the second combination, a minimum age of
1.9 Ga (the first widely accepted fossil oxygenic Cyanobacteria) (69)
and a maximum age of 2.32 Ga (postdating the rise of oxygen
according to ref. 70) were implemented as the oxyphotobacterial root.
In addition, for the second combination another time constraint, the
divergence time between Oxyphotobacteria and Melainabacteria, was
included, which was set from 2.5 Ga (70) to 3.8 Ga (the end of late
heavy bombardment). For all analyses, the age calibration for the
last common ancestor of all taxa used in this study was set between
2.32 and 3.8 Ga (71).
Lin et al. www.pnas.org/cgi/content/short/1614654114 1of5
‘Candidatus Magnetobacterium casensis’ (JMFO00000000)
‘Candidatus Magnetobacterium bavaricum’ (X71838)
‘Candidatus Magnetoovum chiemensis’ CS-04 (JX402654)
‘Candidatus Thermomagnetovibrio paiutensis’ HSMV-1 (GU289667)
Acinetobacter indicus strain A648 (NR 117784)
Acinetobacter seohaensis strain SW100 (NR 115299)
Magnetospira sp. QH-2 (EU675666)
Magnetospira thiophila MMS-1 (EU861390)
Magnetospirillum gryphiswaldense MSR-1 (Y10109)
Magnetospirillum magneticum AMB-1 (D17514)
Magnetospirillum magnetotacticum MS-1 (M58171)
Magnetovibrio blakemorei MV-1 (L06455)
Magneto-ovoid bacterium MO-1 (EF643520)
Magnetococcus marinus MC-1 (L06456)
‘Candidatus Magnetococcus yuandaducum’ (FJ667777)
Fig. S1. Phylogenetic tree of operational taxonomic units (OTUs at 97% threshold similarity) for 16S rRNA gene clone libraries of MTB communities from the
city moat of Xi’an in Shaanxi province (HCH) and Lake Miyun near Beijing (MY). The evolutionary history was inferred by using the maximum-likelihood
method based on the Kimura two-parameter model with 100 bootstraps. On the right-hand side, a heatmap shows the relative abundance and distribution of
each OTU from this study.
Lin et al. www.pnas.org/cgi/content/short/1614654114 2of5
Fig. S2. Bootstrap consensus trees of five magnetosome proteins (MamABEKP) based on the maximum-likelihood method. Only full-length protein sequences were included
in this analysis. Bootstrap values are expressed as percentages, and only values of more than 75% are shown. MSR-1, Magnetospirillum gryphiswaldense MSR-1; AMB-1,
Magnetospirillum magneticum AMB-1; SO-1, Magnetospirillum sp. SO-1; QH-2, Magnetospira sp. QH-2; MC-1, Magnetococcus marinus MC-1; BW-1, Candidatus Desulfamplus
magnetomortis BW-1; MMP, Ca. Magnetoglobus multicellularis; RS-1, Desulfovibrio magneticus RS-1; ML-1, alkaliphilic magnetotactic strain ML-1; MV-1, Magnetovibrio
blakemorei MV-1; IT-1, Ca. Magnetofaba australis strain IT-1; SS-5, Gammaproteobacteria magnetotactic strain SS-5; HK-1, Ca. Magnetomorum sp. HK-1; B13, Latescibacteria
bacterium SCGC AAA252-B13.
Lin et al. www.pnas.org/cgi/content/short/1614654114 3of5
Fig. S3. Bootstrap consensus trees of magnetosome proteins MamABEKP from the four additional Nitrospirae MGCs of MY and those full-length proteins
from all available Nitrospirae MTB based on the maximum-likelihood method. Bootstrap values are expressed as percentages, and only values of more than
75% are shown.
Mean divegence time 95% mean confidence intervals
Archean Proterozoic Phanerozoic
Calibartion constraints and
molecular clock models:
Great Oxidation Event (GOE)
Fig. S4. Summary of mean divergence dates for the Nitrospirae and Proteobacteria phyla estimated using Bayesian relaxed molecular-clock analyses with two
different time constraints and two different molecular clock models (see SI Materials and Methods for details). Two replicated chains were run for each
condition. The input phylogenomic tree used here is shown in Fig. S5.
Lin et al. www.pnas.org/cgi/content/short/1614654114 4of5
Fig. S5. Phylogenomic maximum-likelihood tree of 64 bacterial genomes. Bootstrap values are expressed as percentages, and only values of >75% are shown.
Magnetotactic bacteria are displayed in blue. The Aquificae strains were used as outgroup.
Other Supporting Information Files
Table S1 (DOCX)
Table S2 (DOCX)
Table S3 (DOCX)
Lin et al. www.pnas.org/cgi/content/short/1614654114 5of5