of 6
Localizing transcripts to single cells suggests an
important role of uncultured deltaproteobacteria
in the termite gut hydrogen economy
Adam Z. Rosenthal
a,1
, Xinning Zhang
a,1
, Kaitlyn S. Lucey
a
, Elizabeth A. Ottesen
a
, Vikas Trivedi
b
, Harry M. T. Choi
b
,
Niles A. Pierce
b,c
, and Jared R. Leadbetter
a,2
a
Ronald and Maxine Linde Center for Global Environmental Science,
b
Department of Bioengineering, and
c
Department of Computing and Mathematical
Sciences, California Institute of Technology, Pasadena, CA 91125
Edited by James M. Tiedje, Michigan State University, East Lansing, MI, and approved August 13, 2013 (received for review April 29, 2013)
Identifying microbes responsible for particular environmental
functions is challenging, given that most environments contain
an uncultivated microbial diversity. Here we combined approaches
to identify bacteria expressing genes relevant to catabolite
fl
ow
and to locate these genes within their environment, in this case
the gut of a
lower,
wood-feeding termite. First, environmental
transcriptomics revealed that 2 of the 23 formate dehydrogenase
(FDH) genes known in the system accounted for slightly more than
one-half of environmental transcripts. FDH is an essential enzyme
of H
2
metabolism that is ultimately important for the assimilation
of lignocellulose-derived energy by the insect. Second, single-cell
PCR analysis revealed that two different bacterial types expressed
these two transcripts. The most commonly transcribed FDH in situ
is encoded by a previously unappreciated deltaproteobacterium,
whereas the other FDH is spirochetal. Third, PCR analysis of frac-
tionated gut contents demonstrated that these bacteria reside in
different spatial niches; the spirochete is free-swimming, whereas
the deltaproteobacterium associates with particulates. Fourth, the
deltaproteobacteria expressing FDH were localized to protozoa via
hybridization chain reaction-FISH, an approach for multiplexed,
spatial mapping of mRNA and rRNA targets. These results under-
score the importance of making direct vs. inference-based gene
species associations, and have implications in higher termites, the
most successful termite lineage, in which protozoa have been lost
from the gut community. Contrary to expectations, in higher ter-
mites, FDH genes related to those from the protozoan symbiont
dominate, whereas most others were absent, suggesting that a
successful gene variant can persist and
fl
ourish after a gut pertur-
bation alters a major environmental niche.
acetogenesis
|
RNA-Seq
|
micro
fl
uidic digital PCR
A
large body of work in microbial ecology is centered on
documenting gene and organism diversity into large cata-
logs. The development of tools that unambiguously identify en-
vironmentally active organisms within these immense catalogs
has lagged, however. Phylogenetic inventories of ribosomal RNA
and physiology-relevant genes are often constructed to examine
the evolutionary relationship and full diversity of bacteria and
their many functional roles in nature. Linking species marker
genes for given microbes in the environment with genes for im-
portant ecological functions has proven more challenging. Gene
inventory-based studies often attempt to assign functional genes
to their encoding organisms by inference, for example, based on
parallels in the branching order of the respective phylograms
prepared from different inventories, noting the positions of
genes belonging to reference organisms on those same trees.
Phylogenetic inference is limited, however, because genetic
transfers can occur across phylogenetic lines of descent, and is
even less predictive if the availability of reference organisms for
internal calibration of phylograms is limited.
In the present study, gene inventories and community-wide
transcriptional pro
fi
les were combined with single-cell analyses
and in situ assays to address such matters directly. We interro-
gated a tiny, yet complex environment that accommodates robust,
stable, and species-rich microbial communities
the hindgut of a
wood-feeding lower termite,
Zootermopsis nevadensis
(1).
Termites and their gut microbiota digest lignocellulose, the
most abundant natural composite material on Earth. For some
time now, it has been known that a key activity in this nutritional
mutualism involves the bacterial conversion of H
2
+
CO
2
, gen-
erated during wood polysaccharide fermentation, into acetate in
a process called CO
2
-reductive acetogenesis (2, 3). The acetate
originating from CO
2
-reductive acetogenesis fuels up to one-
fi
fth
to one-third of the termites
carbon and energy needs (4, 5). The
H
2
concentration and turnover measured in termite hindguts is
among the highest recorded in any biological system, and H
2
gas
is the central free intermediate during lignocellulose degradation
(6). Thus, identifying the microbes involved in H
2
+
CO
2
me-
tabolism during acetogenesis or other H
2
-using reactions, and
localizing their activity in situ, has great relevance in under-
standing both host nutrition and the global carbon cycle.
All H
2
+
CO
2
acetogenic bacteria isolated from termites to
date have been members of the bacterial phyla
Spirochetes
or
Firmicutes
(4, 7, 8). Gene-based community analyses have led to
a guiding hypothesis that some, most, or all of the contributors to
gut acetogenesis may be spirochetes, many of which have not yet
been cultivated. In contrast, there has been no support for an
important homoacetogenic role of endospore-forming
Firmicutes
species isolated at low abundances from termites or their close
relatives (6, 8
12). The hypothesis that spirochetes may account
for most or all of termite gut acetogenesis remains largely
Signi
fi
cance
Most environments host a poorly understood microbial di-
versity. In recent years, work on gene inventories and meta-
genomics has revealed much about the microbial species and
metabolic genes that may be present in situ; however, con-
necting microbial species with environmental function has
lagged. Here a combination of emerging single-cell and other
approaches revealed the gut microbes that may catalyze a key
activity in their termite hosts. The results implicate a previously
unappreciated deltaproteobacterium living on a gut protist.
Author contributions: A.Z.R., X.Z., E.A.O., H.M.T.C., N.A.P., and J.R.L. designed research;
A.Z.R., X.Z., K.S.L., E.A.O., V.T., and H.M.T.C. performed research; A.Z.R., X.Z., K.S.L., E.A.O.,
V.T., H.M.T.C., N.A.P., and J.R.L. analyzed data; and A.Z.R., X.Z., N.A.P., and J.R.L. wrote
the paper.
The authors declare no con
fl
ict of interest.
This article is a PNAS Direct Submission.
Data deposition: 16S rRNA gene sequences, FDH sequences, and FTHFS sequences have
been deposited in the GenBank database (accession nos.
JX974519
,
GU563433
GU563485
,
GQ922349
GQ922449
, and
JX974463
JX974518
).
1
A.Z.R. and X.Z. contributed equally to this work.
2
To whom correspondence should be addressed. E-mail: jleadbetter@caltech.edu.
This article contains supporting information online at
www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1307876110/-/DCSupplemental
.
www.pnas.org/cgi/doi/10.1073/pnas.1307876110
PNAS Early Edition
|
1of6
MICROBIOLOGY
untested, however, because much of the available data are based
on accumulated indirect phylogenetic inferences, relying heavily
on the limited availability of cultivated homoacetogenic species
for use as requisite internal references.
Here we have focused on the analysis of genes and transcripts
for hydrogenase-linked formate dehydrogenase (FDH) and, to
a lesser extent, formyl-tetrahydrofolate synthetase (FTHFS) (6,
9
11, 13, 14), which together catalyze the
fi
rst and second steps
of the folate branch of the CO
2
-reductive acetogenesis pathway:
H
2
+
CO
2
=>
formate, and formate
+
THF
=>
formyl-tetra-
hydrofolate (15). We also note that FDH genes may play roles in
other processes that consume or generate H
2
+
CO
2
, for ex-
ample, as a follow-up to the pyruvate-formate lyase reaction in
enterics (16), during purine fermentation (17), and during in-
terspecies formate transfer (18). Of these reactions, FDH is
known to be important to symbiosis only in CO
2
-reductive ace-
togenesis; however, in principle, all of the reactions are relevant,
inasmuch as both H
2
and formate serve as substrates for CO
2
-
reductive acetogenesis (15). To move beyond phylogenetic pre-
diction and to physically identify the microbial species re-
sponsible for H
2
+
CO
2
metabolism in termite guts, we sought to
circumvent the limitation of indirect inference using a combina-
tion of gene-based approaches (
SI Appendix
, Fig. S1
).
Results
Community Transcripts of Two Genes Essential to H
2
+
CO
2
Metabolism
Are Dominated by Only a Few Phylotypes.
Leveraging the predictive
power of existing hydrogenase-linked FDH (
fdhF
) and FTHFS
gene inventories (9, 11, 19), gut environmental metatranscriptomic
data were used to identify highly ex
pressed alleles for each of these
functional gene types. High-throughput cDNA sequencing (RNA-
Seq) is increasingly used to quantif
y transcriptional intensity. In
environmental transcriptomics, previous knowledge of endog-
enous gene sequences may serve as a
scaffold
onto which
sequence fragments are mapped. Total gut RNA was extracted
from the gut tracts of fresh
fi
eld-collected and laboratory-
maintained specimens of
Z. nevadensis
workers, converted to
cDNA, and sequenced (
Materials and Methods
and
SI Appendix
,
Table S1
), generating datasets of 14,043,698 (
fi
eld specimens)
and 13,913,270 (laboratory specimens) 37-mer reads. To increase
read density, the sample from freshly collected specimens was
sequenced more extensively, providing an additional 217,687,073
50-mer reads. All three libraries were combined in silico to
compile a large dataset (
245 million reads). Taking into account
ribosomal RNA (
90
95% total RNA) and total RNA from
protozoa, which dominate as much as 90% of the volume of the
gut lumen (1), the number of reads corresponding to bacterial
mRNA transcripts was
2,500,000.
To examine the entire dataset for evidence of gut transcription
of FDH alleles, comparisons were made to a scaffold library
containing 44 known
fdh
F alleles representing 23 phylotypes
previously inventoried in
Z. nevadensis
gut. Only reads perfectly
matching a scaffold were counted as positive hits, and these were
considered unique only when they could be mapped solely to
alleles corresponding to a single FDH-deduced phylotype. A total
of 920 unique reads mapped onto the
Z. nevadensis fdhF
scaffold
dataset (Table 1). More than 40% of these reads (
n
=
381) map-
ped to a single phylotype, ZnD2sec (selenocysteine containing
FDH; GenBank accession no. GU563467), aligning to 207 distinct
locations along the reference gene sequence. The remaining 539
reads mapped to 18 different deduced FDH phylotypes: in order
of abundance (Table 1), ZnHcys (cysteine containing FDH;
GenBank accession no. GQ922420), Zn2cys (GenBank accession
no. GQ922431), alleles identical to the known
fdh
F
Sec
and
fdh
F
Cys
genes of cultivated strain
Treponema primitia
ZAS-2 (GenBank
accession nos. ADJ19611 and ADJ19610) (11), and 14 other
phylogenetically
Treponeme
-like
FDH genes.
The foregoing results verify that the expression of previously
inventoried functional genes can be reliably detected in total en-
vironmental RNA pools using either speci
fi
c or degenerate PCR
primers targeting particular gene types. The RNA-Seq mapping
approach proved to be highly precise for the system at hand;
comparison of all reads with an additional 180 FDH genotypes
identi
fi
ed from guts of insects other than
Z. nevadensis,
as well as
all those available in public databases (
SI Appendix
,TableS2
),
yielded only 17 additional hits (1.8%) (
SI Appendix
, Table S3
). The
most highly expressed phylotypes generally were equally repre-
sented across both
fi
eld and laboratory specimens (
SI Appendix
,
Table S3
). Interestingly, ZnHcys, the second most highly expressed
variant in samples originating from freshly collected
fi
eld speci-
mens, was not detected in the RNA-Seq library from laboratory-
maintained specimens. Further analysis of this was beyond the
scope of this study.
As an internal check of consistency, using different methods, the
relative abundances of the top alleles were con
fi
rmed via cDNA
libraries and quantitative RT-PCR (qPCR) (
Materials and Methods
and
SI Appendix
, Table S3
). RNA-Seq reads were also mapped to
libraries of FTHFS gene variants (
SI Appendix
,TableS4
). FTHFS
variant P37-7A (33.6%) and an FTHFS sequence identical to that
encoded by the cultivated homoacetogen
T. primitia
ZAS-2 (24.5%)
were the most highly expressed phylotypes, together accounting for
58% of all reads. The observation that the fourth and
fi
fth most
highly expressed FDHs and the second most highly expressed
FTHFS are identical to those from termite isolate
T. primitia
ZAS-
2, which is known to encode two FDHs (13), directly demonstrates
the relevance of this species to CO
2
-reductive acetogenesis activity
in the gut environment. This species has served as a representative
symbiont in studies over the past decade (8, 10).
Micro
fl
uidic Digital PCR Identi
fi
es Termite Gut Spirochetes Encoding
Highly Expressed FDH Phylotypes.
Subsequent efforts focused on
learning more about the species identity and nature of the bac-
teria encoding the two most highly expressed hydrogenase-linked
FDH phylotypes, ZnD2sec and Zn2cys, in laboratory-maintained
insects. We turned to micro
fl
uidic multiplex digital-PCR single-
particle analysis, an approach that has been used to identify
the rRNA phylotypes of termite gut bacteria associated with
speci
fi
c metabolic or phage genes (14, 20). This platform works
by sequestering individual particles from a diluted microbial
community into 6.25-nL chambers, wherein multiplex qPCR is
performed using oligonucleotides targeting both functional genes
of interest and ribosomal small subunit (SSU rRNA) genes. The
PCR products generated in chambers containing positive signal(s)
can be retrieved manually and subsequently analyzed. Here
degenerate PCR primer and probe sets targeting either FDH
or FTHFS were used in concert with those targeting the SSU
Table 1.
Z. nevadensis
gut community FDH gene transcription
Phylotype
FDH type
RNA-Seq unique hits
% of total hits
ZnD2sec
fdh
F
sec
381
41.41
ZnHcys
fdhF
cys
129
14.02
Zn2cys
fdh
F
cys
84
9.13
T. primitia
fdh
F
sec
67
7.28
T. primitia
fdh
F
cys
60
6.52
Zn61sec
fdh
F
sec
41
4.46
ZnF7sec
fdh
F
sec
28
3.04
ZnB5sec
fdh
F
sec
27
2.93
ZnB9cys
fdh
F
cys
24
2.61
ZnLsec
fdh
F
sec
20
2.17
ZnB8sec
fdh
F
sec
13
1.41
ZnD3cys
fdh
F
cys
11
1.20
ZnC6sec
fdh
F
sec
8
0.87
Zn36sec RT
fdh
F
sec
7
0.76
Zn51sec
fdh
F
sec
6
0.65
ZnPcys
fdh
F
cys
5
0.54
Zn75cysRT
fdh
F
cys
4
0.43
Zn72secRT
fdh
F
sec
4
0.43
ZnH8cys
fdh
F
cys
1
0.11
2of6
|
www.pnas.org/cgi/doi/10.1073/pnas.1307876110
Rosenthal et al.
rRNA genes from all bacteria. From this, robust associations
between several FDH genes and bacterial species were
drawn. The Zn2cys FDH variant, the third most abundant
transcript in fresh
fi
eld-collected insects and the second most
abundant in laboratory-maintained specimens, was strongly as-
sociated with a spirochetal ribotype,
Treponema
genomovar
ZnR11 (GenBank accession no. JX571935) (
SI Appendix
, Fig S2
).
Using these same methods, we found that this spirochete also
colocalized with the gene for ZNF7sec, the seventh most highly
expressed FDH in
fi
eld-collected specimens (Table 1 and
SI
Appendix
,Fig.S1
), suggesting that, as is known for cultivated
strains of the bacterium
T. primitia
(11, 13), this organism
encodes both seleno and cysteinyl variants of FDH.
Treponema
genomovar ZnR11 also associated with a speci
fi
cFTHFSal-
lele (
SI Appendix
,Fig.S2
), the fourth most highly expressed
FTHFS in situ, accounting for
8% of hits to the FTHFS
scaffold library (
SI Appendix
, Table S4
), a
fi
nding corroborated
by a previously published FTHFS expression pro
fi
le (6). The
micro
fl
uidic platform successfully linked four different genes
to a single genomovar, a feat not previously accomplished using
digital PCR. Several other FDH and FTHFS phylotypes were
also successfully linked to speci
fi
c SSU rRNA genes, all of which
were spirochetal (
SI Appendix
, Figs. S2 and S3
).
The ZnD2sec FDH Gene Is Overrepresented Within the Gut Particulate
Fraction, Which Contains Wood Particles and Protozoa.
Curiously, all
initial attempts to forge links between either the most highly
expressed FTHFS phylotype (P37-7A) or the most highly
expressed FDH phylotype (ZnD2sec) and speci
fi
c SSU rRNA
genes or their complementary functional genes were unsuccess-
ful. We investigated why ZnD2sec, the most highly expressed
FDH in situ, was not captured in these initial micro
fl
uidic trials.
One possibility was that the source organism encoding ZnD2sec,
although highly active, might be insuf
fi
ciently abundant in gut
contents to enable detection in micro
fl
uidic assays. To examine
this possibility, we prepared libraries of selenocysteine FDH
(
fdh
F
Sec
) genes using standard PCR, low-degeneracy primers
optimized to minimize primer bias (19), and total gut DNA
preparations. The ZnD2sec FDH gene accounted for 64% of the
fdh
F
Sec
clones in the resulting libraries from total gut DNA,
implying that it is perhaps the most abundant
fdh
F
Sec
-encoding
member of the community. However, termite gut bacteria are
known to strongly attach to the gut wall (21, 22) and to associate
with the surfaces or insides of gut protozoa (23), in addition to
being free-swimming. Through the micro
fl
uidic device loading
method, larger wood particles, protozoal particles, gut epithe-
lial tissues, and other aggregates were routinely excluded from
our initial micro
fl
uidic protocols, to prevent channel clogging in
the device. This might explain the lack of ampli
fi
cation of the
ZnD2sec gene from density-clari
fi
ed samples.
To examine this possibility, qPCR was performed on both
particle-associated fractions and density-clari
fi
ed fractions of gut
luminal contents. Primers speci
fi
c for either the ZnD2sec gene
or the
fdh
F
sec
of the spirochete
T. primitia
ZAS-2 were used. The
T. primitia
gene was 6.5-fold more abundant in the density-
clari
fi
ed luminal
fl
uid compared with the particle-associated
fraction, whereas the ZnD2sec was 86-fold more abundant in the
particle-associated fraction. These results are consistent with the
idea that the spirochete
T. primitia
ZAS-2 is free-swimming,
whereas ZnD2sec is associated with either wood particles or
protozoa. These
fi
ndings largely explain the initial failure to
amplify ZnD2sec using micro
fl
uidics (
SI Appendix
, Table S5
).
ZnD2sec, the Most Highly Transcribed Hydrogenase-Linked FDH Gene
in both Laboratory and Field-Collected Specimens of
Z. nevadensis
,Is
Encoded by a Protist-Associated Deltaproteobacterium.
To identify
cells carrying the ZnD2sec gene, we performed micro
fl
uidic
digital PCR on homogenized particulate fractions to detect
positive ampli
fi
cation of ZnD2sec. In multiplex reactions, the
ZnD2sec gene was coampli
fi
ed with the same SSU rRNA gene
(GenBank accession no. JX974519) in
fi
ve of seven wells that
returned sequences for both genes (
SI Appendix
, Table S8
). Se-
quence analysis of wells with coampli
fi
cation revealed that the
ZnD2sec FDH gene is encoded by a novel deltaproteobacterium
(Fig. 1 and
SI Appendix
, Table S3
).
Deltaproteobacteria encompass many diverse subgroups and
physiotypes, including anaerobic dissimilatory sulfate-, sulfur-,
iron-, and manganese-reducing bacteria, as well as strictly oxy-
gen-respiring predatory myxobacteria and bdellovibrios. At least
one species of deltaproteobacteria is known to be capable of
bona-
fi
de CO
2
-reductive, acetogenic growth (24). The not-yet-
cultivated phylotype encoding ZnD2sec is distinct from any
characterized deltoproteobacteria in the literature, but clusters
with marker sequences from several not-yet-cultivated organisms
resident in diverse termite gut tracts (Fig. 1). The closest culti-
vated and described species belong to the order Desulfarculales,
speci
fi
cally
Desulfarculus baarsii,
a poorly studied bacterium ca-
pable of sulfate reduction as well as CO
2
fi
xation into acetate via
the acetyl-CoA pathway (25). Although
D. baarsii
was originally
named
Desulfovibrio baarsii
, neither it nor the sequences related
to the organism encoding ZnD2sec are closely related to mem-
bers of the genus
Desulfovibrio
, several species of which inhabit
termite guts, including on the surface of certain protozoa, but
none of which are homoacetogenic (26).
Our
fi
ndings imply a major role for deltaproteobacteria in
the H
2
+
CO
2
economy of termite hindgut communities. They
also contradict initial phylogenetic inference, which in the ab-
sence of the proper reference organism might broadly af
fi
liate
the ZnD2sec gene with a spirochete (11). Here, without any
interference from indirect phylogenetic hypothesis, direct single-
cell analyses revealed an organism that would not have been on
any previous list of candidates. T
his result raises the possibility
that these deltaproteobacteria may play a major role in H
2
+
CO
2
acetogenesis, which if true would substantially increase
our understanding of bacterial acetogenesis in termites (8
11).
Fig. 1.
Micro
fl
uidic digital PCR links a key FDH phylotype to a deltaproteo-
bacterium. Phylograms of (
A
) deduced FDH protein and (
B
) 16s rRNA gene
sequences. Sequences from evolutionarily lower termites and roaches are
shown in green, and sequences recovered from higher termites are in blue.
Higher termite FDH sequences group with the lower termite protozoan-
associated fdh variant ZnD2sec (green box). Tree construction parameters and
sequence accession numbers are listed in
SI Appendix
,TableS2
.Eachbar
indicates 0.1 change per alignment position.
Rosenthal et al.
PNAS Early Edition
|
3of6
MICROBIOLOGY
However, it remains possible that the species encoding ZnD2sec
uses FDH in another anaerobic mode, such as pyruvate oxida-
tion, purine fermentation (17), or interspecies formate transfer
(18), all of which nevertheless would have relevance to mutual-
istic energy
fl
ow in the symbiosis. Hereinafter, we provisionally
refer to the deltaproteobacterium expressing these ZnD2sec and
SSU rRNA genes as genomovar ZnDP-F1.
HCR-FISH Localizes Genomovar ZnDP-F1
Expressing ZnD2sec mRNA
and ZnDP-F1 rRNA to the Surface of a Protist.
To independently
>support the idea that genomovar ZnDP-F1 deltaproteobac-
teria encodes and coexpresses these mRNA and rRNA sequen-
ces, and to determine the exact association of these bacteria with
speci
fi
c particulates such as termite gut protozoa, we turned to
a new multiplexed in situ hybridization technology based on the
mechanism of hybridization chain reaction (HCR) (27). In this
approach, mRNA and rRNA targets are detected using RNA
probes that trigger hybridization cascades in which
fl
uorophore-
labeled RNA hairpins self-assemble into tethered
fl
uorescent
ampli
fi
cation polymers. Straightforward multiplexing follows
from the simultaneous use of multiple HCR ampli
fi
ers pro-
grammed to operate independently.
We
fi
rst performed three-channel colocalization studies in
environmental bacterial samples isolated from homogenized
termite gut samples (Fig. 2 and
SI Appendix
,Fig.S4
). The
ZnD2sec mRNA, the genomovar ZnDP-F1 rRNA, and the rRNA
of all bacteria were targeted using single probes carrying dif-
ferent HCR initiators (Fig. 2
A
C
). The composite image of
Fig. 2
D
reveals strong signal colocalization for the ZnD2sec
mRNA and the genomovar ZnDP-F1 d
eltaproteobacterial rRNA,
with bacterial clusters stained either for both targets or for neither
target. Staining for the rRNA of all bacteria con
fi
rmed that bac-
teria unstained for ZnD2sec mRNA and genomovar ZnDP-F1
rRNA were nonetheless amenable to HCR staining (Fig. 2
C
).
To gain further insight into these colocalization relationships,
we applied a background threshold to the
fl
uorescence in each
channel to identify interesting pixels (Fig. 2
G
I
), selected salient
bacterial clusters with contrasting morphologies and costaining
properties (orange rectangle, costaining for all three targets; blue
rectangle, elongated spirochete-like morphology with pre-
dominant staining only for the rRNA of all bacteria), and created
scatterplots of these pixel intensities for each pair of channels
(Fig. 2
J
L
). The results are striking. As shown in Fig. 2
J
, signals
for the ZnD2sec mRNA and genomovar ZnDP-F1 deltapro-
teobacteria rRNA were minimal for the blue cluster and scaled
linearly for the orange cluster. At high signal intensity, a sub-
population of pixels deviated from the linear relationship,
exhibiting a disproportionate mRNA signal. Fig. 2
K
compares
signals for the ZnD2sec mRNA and the rRNA of all bacteria,
revealing a vertical blue distribution that reaf
fi
rms the absence of
ZnD2sec staining for this elongated bacterial cluster. The pre-
dominant linear relationship, with a deviating subpopulation at
high signal, is again evident for the orange cluster. Fig. 2
L
compares the signals for genomovar ZnDP-F1 rRNA and the
rRNA of all bacteria, revealing linear relationships for both the
orange and blue distributions. The blue scatter deviates slightly
from vertical, suggesting that the ZnDP-F1 deltaproteobacteria
rRNA probe may be binding with low ef
fi
ciency to off-target
rRNAs in the blue bacterial cluster. For the orange cluster, the
clean linear relationship between the rRNA signals contrasts
with the excursion from linearity of the rRNA vs. mRNA dis-
tributions of Fig. 2
J
and
K
, suggesting that these excursions are
likely biological rather than artifacts of the assay. Overall, the
predominance of single linear relationships in the orange and
blue distributions of Fig. 2
J
L
suggests that the probes selec-
tively bind single target sequences (
SI Appendix
, Figs. S8 and S9
).
The close correlation between ZnD2sec mRNA staining and
genomovar ZnDP-F1 rRNA staining is consistent with expect-
ations based on the foregoing micro
fl
uidics results.
To search for evidence of a spatial association between
genomovar ZnDP-F1 deltaproteobacteria and termite gut protists,
we performed three-channel in situ HCR on particulate and pro-
tozoan-containing samples removed intact from
Zootermopsis
guts
(Fig. 3 and
SI Appendix
,Fig.S5
). The signal for the ZnD2sec
mRNA colocalized with the signal for the ZnDP-F1 rRNA, and
these signals were localized to protozoa resembling the
Z. neva-
densis
protozoans
Trichonympha campanula
and
Trichonympha
sphaerica
(based on morphology; Fig. 3
A
C
). Our
fi
nding that not
all rRNA-stained cells (or clusters of cells) stained with the mRNA
probe is not surprising, considering the known differences in both
the quantity and stability between bacterial mRNAs and rRNAs.
Similar patterns, in which not all cells stained by rRNA probes stain
with an mRNA probe, have been found in similar FISH experi-
ments performed on single cells of laboratory-grown cultures (28)
(
SI Appendix
,Fig.S9
). The signal for the rRNA of all bacteria also
colocalized with the signal from the ZnDP-F1 rRNA (Fig. 3
D
and
E
). The high signal correlation using speci
fi
c and general SSU
rRNA probes suggests that this protist-associated bacterial niche
may be dominated by a single deltaproteobacterial species.
The HCR-FISH data presented in Figs. 2 and 3 provide in-
dependent support for the micro
fl
uidic digital PCR data, with
both indicating high expression of ZnD2sec by a previously un-
appreciated deltaproteobacterium, genomovar ZnDP-F1, which
is associated with termite gut protozoa. This independent sup-
port for colocalization was not reported in previous studies using
multiplexed micro
fl
uidic digital PCR (14, 20), owing largely to the
Fig. 2.
Colocalization of ZnD2sec mRNA, ZnDP-F1 deltaproteobacterial
rRNA, and the rRNA of all bacteria in lysed protozoal preparations. (
A
) Signal
for ZnD2sec mRNA. (
B
) Signal for
ZnDP-F1
-rRNA. (
C
) Signal for rRNA of all
bacteria. (
D
F
) Composite of each pair of channels with phase. (
G
I
) White
pixels depict all areas in the image above a background threshold in a given
channel. White pixels within the orange and blue rectangles are used for the
scatterplots in
J
L
.(
J
L
) Scatterplots comparing pixel intensities for each pair
of channels localized at each pixel within a given area. Orange and blue data
are for bacterial clusters with contrasting morphologies and costaining
properties. (Scale bar: 5
μ
m.) More samples shown in
SI Appendix
, Fig. S4
.
4of6
|
www.pnas.org/cgi/doi/10.1073/pnas.1307876110
Rosenthal et al.
long-standing dif
fi
culty in visualizing mRNA expression in bac-
teria harvested from the environment (28). Here, using HCR-
FISH, we were able to visualize multiple targets within bacterial
cells harvested from the environment and measuring 0.3
×
0.6
μ
m, roughly 1/10th the volume of
Escherichia coli
.
Two additional independent studies support the localization of
genomovar ZnDP-F1 deltaproteobacteria on
Trichonympha
pro-
tists. First, a metagenomic database compiled from cytoplasmic
preparations of
Trichonympha
cells from
Z. nevadensis
guts opti-
mized for recovery of the bacterial endosymbiont
Endomicrobium
trichonymphae
(
genome.jgi.doe.gov
; GoldID Gm00182) clearly
also contained small amounts of deltaproteobacterial DNA, in
1,000 out of
40,000 gene calls. One gene was found to be
99.9% identical to the ZnD2sec FDH gene, whereas another
was 99.9% identical to the SSU RNA of genomovar ZnDP-F1.
Second, a recent study of
Z. nevadensis
trichonymphas using
micropipettes found associated bacterial SSU rRNA sequences that
are 99% similar to the sequence of genomovar ZnDP-F1 (29).
Discussion
Taken together, the results of this study increase our under-
standing of bacteria expressing key genes of CO
2
-reductive
acetogenesis in the hindguts of the dampwood termite
Zoo-
termopsis
. The guts of this insect exhibit high rates of this activity,
which is highly relevant to the mutualistic, dietary symbiosis of
termites (4, 5).
Protozoa as Niche.
Diverse microbes attached to the surface of,
or resident in the cytoplasm of, protozoa may compose a large
fraction of the total gut bacterial load, considering that protozoa
dominate up to 90% of the hindgut volume of lower termites
(21). Previous studies have proposed roles for these bacteria (26,
30
32), but the challenges faced when studying their physiology
in vitro or in situ makes many proposed interactions only spec-
ulative. Nonetheless, it has long been known that many termite
gut protozoa produce H
2
as a fermentation product (1, 33), and
thus direct interspecies hydrogen exchanges between protozoa
and attached, epibiotic, H
2
-consuming spirochetes has been hy-
pothesized (8, 31). The present study provides direct genetic
evidence of H
2
+
CO
2
metabolic potential by a bacterial epibiont
of a H
2
-producing termite gut protozoan.
Presence of Genomovar ZnDP-F1
Like Deltaproteobacterial Ribotypes
and Hydrogenase-Linked FDH Genes in Termites Wherein Gut Flagellate
Protists Are Absent.
Our results also have direct relevance to
achieving a better understanding of lignocellulose processing by
the most successful lineages of termites worldwide. In terms of
total number of species and biomass, most termites are members
of the family
Termitidae
, the so-called
higher
termites. Unlike
all known species of the less-derived wood roaches (
Crypto-
cercus
) and
lower
termite families, all of which contain diverse
cellulose-decomposing and other gut
fl
agellates, higher termite
gut communities are devoid of these
fl
agellates (1, 12). With the
loss of protozoa in ancestral higher termites, affects on ecto-
symbiotic and endosymbiotic or other bacteria metabolically
connected to protozoa might be expected. Indeed, signals for
sweeping bacterial gene losses were recently observed in hy-
drogenase-linked FDH diversity
analyses in higher termites.
Four of
fi
ve major
fdh
F
sec
and all known
Fdh
cys
lineages observed
in lower termites and the wood roach are absent in multiple
species of higher termites (19). Protozoa-associated ZnDP-F1
deltaproteobacterial species, along with the genes that they en-
code, might have been predicted to be among those lost from
higher termite gut systems; surprisingly, however, the loss of
their genes was not observed. Genomovar ZnDP-F1
like delta-
proteobacterial ribotypes have been documented in higher termite
gut communities (Fig. 1) (12, 29). Moreover, ZnD2sec (the FDH
type linked to ZnDP-F1 here) lies at the base of, and shares
evolutionary relatedness to the exclusion of all other genes in the
databases with, the single surviving lineage of FDH that has since
radiated in higher termites (Fig. 1) (19). All other FDH lineages
observed in other termites and the related wood roach, including
those associated with spirochetes here or in other studies (11, 13,
19), are absent, and metagenomic analyses of higher termite gut
communities (12, 34) have identi
fi
ed no alternative FDH can-
didates. Thus, to date, ZnD2sec-like FDHs remain the sole
Fig. 3.
Colocalization of ZnD2sec mRNA and ZnDP-F1 deltaproteobacteria
rRNA in association with termite gut protozoa. Two different
Trichonympha
-
like protozoans are shown, one protozoan per column. (
A
) Signal for
ZnD2sec mRNA. Note the auto
fl
uorescent haze properties of protozoa in the
upper right corner (posterior); a similar auto
fl
uorescent signal was seen in
unstained samples (
SI Appendix
, Fig. S6
). (
B
) Signal for ZnDP-F1 rRNA. (
C
)
Composite of the ZnD2sec mRNA signal and ZnDP-F1 rRNA signal with phase
contrast. (
D
) Signal for rRNA of all bacteria. (
E
) Composite of the ZnDP-F1
rRNA signal and all-bacteria rRNA signal with phase contrast. (Scale bar:
10
μ
m.) More samples in
SI Appendix
,Fig.S5
and
Movie S1
.
Rosenthal et al.
PNAS Early Edition
|
5of6
MICROBIOLOGY
candidate genes in higher termites for an important step in CO
2
-
reductive acetogenesis known to occur at high rates in all wood-
feeding, higher termites examined (12, 35).
Our present results suggest the possibility that relatives of
protozoan-associated ZnDP-F1 deltaprotebacteria succeeded
in weathering a major environmental perturbation by making
a major shift in their spatial niche in situ, that is, during the
major events that led to loss of the protozoal community. Fur-
thermore, their subsequent radiation into newly available, di-
verse acetogenic niches in the gut expanded their activities as
major players in H
2
+
CO
2
metabolism in higher termites (blue
arrow hypothesis; Fig. 1). This possibility, guided by indirect
phylogenetic inference, may be examined in the future using
direct approaches similar to those applied here.
In conclusion, only a fraction of the available FTHFS and FDH
genes were highly expressed in the microbial community of the
host; for example,
fi
ve or fewer bacterial species accounted for
>
85% of the observed FDH transcript pool (Table 1 and
SI Ap-
pendix
, Tables S3 and S4 and Figs. S2 and S3
). One of these
species is an active, uncultivated, and previously unappreciated
deltaproteobacterium, and the other species are speci
fi
c spi-
rochetes encoding
>
30% of the FTHFS and FDH transcripts
observed in the gut of
Zootermopsis
. These two different types live
different lifestyles and occupy distinct spatial niches in situ, that
is, swimming free in the luminal
fl
uid vs. attached directly to H
2
-
producing cellulolytic protists. Con
fi
rmation of the latter associ-
ation using HCR-FISH in a tiny deltaproteobacterium in its
natural habitat opens many doors in environmental and medical
microbiology, making it now tenable to localize in situ distinct
transcripts expressed by important bacterial cells of interest.
Materials and Methods
Termite Collection.
Worker
Z. nevadensis
termites were collected in the San
Gabriel Mountains (California) and kept as detailed in
SI Appendix
.
RNA-Seq.
Samples were prepared as described in
SI Appendix
and previously
(36). Libraries were constructed and sequenced using the standard Illumina
protocol and pipeline and aligned to inventories of FDH and FTHFS
sequences. Details, controls, and validation are provided in
SI Appendix
.
Micro
fl
uidic Multiplex Digital PCR.
Experiments were performed as described
previously (14, 20). When particulates (pellet) samples were processed, the
supernatant was discarded and pellets were resuspended in double-distilled
H
2
O, and repeatedly pipetted and lightly vortexed to disrupt large particles
and protozoa. Probe information and reaction concentrations are provided
in
SI Appendix
.
HCR-FISH.
The HCR-FISH protocol was based on a previously described pro-
tocol (27). Modi
fi
cations, probe and ampli
fi
er sequences and details of the
protocol and experimental samples are provided in
SI Appendix
. Pixel in-
tensities, the
fl
uorescent value at each pixel of the
fi
eld for each exposure
(image), were calculated for all
fl
uorescent channels.
ACKNOWLEDGMENTS.
We thank Dr. Scott Fraser and staff at the Biological
Imaging Center, Colby Calvert and staff at the Molecular Instruments
Resource, Igor Antoscheckhin and staff at the Jacobs Genetics and Genomics
Laboratory, and Drs. Eric Matson and Victoria Orphan for comments.
Research support was provided by the US Department of Energy (Grant
DE-FG02-07ER64484), the Center for Environmental Microbial Interactions at
Caltech, the National Institutes of Health (Grant 5R01 EB006192, to H.M.T.C.
and N.A.P.), the Beckman Institute at Caltech (H.M.T.C. and N.A.P.), and
a Caltech Rosen Center Bioengineering Scholarship (to V.T.).
1. Abe T, Bignell DE, Higashi M (2000)
Termites: Evolution, Sociality, Symbioses, Ecology
(Kluwer Academic, Dordrecht), pp 1
42.
2. Breznak JA, Kane MD (1990) Microbial H
2
/CO
2
acetogenesis in animal guts: Nature
and nutritional signi
fi
cance.
FEMS Microbiol Rev
7(3-4):309
313.
3. Shively JM, English RS, Baker SH, Cannon GC (2001) Carbon cycling: The prokaryotic
contribution.
Curr Opin Microbiol
4(3):301
306.
4. Breznak JA, Switzer JM (1986) Acetate synthesis from H
2
plus CO
2
by termite gut
microbes.
Appl Environ Microbiol
52(4):623
630.
5. Pester M, Brune A (2007) Hydrogen is the central free intermediate during lignocel-
lulose degradation by termite gut symbionts.
ISME J
1(6):551
565.
6. Pester M, Brune A (2006) Expression pro
fi
les of fhs (FTHFS) genes support the hy-
pothesis that spirochaetes dominate reductive acetogenesis in the hindgut of lower
termites.
Environ Microbiol
8(7):1261
1270.
7. Kane MD, Breznak JA (1991)
Acetonema longum
gen. nov. sp. nov., an H
2
/CO
2
ace-
togenic bacterium from the termite,
Pterotermes occidentis
.
Arch Microbiol
156(2):
91
98.
8. Leadbetter JR, Schmidt TM, Graber JR, Breznak JA (1999) Acetogenesis from H
2
plus
CO
2
by spirochetes from termite guts.
Science
283(5402):686
689.
9. Ottesen EA, Leadbetter JR (2010) Diversity of formyltetrahydrofolate synthetases in
the guts of the wood-feeding cockroach
Cryptocercus punctulatus
and the omnivo-
rous cockroach
Periplaneta americana
.
Appl Environ Microbiol
76(14):4909
4913.
10. Salmassi TM, Leadbetter JR (2003) Analysis of genes of tetrahydrofolate-dependent
metabolism from cultivated spirochaetes and the gut community of the termite
Zootermopsis angusticollis
.
Microbiology
149(Pt 9):2529
2537.
11. Zhang X, Matson EG, Leadbetter JR (2011) Genes for selenium-dependent and in-
dependent formate dehydrogenase in the gut microbial communities of three lower,
wood-feeding termites and a wood-feeding roach.
Environ Microbiol
13(2):307
323.
12. Warnecke F, et al. (2007) Metagenomic and functional analysis of hindgut microbiota
of a wood-feeding higher termite.
Nature
450(7169):560
565.
13. Matson EG, Zhang X, Leadbetter JR (2010) Selenium controls transcription of paral-
ogous formate dehydrogenase genes in the termite gut acetogen,
Treponema
primitia
.
Environ Microbiol
12(8):2245
2258.
14. Ottesen EA, Hong JW, Quake SR, Leadbetter JR (2006) Micro
fl
uidic digital PCR
enables multigene analysis of individual environmental bacteria.
Science
314(5804):
1464
1467.
15. Ragsdale SW, Pierce E (2008) Acetogenesis and the Wood-Ljungdahl pathway of CO(
2
)
fi
xation.
Biochim Biophys Acta
1784(12):1873
1898.
16. Zinoni F, Birkmann A, Stadtman TC, Böck A (1986) Nucleotide sequence and expres-
sion of the selenocysteine-containing
polypeptide of formate dehydrogenase
(formate-hydrogen-lyase-linked) from
Escherichia coli
.
Proc Natl Acad Sci USA
83(13):
4650
4654.
17. Thauer RK (1973) CO
2
reduction to formate in
Clostridium acidi-urici
.
J Bacteriol
114(1):443
444.
18. Stams AJ, Plugge CM (2009) Electron transfer in syntrophic communities of anaerobic
bacteria and archaea.
Nat Rev Microbiol
7(8):568
577.
19. Zhang X, Leadbetter JR (2012) Evidence for cascades of perturbation and adaptation
in the metabolic genes of higher termite gut symbionts.
mBio
3(4), e00223-12.
20. Tadmor AD, Ottesen EA, Leadbetter JR, Phillips R (2011) Probing individual environ-
mental bacteria for viruses by using micro
fl
uidic digital PCR.
Science
333(6038):58
62.
21. Berchtold M, et al. (1999) Differential enumeration and in situ localization of mi-
croorganisms in the hindgut of the lower termite
Mastotermes darwiniensis
by hy-
bridization with rRNA-targeted probes.
Arch Microbiol
172(6):407
416.
22. Leadbetter JR, Breznak JA (1996) Physiological ecology of
Methanobrevibacter cuti-
cularis
sp. nov. and
Methanobrevibacter curvatus
sp. nov., isolated from the hindgut
of the termite
Reticulitermes
fl
avipes
.
Appl Environ Microbiol
62(10):3620
3631.
23. Ohkuma M (2008) Symbioses of
fl
agellates and prokaryotes in the gut of lower ter-
mites.
Trends Microbiol
16(7):345
352.
24. Schink B, Thiemann V, Laue H, Friedrich MW (2002)
Desulfotignum phosphitoxidans
sp. nov., a new marine sulfate reducer that oxidizes phosphite to phosphate.
Arch
Microbiol
177(5):381
391.
25. Jansen K, Fuchs G, Thauer RK (1985) Autotrophic CO
2
fi
xation by
Desulfovibrio
baarsii
: Demonstration of enzyme-activities characteristic for the acetyl-CoA pathway.
FEMS Microbiol Lett
28(3):311
315.
26. Sato T, et al. (2009) Candidatus
Desulfovibrio trichonymphae
, a novel intracellular symbiont
of the
fl
agellate
Trichonympha agilis
in termite gut.
Environ Microbiol
11(4):1007
1015.
27. Choi HMT, et al. (2010) Programmable in situ ampli
fi
cation for multiplexed imaging
of mRNA expression.
Nat Biotechnol
28(11):1208
1212.
28. Pernthaler A, Amann R (2004) Simultaneous
fl
uorescence in situ hybridization of
mRNA and rRNA in environmental bacteria.
Appl Environ Microbiol
70(9):5426
5433.
29. Strassert JF, et al. (2012)
Candidatus
Ancillula trichonymphae,
a novel lineage of
endosymbiotic
Actinobacteria
in termite gut
fl
agellates of the genus
Trichonympha
.
Environ Microbiol
14(12):3259
3270.
30. Hongoh Y, et al. (2008) Complete genome of the uncultured Termite Group 1 bacteria
in a single host protist cell.
Proc Natl Acad Sci USA
105(14):5555
5560.
31. Hongoh Y, et al. (2008) Genome of an endosymbiont coupling N
2
fi
xation to cellu-
lolysis within protist cells in termite gut.
Science
322(5904):1108
1109.
32. Noda S, Hongoh Y, Sato T, Ohkuma M (2009) Complex coevolutionary history of
symbiotic Bacteroidales bacteria of various protists in the gut of termites.
BMC Evol
Biol
9:158.
33. Odelson DA, Breznak JA (1983) Volatile fatty acid production by the hindgut micro-
biota of xylophagous termites.
Appl Environ Microbiol
45(5):1602
1613.
34. He S, et al. (2013) Comparative metagenomic and metatranscriptomic analysis of
hindgut paunch microbiota in wood- and dung-feeding higher termites.
PLoS ONE
8(4):e61126.
35. Brauman A, Kane MD, Labat M, Breznak JA (1992) Genesis of acetate and methane by
gut bacteria of nutritionally diverse termites.
Science
257(5075):1384
1387.
36. Rosenthal AZ, Matson EG, Eldar A, Leadbetter JR (2011) RNA-Seq reveals cooperative
metabolic interactions between two termite-gut spirochete species in co-culture.
ISME J
5(7):1133
1142.
6of6
|
www.pnas.org/cgi/doi/10.1073/pnas.1307876110
Rosenthal et al.