Introduction
In shallow and intertidal marine habitats, endolithic microbes colonize a
variety of carbonaceous and phosphatic substrates, such as bone, shell,
coraline carbonate, ooliths, limestones, dolostone and
phosphorite outcrops (Campbell, 1983). Some of these microbes
take advantage of the natural pores or crevices in the solids, but some have
the ability to actively bore their way into the substrate. Such microborers,
also known as euendoliths (Golubic et al., 1981), build
communities that can cover as much as 50 % of the exposed solid surface
(Golubic et al., 2000) with full colonization times of virgin
substrate on the order of months
(Gektidis, 1999; Grange et
al., 2015). Several long-term geological phenomena are driven by
microborers, from the erosive morphogenesis of coastal limestones
(Purdy and
Kornicker, 1958; Schneider, 1983; Torunski, 1979; Trudgill, 1987) and the
destruction of coral reefs and other biological carbonates
(Le Campion-Alsumard et al., 1995; Ghirardelli, 2002) to
the formation of lithified laminae of welded carbonate grains in coastal
stromatolites
(MacIntyre et
al., 2000; Reid et al., 2000). Additionally, phototrophic euendoliths can
cause significant damage and shell weakening to bivalve populations
(Kaehler and McQuaid, 1999). Long-term rates of microborer-driven
carbonate dissolution, the “bioerosion” process, range between 20 and 930 g CaCO3 m-2 d-1, of clear geologic significance
(Grange
et al., 2015; Peyrot-Clausade et al., 1995; Tudhope and Risk, 1985; Vogel et
al., 2000) and may increase under future scenarios of increased atmospheric
CO2 and ocean acidification
(Tribollet et al.,
2009).
There exists a very large body of descriptive literature spanning 18
decades, largely based on microscopic observations, documenting the
biodiversity of microborers, with contributions in the microbiological,
ecological, sedimentological and paleontological fields
(Acton,
1916; Al-Thukair et al., 1994; Bachmann, 1915; Batters, 1892; Bonar, 1942;
Bornet and Flahault, 1888; Budd and Perkins, 1980; Le Campion-Alsumard et
al., 1995; Chodat, 1898; Duerden, 1902; Duncan, 1876; Ercegovic, 1925, 1927,
1930; Frémy, 1936, 1941; Ghirardelli, 2002; Golubic, 1969; Kölliker,
1859; Lehmann, 1903; May and Perkins, 1979; Nadson, 1927; Pantazidou et al.,
2006; Perkins and Tsentas, 1976; Wisshak et al., 2011). Euendoliths have
been reported among eukaryotes (fungi, green and red algae) and prokaryotes
(cyanobacteria), taxa where it may have been selected as a strategy to
escape predation from grazers, protect from UV radiation or acquire
nutrients as a tradeoff for the boring energetic cost
(Cockell and Herrera, 2008). The most common genera of
phototrophic eukaryotic euendoliths are Ostreobium and Phaeophila in the green algae, as well as
the red algal genus Porphyra (in its filamentous diploid generation, known also as
Conchocelis stage). In the cyanobacteria, the pseudofilamentous genera Hyella and Solentia are quite
common
(Al-Thukair,
2011; Al-Thukair et al., 1994; Al-Thukair and Golubic, 1991; Brito et al.,
2012; Campion-Alsumard et al., 1996; Foster et al., 2009; Golubic et al.,
1996), as are some forms in the simple filamentous genus Plectonema (Chacón et
al., 2006; Pantazidou et al., 2006; Tribollet and Payri, 2001; Vogel et al.,
2000). Morphologically complex cyanobacteria such as Mastigocoleus testarum (Golubic and
Campion-Alsumard, 1973; Nadson, 1932; Ramírez-Reinat and Garcia-Pichel,
2012a) complete the list of common euendoliths. Less common genera of
euendolithic cyanobacteria include Cyanosaccus
(Pantazidou et al.,
2006), Kyrtuthrix (Golubic and Campion-Alsumard, 1973) and Matteia
(Friedmann et al., 1993). To date, these genera were all
assigned based upon morphological criteria and could represent morphological
variations of the same types (Le Campion-Alsumard and Golubic,
1985), highlighting the need to reassess the diversity of euendolithic
cyanobacteria using a combination of characters including genetic markers.
Modern genomic methods for community fingerprinting have, more recently,
been applied to provide a complementary and more comprehensive description
of endolithic communities. Some studies, focused on phototrophs from marine
carbonates, revealed that, while some biodiversity had been missed by
deploying morphological studies, there was also congruency between DNA-based
surveys and the traditional literature
(Chacón et al.,
2006; Ramírez-Reinat and Garcia-Pichel, 2012b). DNA-based studies have
revealed that the endolithic habitat at large can harbor complex communities
of microbes, in addition to euendoliths, particularly when the substrate
rocks are naturally porous or when they have been rendered porous by the
action of euendoliths themselves. Horath et al. (2006), for example,
investigating terrestrial endolithic communities in dolomite outcrops in the
Alps, found a large diversity of presumably chemotrophic bacteria and
archaea, in addition to expected green algae and cyanobacteria. Similar
conclusions could be drawn from the work of De la Torre et al.
(De la Torre et al., 2003) on Antarctic sandstone
cryptoendoliths, those of Walker and colleagues (Walker et
al., 2005; Walker and Pace, 2007) on terrestrial limestones, sandstones and
granites or the recent contribution of (Crits-Christoph
et al., 2016) who used a metagenomic approach to investigate the
chasmoendolithic communities of the hyper-arid Atacama Desert. However, no
high-throughput sequencing studies are available on the globally significant
intertidal endolithic communities.
Tribollet (2008) provided an account of the dynamic changes in microborer
community composition taking place after coral death, which obviously
constitute a true succession in the ecological sense, with pioneer
euendoliths (such as Mastigocoleus testarum) and secondary colonizers such as Ostreobium quekettii
and Plectonema terebrans, as well as
fungi
(Grange
et al., 2015; Tribollet, 2008). During laboratory studies with the
cultivated strain of Mastigocoleus testarum strain BC008, used as a model to understand the
physiology of cyanobacterial boring
(Garcia-Pichel
et al., 2010; Guida and Garcia-Pichel, 2016; Ramírez-Reinat and
Garcia-Pichel, 2012b), we found that, among the carbonates, this strain
excavated most rapidly into various types of calcite and aragonite minerals
(CaCO3). It could bore slowly into strontianite (SrCO3) but was
unable to penetrate into magnesite (MgCO3), dolomite (CaMgCO3),
witherite (BaCO3), rhodochrosite (MnCO3), siderite (FeCO3) or
ankerite (CaFe(CO3)2; Ramírez-Reinat and
Garcia-Pichel, 2012a). However, literature reports do exist detailing
microborings in modern and fossil dolomitic substrates (see, e.g.,
Campbell, 1983; Golubic and Lee, 1999). Similar substrate
preferences have also been observed for phosphates: M. testarum strain BC008 did not
bore into calcophosphatic substrates, including hydroxyapatite, vivianite or
dentine, yet the literature is replete with reports of cyanobacterial
microborings on biotic and abiotic phosphatic rocks
(Soudry
and Nathan, 2000; Underwood et al., 1999; Zhang and Pratt, 2008). The
expression of such a mineral substrate preference among the pioneer
euendolithic cyanobacteria could principally drive the whole community
towards a different successional sequence with distinct mature community
assemblages and metabolic potentialities. We wanted to find out whether
evolutionary specialization has resulted in a highly adapted endolithic
flora for each type of mineral substrate and whether there exist specialized
apatite-borers, dolomite-borers or carbonate-borers in nature.
In order to answer these questions, we investigated in depth the marine
endolithic communities of Isla de Mona (PR), a small, uninhabited Caribbean
island offering a variety of coastal cliffs composed of dolomite and
limestone, as well as raised aragonitic and phosphatic reefs, with the dual
purpose to (i) describe the microbial diversity of intertidal endolithic
community at high resolution and (ii) test the effects of substrate
composition on community structure in a single geographic location with
common bathymetry (the intertidal notch), controlling for other known major
determinants of community composition.
Materials and methods
Sampling site and procedure
Samples were obtained from Isla Mona (18.0867∘ N,
67.8894∘ W), a small (11 by 7 km) carbonate island 66 km W of
Puerto Rico. Isla Mona is a protected habitat and all necessary permits were
acquired from the Departamento de Recursos Naturales y Ambientales prior to
arrival. The present study did not involve endangered or protected species.
Endolithic communities were obtained by sampling different locations from
nine separate island localities. Rock samples containing endolithic biomass,
verified using a digital field microscope, were chipped off from large
boulders and rock walls using a standard geological hammer. The hammer was
thoroughly washed with surrounding sea water at each sampling point.
Material was predominantly collected within the boring notch of the
intertidal zone. Bathymetric samples were collected via SCUBA diving at
sample site K at depths of 3.5, 4.6, 7 and 9.1 m. Each sample was
broken into three pieces and each biological replicate was stored in a sterile
50 mL falcon tubes; one replicate was air dried for mineralogical analysis,
one was kept viable in seawater for strain isolation and another was
preserved in situ in 70 % ethanol for DNA extraction. Air drying and alcohol
preservation were done in the field. Samples were shipped at room
temperature, in the dark for 5 days, and, upon arrival in the lab, the
preserved samples were immediately stored at -20 ∘C until extractions
were performed. Aliquots of local seawater were collected at sample site K
and filtered through 0.22 µm syringe filters into sterile 50 mL
falcon tubes. After 5 days of transit at room temperature in the dark, the
seawater sample was stored at 4 ∘C in the dark for an additional week
before being processed for physicochemical analysis.
Bulk powder X-ray diffraction and elementary analyses
A fragment of each sample was ground down to powder in 100 % ethanol. XRD
patterns were collected using Panalytical X'Pert Pro diffractometer mounted
in the Debye–Scherrer configuration with a CuKα monochromatic X-ray
source. Data were recorded in continuous scan mode within a
10–90∘ 2θ range. X'Pert High Score plus software was used
to identify mineral phases and their relative concentration using the
automatic Rietveld refinement method implemented in the software under
default parameters. The elementary composition of the rocks and water sample
analyses were performed by the Goldwater Center at Arizona State University
using an inductively coupled plasma optical emission spectrometer (ICP-OES),
Thermo iCAP6300.
Isla de Mona setting: (a) simplified geological map modified from
that of Briggs and Seiders (1972), showing the locations of
the sampling sites. (b) Sky view of Isla de Mona: the cliff is composed of
the Isla de Mona Dolomite topped by the Lirio limestone and the Isla de Mona
lighthouse is visible. (c–d) Views of Isla de Mona coastal area: samples were
taken from isolated boulders (c) and directly from the cliff (d), at the notch
(white arrows c–d) or on the raised reef flat (c–d).
Total genomic DNA purification
The surface of the ethanol fixed samples was brushed vigorously with a
sterile toothbrush and sterile Milli-Q water to remove epilithic material. A
chip of 8 cm3 was further ground in a sterile mortar as recommended
by Wade and Garcia-Pichel (2003); 0.5 g of
the obtained coarse powder was then transferred into the bead tube of the
MoBio PowerPlant Pro kit (Mo Bio Laboratories, Inc., Carlsbad, CA, USA). The
first lysis step of the kit was modified by homogenizing bead tubes
horizontally at 2200 rev min-1 for 10 min and seven freeze–thaw cycles
(Wade and Garcia-Pichel, 2003). The next
steps of the extraction were conducted following the MoBio PowerPlant Pro
kit following manufacturer's guidelines.
16s rRNA gene library preparation and sequencing
The 16S rRNA gene V3–V4 variable region was targeted using PCR primers
341F (CCTACGGGNGGCWGCAG) and 806R (GGACTACVSGGGTATCTAAT) with a barcoded
forward primer. The PCR amplification was performed using the HotStartTaq
Plus Master Mix Kit (Qiagen, USA) under the following conditions:
94 ∘C for 3 min, followed by 28 cycles of 94 ∘C for
30 s, 53 ∘C for 40 s and 72 ∘C for 1 min,
followed by a final 5 min elongation step at 72 ∘C. PCR product
were further purified and pooled into a single DNA library using the
Illumina TruSeq DNA library preparation protocol. This library was further
sequenced on a MiSeq following the manufacturer's guidelines. The library
preparation, sequencing paired ends assembly and first quality trimming
(with phred score of Q25 cutoff) were performed by MR DNA (www.mrdnalab.com,
Shallowater, TX, USA).
The 16S rDNA sequences from the newly cultured euendolithic strains were
retrieved using the PCR condition and primers described by
Nübel et al. (1997) followed by Sanger sequencing.
Briefly, the primers used were the forward Cya106F (CGGACGGGTGAGTAACGCGTGA)
and an equimolar mixture of the Cya781R(a) (GACTACTGGGGTATCTAATCCCATT) and
Cya781R(b) (GACTAC AGGGGTATCTAATCCCTTT) as reverse. The PCR amplification
was performed using the GoTaq enzyme and master mix (Promega, Madison, USA)
at 1X concentration. The amplification conditions were as follows: after an
initial denaturation step 94 ∘C for 5 min, 35 PCR amplification
cycles were performed, each consisting of a 1 min denaturation step at
94 ∘C, a 1 min annealing step at 60 ∘C and a 1 min
elongation step at 72 ∘C.
Operational taxonomic unit
(OTU) table building and analysis
Sequences were further processed using the Qiime version 1.9
(Caporaso et al., 2010). The sequences were first run through
the split_libraries.py script under the default parameter that includes barcode removal,
quality filtering (sequences of less than 200 bp or with homopolymer runs
exceeding 6 bp were removed) and split of the dataset per sample. The output
file was further processed through the pick_open_reference_otus.py script using the default parameters
except for the taxonomic assignment that was done by the RDP classifier (see
parameter file in Supplement for more details). This step
clustered the sequences at a similarity threshold of 97 %
(Edgar, 2010) to build OTUs, assign their taxonomy and further report specific abundance for each
sample into an OTU table. Because in this case we were not interested into
the rare biosphere but focused on the most abundant OTUs and how they vary,
we filtered the OTU table to remove the rare OTUs. The OTUs retained were
those that occurred in at least 5 samples among the 34 analyzed or that
represent more than 0.1 % of the total sequences found in a particular
sample. By doing this, we eventually analyzed 90 % of all the single
sequences but only 11 % of the initial OTUs. The Qiime script
summarize_taxonomy_through_plots.py was run on the final OTU table for all the prokaryotes and for the
cyanobacteria only (filtering out the chloroplasts) in order to build the
summarized microbial community composition bar graphs displayed in
Fig. 2.
Accession numbers
One representative sequence per OTU was deposited to gene bank under the
accession numbers KT972744-KT981874. The 16S rDNA sequences of the new
euendolithic strains described in this article received the following
accession numbers: Ca. Pleurinema perforans IdMA4 [KX388631], Ca. Mastigocoleus
perforans IdM [KX388632] and Ca. Pleurinema testarum RPB [KX388633].
Mineral composition and microbial community structure of Isla de
Mona intertidal outcrops. Each line corresponds to one sample. (a) Mineralogical composition as retrieved
by bulk powder XRD. (b) Distribution
of 16 rDNA OTUs taxonomically assigned at the phylum level and (c) associated
chao1 richness metric. This reflects the total microbial community
structure. (d) Distribution of the cyanobacterial 16 rDNA OTUs assigned at
the phylum level, excluding chloroplasts and associated chao1 richness
metric for cyanobacteria (e).
Meta-analysis of microbial communities
For comparison, raw sequences from datasets ID 662/678/809/627/713/925 were
retrieved from the Qiita repository along with their mapping table. All
these studies used comparable sequencing depth, technology and targeted the
same region of the 16 rRNA gene compared to the present study. Two samples
from Alchichica cyanobacteria-dominated microbialites communities
(Couradeau et al., 2011) were
processed in parallel to the Isla de Mona samples (same extraction
methodology, sequenced in the same MiSeq run) and also included in this
analysis. The sequences were all aggregated into a master file that was
processed in Qiime version 1.9 (Caporaso et al., 2010). The
same exact procedure than the one described above was used to pick OTUs.
Again we retained the OTUs that occurred at least in five samples. We ran the
jackknifed_beta_diversity.py pipeline using the Bray–Curtis metrics under default parameters. The
obtained distances were used to cluster samples under a UPGMA hierarchical
clustering method and 5000 sequences were included in each jackknifed subset
in order to generate nodes support.
Differential abundance of OTUs analyses
To determine if some OTUs were more associated with certain type of substrates
we ran the differential_abundance.py of the Qiime 1.9 package (Caporaso et al., 2010)
using the DESeq2 method (Love et al., 2014), under
a negative binomial generalized linear model. This method was initially
developed to assess the differential gene expression from RNA-seq data but
can be applied to any count matrix data such as OTU tables
(Love et al., 2014). It was recently implemented
for the treatment of 16S rDNA OTU table and has been widely used since (e.g.,
Debenport
et al., 2015; Pitombo et al., 2015) because it (i) is a sensitive and
precise method, (ii) controls the false positive rate
(Love et al., 2014) and (iii) uses all the
power of the dataset without the need to rarefy the OTU table
(McMurdie and Holmes, 2014). After checking the good agreement
between the fit line and the shrunken data on the dispersion plot, a Wald
test was applied to each OTU to reject the null hypothesis (p < 0.05)
because the logarithmic fold change between treatments (i.e., in our case
type of mineral substrate) for a given OTU is null.
Phylogeny reconstruction
In order to determine which of the cyanobacterial OTUs of the dataset were
possible euendolithic organisms, we built a phylogeny to assess their
proximity to proven boring cultured strains. The maximum-likelihood
phylogenetic reconstruction was performed using TREEFINDER
(Jobb et al., 2004) under a general time reversible
(GTR) and a four-category discrete approximation of a Γ
distribution. Bootstrap values were inferred from 1000 replicates. The
sequence dataset used for the reconstruction was first aligned with MAFFT
(Katoh et al., 2005) and then manually
checked and trimmed using the MUST package
(Philippe, 1993).
Results and discussion
Geological setting of Isla de Mona outcrops
The island is an 11 by 7 km emerged platform of Miocene Isla de Mona
Dolomite (up to 80 m thick) topped by a thinner (up to 40 m) layer of
Miocene Lirio limestone (Briggs and Seiders, 1972;
Frank et al., 1998). It is partially surrounded in its southern and
southwestern shores by a Pleistocene raised reef flat, mostly composed of
biogenic carbonates (Fig. 1). The island also harbors secondary phosphorite
deposits formed by the diagenetic alteration of guano, most typically
associated with an extensive system of karstic caves at the interface of
limestone and dolostone (Briggs, 1959). Isla de Mona was never
continuously inhabited. The island was mostly used as a guard post over the
Mona Passage throughout the 20th century, and declared a nature
preserve in 1993 (National Parks Register, USA). The coastal area has been
protected from disturbance ever since. We took advantage of this unique and
pristine geological setting to sample dolostones, limestones and
phosphorites exposed to similar environmental conditions. We analyzed a set
of 34 samples consisting of pieces of exposed rock, in most cases taken
directly at the intertidal notch. Location of sampling sites are in the
simplified geological map in Fig. 1a. The mineralogical composition of
each sample (Fig. 2), determined using bulk powder X-ray diffraction,
confirmed the presence of apatite (Ca5(PO4)3(OH,Cl,F)),
dolomite (CaMg(CO3)2), calcite (CaCO3) and
aragonite(CaCO3) in various proportions depending of the sampling site
(Fig. 2a).
Euendolithic cyanobacterial strains used to assign potential roles
to OTUs.
Strain name
Order
Reference
presence in
Isolation source
Bores in culture
Reference
sequence
this dataset
culture
Mastigocoleus testarum
Stigonematales
DQ380405
yes
Cabo Rojo carbonate,
yes
Chacón et al. (2006)
Puerto Rico
Solentia sp. HBC10
Pleurocapsales
EU249126
no
Stromatolite bahamas
yes
Foster et al. (2009)
Hyella sp. LEGE 07179
Pleurocapsales
HQ832901
yes
Rocky Moledo do
not tested*
Brito et al. (2012)
Minho beach (Portugal)
Ca. Pleurinema perforans IdMA4
Pleurocapsales
KX388631
yes
Isla de Mona outcrop
yes
this study
Ca. Mastigocoleus perforans IdM
Stigonematales
KX388632
yes
Isla de Mona outcrop
yes
this study
Ca. Pleurinema testarum RPB
Pleurocapsales
KX388633
Yes
Puerto Peñasco
yes
this study
Coquina reef
* Hyella sp. LEGE 07179 was isolated from inside a patella shell where it was
identified as a true borer by the authors but its boring ability was never
tested again in the lab.
Hierarchical clustering analysis (UPGMA) of bacterial community
composition in various settings based on pairwise Bray–Curtis distance
metrics. The robustness of the topology was assessed through jackknife
repeated resampling of 5000 sequences. The number of samples in a given
collapsed tree branch are in parentheses, while the numbers in brackets are
the Qiita dataset ID number.
The endolithic microbial communities
We studied the endolithic community composition by analyzing the 16S rDNA
diversity present in total genomic DNA extracted from the rock after
aggressively brushing away epilithic growth from the external sample
surface. The 16S rDNA sequences were obtained after specific PCR
amplification and Illumina-based high-throughput sequencing, with one
library per sample (Table S2 in the Supplement). We clustered sequences into OTUs based on a 97 % similarity criterion and further
filtered the dataset to remove the rare OTUs, focusing our study on OTUs
that occurred in at least five separate samples or those that made up more
than 0.1 % of all sequences in any one sample. Bacterial OTU richness in
these samples was 4058 ± 1252, as given by the chao1 metric (Fig. 2c). Thus, comparatively our endolithic communities are of rather low
diversity, an order of magnitude lower than current estimates assigned to
bulk soil bacterial communities (Roesch et al., 2007), but
similar to other microbial communities such as biological soil crusts
(Couradeau et al., 2016), microbial mats
(Hoffmann et al., 2015) or stromatolites
(Mobberley et al., 2011) that are dominated by cyanobacterial
primary producers. This suggests that endolithic habitat nurtured by the
presence of cyanobacterial primary producers can support the development of
a high diversity of microorganisms even when this type of habitat is expected
to be nutrient limited due to its low connectivity with sea water
(Cockell and Herrera, 2008). Taxonomic assignment of the
OTUs on the basis of the Greengene database
(McDonald et al., 2012) allowed us
to reconstruct the endolithic prokaryotic communities from Isla de Mona at
various level of taxonomic resolution. At the phylum level (Fig. 2b), the
analysis revealed complex microbial communities with numerically significant
populations of bacteria other than cyanobacteria: Proteobacteria, Chloroflexi, Actinobacteria and Bacteroidetes. In fact, the
contribution of cyanobacteria to the total sequence richness was only 12 ± 3 %. These communities clearly host not only a large number of
bacterial types but also a wide diversity of phylogenetic and metabolic
potential beyond oxygenic photosynthesis. Clearly, mature endolithic
cyanobacterial communities in this study are much more complex than the
majority of the literature to date (for example, the exhaustive descriptive
literature review in the introduction does not report beyond cyanobacteria
and eukaryotic algae). While it is proven that some axenic cyanobacteria are
able to initiate excavation on virgin substrate
(Ramírez-Reinat and Garcia-Pichel, 2012a), it is interesting
to entertain that in such complex communities, other metabolic activities
(of co-occurring microorganisms), particularly those that result in pH
changes, might play a significant role on the determination of the local
saturation index of the carbonate mineral
(Baumgartner
et al., 2006; Dupraz et al., 2009; Dupraz and Visscher, 2005) and in this
way influence the overall mineral excavation yield or rates. At this level
of taxonomic resolution, we did not detect any significant association of
substrate mineralogy and community composition (as judged by nonsignificant
Spearman's ρ when comparing each phylum's relative abundance to
mineralogical composition; not shown).
Because endolithic communities have not received much attention, we
integrated our dataset into a meta-analysis of various cognate microbial
communities, for which technically comparable datasets were publicly
available (http://qiita.microbio.me). To do so, we aggregated all the
sequences from the selected Qiita datasets into a single file that was used
to pick and cluster 16S rDNA OTUs anew and conducted similarity analyses.
The meta-community analysis revealed that endolithic communities clustered
together and apart from other types of phototrophic microbial communities
in terms of composition (beta-diversity). The fact that they clustered
together indicates that their microbial assemblages are recognizable and
distinct beyond just their belonging to the marine habitat itself, in a
microbiological and presumably adaptive way. However, at this stage we cannot
exclude that the observed pattern could represent a biogeographical island
effect. Further studies involving a larger dataset of endolithic communities
will be necessary to disentangle the local signature controlled by
environmental parameters from the endolithic signature presumably universal
to all endolithic communities. Interestingly, our endolithic community
samples could be separated into two self-similar clades (A and B in Fig. 3) but
so far we cannot ascertain a factor that would drive the observed separation
beyond the fact that it is not substrate type. While it would be of interest
to compare our communities to other endolithic communities, such as those
studied by
Chacón
et al. (2006), Crits-Christoph et al. (2016), Horath and Bachofen (2009) and De la
Torre et al. (2003), this is not technically possible given that all of
those studies used alternative methods for community analyses (Clone
libraries, DGGE, metagenomes) that do not allow direct comparisons.
A diverse cyanobacterial community dominated by likely euendoliths
Because they comprise the pioneer microborers and primary producers within
many endolithic communities, cyanobacteria are of particular interest in
this study. We therefore analyzed cyanobacteria at a higher resolution. The
cyanobacterial community appeared quite diverse with a specific chao1
richness of 484 ± 184, certainly much more genetic diversity among
this group than could be surmised from the wealth of microscopically based
accounts in the botanical literature
(Chazottes
et al., 1995; Pantazidou et al., 2006; Sartoretto, 1998; Tribollet et al.,
2006). In these studies, one typically finds reports of anywhere from one to
five
morphotypes. Even accounting for the fact that morphotypes typically
underestimate genetic diversity by a significant fraction
(Nübel et al., 1999), this is a very large
underestimation of oxygenic phototroph diversity. Phylotypes assignable to
the orders Pseudanabaenales, Chrooccocales, Nostocales and Stigonematales were most common and widespread. Again, no pattern linking
mineralogy to microbial community composition arose at this taxonomic level,
as judged by the nonsignificant Spearman's ρ when comparing the
relative abundance of each cyanobacterial to mineralogical composition (not
shown). A combination of literature search and additional efforts of
cultivation and genetic characterization of isolates allowed us to attempt
the assignment of a true-boring (euendolithic) role to some of our
cyanobacterial OTUs (Table 1 and Figs. S2–S3 in the Supplement). Interestingly, out of the
five most abundant OTUs in our combined dataset, four (NR_OTU741, OTU 842393, NR_OTU193 and OTU 351529) could be deemed
as likely euendoliths, given their close phylogenetic affiliation to
cultivated isolates proven in the laboratory to be able to bore. The fifth
most abundant OTU (OTU 186537) fell between Mastigocoleus testarum BC008 (a proven euendolith) and
Rivularia atra (not described as boring in the literature), so its capacities remain
unclear. Notably, the most abundant OTU, NR_OTU741 in our set
is virtually indistinguishable from one of our isolates obtained from the
same samples, the boring strain Ca. Pleurinema perforans IdMA4 (similarity
> 99 %), which is not only the most abundant cyanobacterial OTU
but also the second most abundant bacterial OTU overall in our dataset.
Overall the seven OTUs that could be assigned as possible euendolith based on
their phylogenetic proximity to known microborers account for 0.8 to
73 % (average value 29 %) of the total number of sequences depending on
the sample considered. These results suggest that euendoliths compose a
major fraction of the community, one that not only represents an initial set
of pioneers but also maintains relevance even after bioerosive
degradation and reworking of the mineral substrates allow the colonization
of newly made pore spaces by non-boring endoliths.
On analyzing the diversity of the possible euendoliths detected in this
dataset, we realized that while many of the most common known genera of
cyanobacterial microborers are represented and abundant, the thin,
filamentous Plectonema terebrans is not. This was surprising because Plectonema terebrans has always been described as
an important member of the euendolithic community, accounting for up to
80 % of the total of microborer biomass
(Tribollet,
2008), and is found associated with Mastigocoleus testarum. This apparent paradox is likely not due
to the absence of the organism but rather to failure to properly identify it
molecularly due to the lack of reference sequences in the databases. Indeed
morphotypes resembling Plectonema terebrans were visually recognized but not detected
molecularly in the extensive study of euendolithic cyanobacteria from
various locations by (Ramírez-Reinat and
Garcia-Pichel, 2012b). In the present dataset, Plectonema could have been assigned to
another member of the Oscillatoriales, such as Phormidium or Halomicronema, which represent 10 and
4.6 %, respectively, of the cyanobacterial sequences. A bona fide isolate proven
to bore in the lab will be needed before we can advance regarding the
presence and abundance of simple filamentous euendolithic cyanobacteria
anywhere. Among the cyanobacterial taxa detected, the following have never
been reported to be true borers: Gloeobacterales, Nostocaceae,
Acaryochlorales, Cyanobacteriaceae, Spirulinaceae and Pseudanabaenales. In all,
these cyanobacteria contribute at least to some 43 ± 20 %,
indicating that a significant proportion of the community is likely made up
of adventitious endoliths. A study of the temporal dynamics of colonization
could help understand the true role of each taxon.
Differential abundance of cyanobacterial OTUs in Ca carbonates
(calcite–aragonite) n= 14 vs. CaMg carbonate (dolomite) n= 13 samples.
This plot was constructed using the DESeq2 method. It displays the average
normalized counts per OTU as a measure of abundance against the log2 fold
difference. The OTUs that were significantly differentially abundant in the
two conditions (p < 0.05) are represented as open circles; the other
ones are displayed as close symbols. Positive values indicate enrichment
towards CaMg carbonate and negative values indicate enrichment towards
Ca Carbonate. The OTU ID and taxonomical assignment of the most abundant
OTUs is displayed on the right. The stars tag the possible euendolithic OTUs
as determined by phylogenetic proximity to known microborers (Fig. S3).
Differential abundance of cyanobacterial OTUs in Ca carbonate
(calcite–aragonite) n= 14 vs. Ca phosphate (apatite) n= 3 samples This
plot was constructed using the DESeq2 method. It displays the average
normalized counts per OTU as a measure of abundance against the log2 fold
difference. The OTUs that were significantly differentially abundant in the
two conditions (p < 0.05) are represented as open circles; the other
ones are displayed as close symbols. Positive values indicate enrichment
towards Ca phosphate and negative values indicate enrichment towards
Ca Carbonate. The OTU ID and taxonomical assignment of the most abundant
OTUs is displayed on the right. The stars tag the possible euendolithic OTUs
as determined by phylogenetic proximity to known microborers (Fig. S3).
Substrate preference among cyanobacteria
We knew from the experimental study of the model euendolith Mastigocoleus testarum strain BC008 that
this particular organism exhibits a clear boring substrate preference. It
bores into Ca carbonates (like aragonite and calcite) and to a lesser extent
Sr carbonate (strontianite), but not into CaMg carbonate like dolomite
(Ramírez-Reinat and Garcia-Pichel, 2012a). This strain
remains the single case where the boring preference has been directly
tested, but it is unknown whether this preferential behavior is representative of
euendoliths at large. Only a few studies examined endolithic communities
colonizing dolostone; Jones (1989) provided the first comparison
of endolithic communities from dolostones and limestones from Grand Cayman
Ironshore. He observed that dolostones were less colonized by endoliths than
limestones and concluded that the bioerosion of limestones was faster due to
the more abundant endolithic flora, while the erosion pattern of the
dolostone was slower and allowed the development of more epiliths. When
looking at the endolithic microbial diversity of terrestrial dolostones,
Horath et al. (2006) found the same
cyanobacterial genera than the ones typically described on freshwater
limestones substrates (Norris and Castenholz, 2006) while
Sigler et al. (2003) concluded that
the endolithic dolostone phototrophic community resembled other
desiccation-tolerant endolithic communities. The question of whether there
truly exists a specialized community associated with dolostone vs. limestone
remains clearly open.
Our own data showed no specificity for substrate at family level,
highlighting the need to analyze this at a phylogenetically deeper
resolution. To do so, we analyzed how cyanobacterial OTUs were
differentially represented in sample subsets from contrasted mineralogical
substrates using the DESeq2 method (Love et al.,
2014). This method was developed to analyze RNA-seq datasets but can be used
on any count matrix, such as an OTU table. This statistical framework is
sensitive and precise and does not involve rarefying the dataset to an even
sampling depth, so that the entire statistical power of the data is
accounted for (McMurdie and Holmes, 2014). We used it to determine
whether any given OTU is significantly differentially represented in a
particular subset of samples sharing a common mineralogical substrate
compared to another set. When comparing OTUs detected in samples which were
mineralogically dominated by Ca carbonates (calcite or aragonite, n= 13)
with those that were dolomitic in nature (CaMg carbonate, n= 14), we found
31 OTUs to be significantly enriched in Ca-carbonate substrates (p <
0.05; corresponding to log2 fold difference > |2.83|), while 22 preferred dolomite with p < 0.05, out of
1039 cyanobacterial OTUs considered. Results suggest that substrate
preferences are found when one looks at fine taxonomic resolution and that
some likely euendoliths show such preference: Mastigocoleus testarum close relative
NR_OTU193 prefers the Ca-carbonate pole (log2 fold
difference =|3.4|) while another possible euendolith
NR_OTU741 belonging to the Pleurocapsales clearly prefers dolomite
(log2 fold difference =|1.7|). It is also clear that
for most of the OTUs, either there is not sufficient resolution at the 16S
rDNA level to detect it or, more parsimoniously, these OTUs represent taxa
that can colonize various substrates. Many in this group of OTUs are not
differentially represented on a particular substrate type, suggesting that
they may be adventitious endoliths that do not bear the burden of boring
into the substrate and can potentially colonize any substrate. However, at
least some of these represent most likely euendoliths (NR_OTU4, OTU 351529 and OTU 842393) and still are not differentially
represented with respect to the mineral phase they colonize.
Using the same method, we then compared Ca-carbonate-dominated samples
(n= 14) to Ca-phosphate-dominated samples (n= 3). Although the paucity of
phosphate samples restricted our statistical power, we were still able to
identify 81 OTUs that were statistically significantly enriched on the
phosphatic substrate (p < 0.05) side, while only 21 were enriched in
carbonates (p < 0.05; Fig. 5). This suggests an asymmetrical
effect of carbonate vs. phosphate substrate types, the latter being a more
powerful driver of differential abundance among cyanobacteria. However, in
this case, the majority of OTUs, including some of the most abundant, were
widespread across different substrate types. Mastigocoleus sp. (NR_OTU193) appeared
clearly enriched in the carbonates (log2 fold difference =|3.8|), while the other potential borers including the Pleurocapsales
OTUs did not exhibit statistically significant differential abundance with
substrate.
In all, these results suggest that some cyanobacteria do have a substrate
preference and that these preferences sometimes occur among closely related
clades (like NR_OTU193 and NR_OTU4), which do
exhibit differential occurrence. These comparisons highlight the
differential role of the cationic vs. the anionic mineral component.
NR_OTU193 for instance showed a higher rate of occurrence
when testing for both components, suggesting that it prefers calcium over
magnesium in terms of cation and carbonate over phosphate as an anion. In contrast, NR_OTU741 only appeared differentially
represented when the cationic part of the mineral varied. Finally, it is
important to note that only a small fraction of the cyanobacterial community
seems to be influenced by the substrate, 3.5 % of the total number of
species on average accounting for 16 ± 4 % of the total number of
cyanobacterial sequences analyzed. These results are consistent with the
idea that borers may be specialized, but ancillary endoliths are not. The
substrate specialization of euendoliths may be due to the physiological
requirements of excavation into specific mineral types. Future endolithic
community metagenomic reconstructions and comparisons could aid in the
identification of alternative pumps that may be specific to mineral types.
Implications for the diversity of the boring mechanism and substrate-driven evolution of euendoliths
A question that follows naturally from the previous findings is how such a
substrate preference may relate to the physiological mechanism of boring.
The model strain Mastigocoleus testarum BC008 is clearly specialized in the excavation of calcium
carbonate through the uptake of calcium anions at the boring front and their
active transport along the filament toward the surface
(Garcia-Pichel
et al., 2010; Garcia-Pichel, 2016). In culture, M. testarum strain BC008 could
not bore into dolomite or magnesite. In agreement with this, the closest
phylogenetic allies to this strain in our communities, NR_OTU193, did also show a higher rate of occurrence in calcium carbonates as
compared to magnesium carbonate. Experiments with natural endolithic
communities using calcium pump inhibitors have shown that the calcium-based
mechanism is commonly at work in many localities but, at least in one case,
boring was impervious to inhibition, pointing to the potential existence of
mechanistic diversity (Ramírez-Reinat and
Garcia-Pichel, 2012b). Because we could not detect preferential enrichment
of bona fide euendoliths in the phosphate compared to the carbonate substrates, we
must assume that the mineral anion is not a strong determinant of substrate
choice in these communities. The boring mechanism described for M. testarum BC008 is in fact
only dependent on the nature of the cation and could work in principle on
calcium phosphates as well, and yet M. testarum strain BC008 did not bore into pure
hydroxyapatite in the laboratory. These contrasted findings highlight that
there must be factors other than the cationic part of the mineral
determining the excavation ability of a particular strain and that the
boring mechanism proposed for M. testarum strain BC008 might be incompletely described.
Other mechanisms have been suggested to explain boring mechanism which have
been invalidated for the model organism M. testarum strain but may prove themselves
valuable for other taxa. The dissolution of carbonate mineral by acid
excretion was proposed by Haigler (1969) and
Golubic et al. (1984). This mechanism
could involve spatial and temporal separation of photosynthesis vs. respiration
by cyanobacteria or acid production as a byproduct of other heterotrophic
bacteria activity (Garcia-Pichel, 2006). These
hypotheses will need to be re-evaluated for other euendoliths as well as in
natural communities.