tales from the tomb: the microbial ecology of exposed rock...

Tales from the tomb: the microbial ecology of exposedrock surfaces

Tess E. Brewer1,2 and Noah Fierer1,3*1Cooperative Institute for Research in Environmental

Sciences, University of Colorado, Boulder, CO, USA.2Departments of Molecular, Cellular, and Developmental

Biology.3Department of Ecology and Evolutionary Biology,

University of Colorado, Boulder, CO, USA.

Summary

Although a broad diversity of eukaryotic and bacterial

taxa reside on rock surfaces where they can influence

the weathering of rocks and minerals, these commu-

nities and their contributions to mineral weathering

remain poorly resolved. To build a more comprehen-

sive understanding of the diversity, ecology and

potential functional attributes of microbial communi-

ties living on rock, we sampled 149 tombstones

across three continents and analysed their bacterial

and eukaryotic communities via marker gene and

shotgun metagenomic sequencing. We found that

geographic location and climate were important

factors structuring the composition of these commu-

nities. Moreover, the tombstone-associated microbial

communities varied as a function of rock type,

with granite and limestone tombstones from the

same cemeteries harbouring taxonomically distinct

microbial communities. The granite and limestone-

associated communities also had distinct functional

attributes, with granite-associated bacteria having

more genes linked to acid tolerance and chemotaxis,

while bacteria on limestone were more likely to be

lichen associated and have genes involved in photo-

synthesis and radiation resistance. Together these

results indicate that rock-dwelling microbes exhibit

adaptations to survive the stresses of the rock

surface, differ based on location, climate and rock

type, and seem pre-disposed to different ecological

strategies (symbiotic versus free-living lifestyles)

depending on the rock type.

Introduction

Mineral and rock weathering are important controls on soil

formation and terrestrial biogeochemical cycles. In addition

to shaping the surface of the planet and the availability of

inorganic nutrients, weathering also affects humans

directly by influencing water quality, soil fertility and the

durability of buildings and monuments (Mailloux et al.,

2009; Mapelli et al., 2012). Weathering is a product of both

abiotic and biotic factors, with microbes playing a signifi-

cant role in regulating mineral weathering rates (Banfield,

1999; Mapelli et al., 2012). Microbes are common inhabi-

tants of rock and mineral surfaces and there have been a

number of previous studies investigating microbial diversity

in these environments (Abdulla, 2009; Frey et al., 2010;

Hutchens et al., 2010; Lapanje et al., 2012) with a particu-

larly large body of literature on lichen-associated

communities (Banfield, 1999; Chen et al., 2000; Villa et al.,

2015). However, exposed rock and mineral surfaces can

also harbour a broad diversity of non-lichen associated

bacteria, fungi and other eukaryotes that have received

less attention (Gorbushina and Broughton, 2009).

Exposed rock surfaces present unique challenges for

microbial growth and survival due to high levels of UV radi-

ation, frequent desiccation and limited resource availability.

Many rock dwelling microbes have evolved symbiotic strat-

egies to survive in this difficult environment, with lichens

being the best-studied example. Lichens are composite

organisms that represent a symbiotic relationship between

fungi and a photobiont (either algae or cyanobacteria),

although recent work has shown that a diversity of bacteria

and other eukaryotes also participate in this symbiosis

(Bates et al., 2011; Grube et al., 2015; Spribille et al.,

2016). Lichens are effective colonizers of rock surfaces:

they can mitigate the effects of solar radiation through UV

absorbing pigments, store water to fend off desiccation, fix

carbon and, in some cases, fix nitrogen (Banfield, 1999).

However, not all microbes found on rock surfaces are

members of lichen symbioses. Free-living heterotrophic

bacteria, fungi and non-lichenized algae and cyanobacteria

also frequently colonize rock surfaces where they can per-

sist in biofilms (Gorbushina, 2007). Additionally, microbes

can dwell within the rock itself as endoliths. By moving into

the pore space of rocks, endoliths are buffered from the

stressors of the rock surface and can exist in environments

Received 18 July, 2017; revised 25 October, 2017; accepted 6December, 2017. *For correspondence. E-mail [email protected]; Tel. 303-492-5615; Fax 303-492-1149.

VC 2017 Society for Applied Microbiology and John Wiley & Sons Ltd

Environmental Microbiology (2017) 00(00), 00–00 doi:10.1111/1462-2920.14024

where epilithic life is inhibited (Bell, 1993; Walker and

Pace, 2007).

Microbial colonization of rock surfaces can accelerate

mineral weathering by non-specific physical (e.g., contrac-

tion and expansion of fungal hyphae) and chemical

processes (e.g., secretion of acidic metabolites) (Gorbush-

ina, 2007; Uroz et al., 2015). Conversely, some microbes

are capable of more direct mineral dissolution. Shewanella

oneidensis, for instance, uses outer-membrane cyto-

chromes to reduce manganese and iron on the surface of

minerals, thereby gaining terminal electron acceptors for

anaerobic respiration and consequently accelerating min-

eral dissolution (Lower et al., 2001). Microbes can also

produce siderophores, chelating compounds capable of

binding iron and other metals with high affinity (Ahmed and

Holmstr€om, 2014). However, the rate and mechanisms of

microbial weathering are highly variable and dependent on

specific characteristics of the rock surface (Mapelli et al.,

2012).

Several factors are known to affect the assembly of rock-

associated microbial communities, including climate, rock

type and age of rock surface. While tropical climates are

more amenable to rapid microbial growth due to high humid-

ity and elevated temperatures (Dakal and Cameotra, 2012),

rock-inhabiting organisms also thrive in more arid environ-

ments (Gorbushina and Broughton, 2009). For example,

bacilli isolated from granite rock in Arizona were shown to be

similar to strains recovered from Antarctic soils (Fajardo-

Cavazos and Nicholson, 2006). Similarly, non-lichenized

fungi first isolated in hot and cold deserts have been shown

to also colonize rock surfaces in temperate regions (Gor-

bushina and Broughton, 2009). Several studies have found

that microbial communities also vary as a function of rock

type, although many of these studies were restricted to the

portion of communities that were readily cultivable (Abdulla,

2009; Frey et al., 2010; Lapanje et al., 2012). Lastly, the

length of time a rock surface has been exposed can also

influence the composition of rock-associated microbial com-

munities via successional processes. For example,

autotrophs are often the initial colonizers of rock surfaces

and their presence leads to elevated water retention and

organic carbon concentrations, setting the stage for other

taxa, including heterotrophs, to successfully colonize rock

surfaces over time (Lan et al., 2010).

Here, we surveyed the diversity of microbial communi-

ties associated with rock surfaces, including both lichen-

associated and lichen-independent bacteria, archaea and

microbial eukaryotes, using high-throughput marker gene

sequencing (targeting 16S and 18S rRNA genes). We also

used shotgun metagenomic sequencing to determine the

functional attributes of a subset of these microbial commu-

nities. Our objectives were to determine how the

composition and functional attributes of microbial commu-

nities living on rock surfaces vary as a function of rock

type, climate, exposure time and geographic location

across a broad sample set. We addressed these questions

by sampling the microbial communities found on 149 tomb-

stones from nine cemeteries located across three

continents. As demonstrated in previous work (Mushegian

et al., 2011; Villa et al., 2015), tombstones are well-suited

for investigating microbial colonization of rock surfaces.

Most tombstones are composed of only a handful of com-

mon rock types and are of similar general design, allowing

us to sample from replicate tombstones within each ceme-

tery location. Also, by sampling from tombstones that were

unlikely to have been cleaned, we can use the date on the

tombstone as a rough approximation of the amount of time

that tombstone surface has been exposed and available

for microbial colonization.

Results

Diversity of tombstone-associated communities

We used a network of volunteers to collect 149 tombstone

surface samples from nine cemeteries across three conti-

nents (Supporting Information Table 1). These locations

represented a broad range of climatic conditions, with

mean annual temperatures ranging from 48C to 248C and

mean annual precipitation ranging from 469 to 1268

mm y21. Nearly 90% of these tombstones were classified

as either limestone or granite, with installation dates rang-

ing from 1838 to 2011. Because we were interested in both

bacterial and eukaryotic communities and their interactions

on rock surfaces, we used two primer pairs, one that

yielded amplicons that were dominated (>95%) by bacte-

rial 16S rRNA gene reads (515f/906r, Klindworth et al.,

2013) and another set of eukaryote-specific 18S rRNA-

targeting primers (F1391/REukBr, Ramirez et al., 2014).

Although primer pair 515f/906r also targets archaeal

16S rRNA genes, archaea were rare (<0.001% of reads)

in the sequenced amplicon pools. The bacterial communi-

ties we found on tombstone surfaces were dominated by

five phyla: Proteobacteria (33% of all 16S rRNA gene

reads), Cyanobacteria (15%), Bacteroidetes (14%), Acti-

nobacteria (13%) and Acidobacteria (6%). We also

examined the bacterial communities at finer levels of taxo-

nomic resolution; in Fig. 1, we show both bacterial and

eukaryotic genera that had at least 1% average abun-

dance across all geographic locations. These genera

represent taxa that are common across rock surfaces in

many geographic locations, although there were few gen-

era shared across all samples. The dominant bacterial

genera were members of the five aforementioned phyla,

along with one genus within the candidate phylum FBP

and one genus from the Deinococcus-Thermus phylum.

Two main groups dominated the eukaryotic 18S sequence

libraries: Fungi (47%) and Chlorophyta (33%). While there

was considerable variation in the relative abundances of

2 T. E. Brewer and N. Fierer

VC 2017 Society for Applied Microbiology and John Wiley & Sons Ltd, Environmental Microbiology, 00, 00–00

the eukaryotic taxa, the most abundant and ubiquitous

eukaryotic genera were members of the algal class Tre-

bouxiophyceae and fungal orders Chaetothyriales and

Pleosporales (Fig. 1).

Next, we identified co-occurrence patterns of eukaryotic

and bacterial taxa that were common across samples to

determine which taxa (defined at the �97% sequence simi-

larity level) consistently appeared together (q> 0.60,

p< 0.01; Spearman correlation). Unsurprisingly, many of

the co-occurring taxa appeared to be lichen symbionts

(Supporting Information Figure S1). Our network correlation

included lichenized fungi within the order Teloschistales that

consistently co-occurred with algal phylotypes within the

class Trebouxiophyceae, known to contain lichenized photo-

bionts. Several bacterial taxa also co-occurred with the

lichenized fungi; these bacterial taxa were from the orders

Sphingomonadales, Deinococcales, Frankiales and the phy-

lum FBP.

Factors structuring the tombstone-associatedcommunities

We found that the overall composition of both the bacterial

and eukaryotic communities significantly differed depend-

ing on cemetery location (bacterial R2 5 0.34, p<0.001;

eukaryotic R2 5 0.29, p< 0.001; PERMANOVA). Ordina-

tion plots for both bacterial and eukaryotic communities

highlight that regions with similar climates tend to harbour

similar rock-associated communities – tropical and sub-

tropical regions (Florida, Colombia) were clearly distinct

from more temperate regions (Barcelona, Belgium, Colo-

rado, Denmark and Maine, Fig. 2). Indeed, mean annual

precipitation and temperature together were significant in

structuring both the bacterial and eukaryotic communities

(bacterial r 5 0.18, p<0.001; eukaryotic r 5 0.34,

p<0.001; Mantel test). Although we tested the effect of a

number of other variables on the composition of the micro-

bial communities, including tombstone age, cardinal

direction of stone face and surface texture (polished versus

rough), all were found to have no significant effect on the

overall composition of either the eukaryotic or bacterial

communities.

We did find that tombstone rock type had significant

effects on the composition of both the bacterial and

eukaryotic communities (bacterial R2 5 0.11, p< 0.001;

eukaryotic R2 5 0.075, p< 0.001; PERMANOVA). How-

ever, because several cemeteries had tombstones of only

a single rock type, it was difficult to disentangle the effects

of rock type and geographic location across our entire

sample set. For example, all of the tombstones sampled in

Colombia were limestone, while the vast majority of tomb-

stones sampled in Spain were granite (Supporting

Information Table S1). To specifically investigate the effects

of rock type on bacterial and eukaryotic communities, we

Fig. 1. The most abundant taxain the bacterial and eukaryoticdatasets, by medianabundance, across all sampletypes.

Only taxa that could be

classified to class level and had

at least 1% average abundance

across the entire dataset are

shown. The number within each

box represents the median

abundance of the specified

genus in the specified location.

When taxa could not be

classified to the genus level,

the genus is labelled as

‘Unclassified’. Sites are ordered

by increasing mean annual

temperature. Genera with low

abundance are coloured dark

blue, higher abundance in

yellow and the highest two

values in pink.

Tales from the tomb 3


focused on two cemeteries where granite and limestone

tombstones were co-located. This subset of our data

included 40 samples in total: 20 from cemeteries in Bel-

gium and 20 from a cemetery in Maine, United States.

Samples at both sites were roughly half limestone and half

granite, with age ranges of 1855–1980 for the Maine sam-

ples and 1914–1991 for the Belgium samples. When the

overall composition of the microbial communities from

these two sites were compared, rock type explained more

of the variation in community composition than geographic

location for both bacteria and eukaryotes (Fig. 3; bacterial:

rock type: p< 0.002, R2 5 0.31; location: p <0.046,

R2 5 0.07; eukaryotic: rock type p 5 0.001, R2 5 0.16;

location p 5 0.001, R2 5 0.09; all PERMANOVA). In other

words, the microbial communities on limestone tomb-

stones in Maine were more similar to limestone

tombstones in Belgium than they were to granite tomb-

stones located just several meters away.

We used the Belgian and Maine subset of our data to

identify microbes that consistently preferred either lime-

stone or granite as a substrate, regardless of geographic

location. We identified 25 differentially abundant

bacterial taxa (11 for granite and 14 for limestone, FDR

corrected Mann-Whitney test, 0.01 p-value cut off) and

16 significant eukaryotic taxa (8 for each rock type, FDR

corrected Mann-Whitney test, 0.05 p-value cut off). The

top rock-specific genera (by degree of significance) are

shown for both the bacterial and eukaryotic datasets in

Fig. 4.

The eukaryotic taxa that were enriched on granite surfa-

ces included the acid-tolerant free-living algae

Apatococcus (Gustavs et al., 2016; Low-D�ecarie et al.,

2016), the melanized, microcolonial rock-inhabiting fungi

Knufia (Noack-Sch€onmann et al., 2014) and Catenulos-

troma (Ruibal et al., 2008). The limestone-enriched

eukaryotes included several taxa associated with lichens;

the fungal order Teloschistales are almost entirely liche-

nized, while the algae Heveochlorella are known lichen

photobionts (Sanders et al., 2016). The remaining

limestone-specializing eukaryotic orders (Capnodiales,

Chaetothyriales and Xylariales) encompass both micro-

colonial rock dwelling and lichenized fungi.

Many of the bacterial taxa that were significantly more

abundant on limestone tombstones were taxa that are

Fig. 2. A PCoA plot of Bray-Curtis dissimilarities highlightingthat both bacterial (p< 0.001,R2 5 0.34, PERMANOVA) andeukaryotic (p< 0.001,R2 5 0.29, PERMANOVA)communities were significantlydifferent in compositiondepending on samplinglocation.

Communities from tropical and

subtropical climates (Colombia

and Florida) cluster apart from

those found in more temperate

locations. The polygons are

drawn to encompass samples

from each location.

Fig. 3. A PCoA plot of Bray-Curtis dissimilarities highlightingthat rock type is more importantthan location in describing bothbacterial (p< 0.001, R2 5 0.31vs. p< 0.046, R2 5 0.07PERMANOVA) and eukaryotic(p 5 0.001, R2 5 0.16 vs.p< 0.001, R2 5 0.09PERMANOVA) communitiesfrom Belgium and Maine.

Rock type appears to have a

stronger effect on bacterial than

eukaryotic communities. The

polygons are drawn to

encompass samples from each

rock type.



known to include radiation-resistant strains (e.g., Truepera-

ceae, Chroococcidiopsis, Spirosoma, Rubellimicrobium;

Billi et al., 2000; Albuquerque et al., 2005; Favet et al.,

2013; Lee et al., 2014;). We also noted that the bacterial

taxa enriched on the granite tombstones tended to include

known acid tolerant groups (e.g., Acidobacteriaceae,

Beijerinciaceae and Methylocystaceae), while bacteria

that were more abundant on limestone tended be from

groups that are considered neutrophilic or alkaliphilic (e.g.,

Spirosoma, Rubellimicrobium and Truepera). Thus, we

hypothesized that differences in bacterial pH preferences

are a major factor driving the differential abundances of

bacterial taxa across the limestone and granite tomb-

stones. To test this hypothesis, we searched the literature

for pH growth ranges of all characterized isolates from bac-

terial families that were significantly enriched on one rock

type. Although the pH growth range data were not for the

exact strains we found on the tombstones, pH tolerance is

a trait conserved at a coarse level of taxonomic resolution

and thus should be reasonably consistent within bacterial

families (Martiny et al., 2015). We found pH growth range

data for 39 and 32 isolates representing bacterial families

Fig. 4. Taxa from the eukaryotic andbacterial datasets that weresignificantly enriched on granite orlimestone in the Maine and Belgiumtombstone samples (bacterialp� 0.001; eukaryotic p< 0.05; Mann-Whitney test, FDR-corrected).

Bacterial isolates from the granite-

enriched genera were generally acid-

tolerant, while the limestone-enriched

bacterial genera included neutrophilic

and radiation resistant isolates. The

eukaryotic limestone genera included

known lichen affiliates (Capnodiales,

Heveochlorella, Teloschistales).



found to be more abundant on granite and limestone tomb-

stones respectively. These bacterial families accounted for

>40% of bacterial 16S rRNA gene sequences on their

respective tombstone rock types. As expected, the cul-

tured representatives from bacterial families that were

more abundant on granite tombstones had a significantly

lower pH threshold for growth than their limestone counter-

parts (Supporting Information Figure S2).

Because we observed strong effects of rock type on the

taxonomic composition of the microbial communities, we

also characterized the functional attributes of the microbial

communities found on granite and limestone surfaces. To

do this, we used shotgun metagenomic sequencing of four

limestone samples and four granite samples from Maine,

United States. As an initial comparison, we used Metaxa2

(Bengtsson-Palme et al., 2015) to confirm that similar taxa

were observed in both our amplicon and shotgun metage-

nomic datasets. The same five bacterial families

dominated both the amplicon and metagenomic datasets-

Methylocystaceae, Xenococcaceae, Acetobacteraceae,

Sphingomonadaceae and Cytophagaceae. Likewise, both

eukaryotic datasets reported high abundances of fungal

classes Dothideomycetes, Eurotiomycetes and Lecanoro-

mycetes, as well as the algal family Trebouxiophyceae.

The relative abundance of bacteria versus eukaryotes was

highly variable across samples; eukaryotic sequences

accounted for 38%–72% of all small-subunit RNA gene

sequences recovered from the shotgun metagenomic

datasets, although these results are difficult to interpret

due to high variation in rRNA gene copy numbers across

eukaryotic taxa (Black et al., 2013). Archaeal SSU rRNA

made up only 0.005% of all the 16S rRNA reads recovered

from the shotgun metagenomic data, supporting the near

complete absence of archaea in our 16S rRNA amplicon

sequencing dataset.

We next used the portion of the metagenomic reads that

could be annotated with KEGG (approximately 12% of

reads on average, Kanehisa et al., 2012) to compare the

potential functional attributes of tombstone communities

living on granite versus limestone tombstones. The low

proportion of annotated reads stems from the few

genomes available from environmental, uncultured

microbes and the conservative nature of the KEGG data-

base which primarily includes well-studied genes. We split

these annotated reads into those assigned as bacterial or

eukaryotic genes to avoid biases caused by differences in

overall bacteria: eukaryote ratios. We also normalized

these read abundances to account for differences in mean

genome sizes within each sample (Nayfach and Pollard,

2015). We found that rock type was a significant factor

explaining differences in overall functional attributes for

bacterial, but not eukaryotic communities (bacterial

p< 0.05, R2 5 0.34; eukaryotic p> 0.05; PERMANOVA;

Supporting Information Figure S3). Next, we identified

KEGG gene pathways that were differentially abundant

between the two tombstone types in the bacterial commu-

nities. There were 10 and 14 pathways that were more

abundant within limestone-associated and granite-

associated bacterial communities respectively (FDR-cor-

rected p< 0.1; Mann-Whitney Test; Supporting Information

Table S3). For example, granite surfaces had more genes

related to acid resistance pathways (sesquiterpenoid triter-

penoids biosynthesis), cell movement (flagellar assembly),

exotic compound degradation (benzoate degradation) and

amino acid metabolism (cysteine and methionine, phenyl-

alanine and tryptophan, Fig. 5). In contrast, limestone

surfaces had significantly more genes associated with pho-

tosynthesis, UV resistance (carotenoid synthesis) and

vitamin and cofactor synthesis pathways (ubiquinone,

folate, pantothenate and thiamine biosynthesis, Fig. 5).

Discussion

Diversity of rock associated communities

Most of the bacterial groups we found to be common on

tombstone surfaces (Fig. 1) have been observed previ-

ously in comparable environments such as stone

monuments. Bacteria within the genera Sphingomonas,

Hymenobacter, Spirosoma, Truepera, Pseudonocardia,

the family Acetobacteraceae and the phyla Acidobacteria

and FBP were found on the surface of stone monuments in

China (Li et al., 2016). Several of these bacterial groups have

also been identified from dust samples (Sphingomonas,

Hymenobacter, Pseudonocardiaceae, Sporichthyaceae,

Acetobacteraceae), which may explain their global ubiq-

uity on rock surfaces (Favet et al., 2013). More generally,

the bacteria we identified as common rock-dwellers

(Fig. 1) have been shown to possess a number of traits

that could help them to survive on exposed rock surfaces,

from resistance to radiation (Hymenobacter, Spirosoma,

Truepera and phylum Actinobacteria: Albuquerque et al.,

2005; Warnecke et al., 2005; Dartnell et al., 2010; Lee

et al., 2014; Ahn et al., 2016), to production of viscous

exopolysaccharides (polymers known to moderate tem-

perature extremes and enhance water retention:

Sphingomonas, Nwodo et al., 2012), to associations with

lichen (Acetobacteraceae and Sphingomonodaceae,

Bates et al., 2011).

Likewise, the eukaryotic taxa we found to be common

on tombstone surfaces have previously been observed on

rock surfaces. Similar to other studies of rock colonizing

fungi, the fungi we found were all from the phylum Asco-

mycota (Li et al., 2016). Of these taxa, two orders

(Chaetothyriales and Capnodiales) include members

known as black melanized micro-colonial fungi (MCF),

which live within subaerial biofilms on rock and have been

found in desert varnish (Gorbushina, 2007; Northup et al.,

2010). MCF can endure long periods of dormancy to



persist under nutrient poor conditions and have dark pig-

mentation that imparts resistance to UV radiation.

However, members of the Chaetothyriales and Capno-

diales groups also include lichenized taxa (Banfield, 1999).

The green algae Trebouxiophyceae was also common on

rock surfaces regardless of geographic location. Like the

fungal orders Chaetothyriales and Capnodiales, these

algae can be components of subaerial biofilms on rock sur-

faces (Gorbushina, 2007; V�azquez-Nion et al., 2016), or

can live as lichen photobionts (Muggia et al., 2013). Mem-

bers of Trebouxiophyceae are often radiation and

desiccation-resistant (Rivasseau et al., 2016), traits that

likely enable effective colonization of rock surfaces.

Given the abundance of lichenized fungi and green

algae, it is not surprising that our co-occurrence analyses

were able to identify strong associations between potential

lichen partners (Supporting Information Figure S1). The

co-occurrence network we assembled contains fungi from

the lichenized order Teloschistales, as well as the order

Capnodiales, which includes lichenized taxa among its

diverse constituents. The algal taxa in our network were

from the class Trebouxiophyceae, whose members are

capable of a free-living or photobiont lifestyle (Muggia

et al., 2013). Several of the bacteria which exhibited strong

co-occurrence patterns with algae or fungi have previously

been observed to associate with lichens (including taxa

from the Sphingomonadaceae and Truperaceae families;

Albuquerque et al., 2005; Bates et al., 2011). However,

many of the bacterial taxa shown in Supporting Information

Figure S1 have not been identified in association with

lichens (including members of the order Frankiales and the

FBP candidate phylum). Additional work is required to

determine if these bacteria consistently associate with

lichens and what role, if any, they might play in lichen

symbioses.

We found that the composition of the bacterial and

eukaryotic communities on tombstone surfaces was vari-

able across the sampled cemeteries. This geographic

variation is likely a product of differences in climatic condi-

tions, as the development of both lichenized and non-

lichenized rock-associated microbial communities is

expected to be sensitive to differences in temperature and

water availability (Macedo et al., 2009). Interestingly,

besides geographic location, climate and rock type, no

other measured factors were significantly correlated with

variation in the rock-associated microbial community com-

position. Even tombstone age (how long a tombstone has

been in a cemetery) was not a significant predictor of

microbial community composition. Colonization of terres-

trial surfaces has been found to follow predictable trends;

free-living algae, bacteria and microcolonial fungi are gen-

erally responsible for primary colonization, followed by

mosses and lichen, then plants and small animals (Gor-

bushina and Broughton, 2009). However, mature

colonization of rock surfaces can occur in as little as 15

years (Pohl and Schneider, 2002). A different sampling

Fig. 5. KO pathways enriched in bacterial communities on either limestone or granite tombstones in Maine (FDR-corrected p< 0.1; Mann-Whitney Test).

Granite had significantly more genes related to substrate transport (ABC transporters, bacterial secretion), acid resistance (sesquiterpenoid/

triterpenoids), cell motility (flagellar assembly) and exotic compound degradation (benzoate), while limestone had more genes related to

photosynthesis and UV resistance (carotenoids). Granite communities also had higher relative abundances of genes encoding the synthesis of

metabolically expensive amino acids (cysteine, methionine, phenylalanine and tryptophan) while limestone communities were enriched in

vitamin and cofactor synthesis pathways (ubiquinone, folate, thiamine and pantothenate). Error bars indicate range of samples, while the

proportional abundances of KO categories were calculated after dividing by genome equivalents in each sample.



strategy focusing more explicitly on tombstones that vary

in age within a single cemetery is likely necessary to detect

these temporal patterns in microbial community succes-

sion on rock surfaces.

Effects of rock type on tombstone microbial communities

Rock type had a significant impact on the composition of

both the bacterial and eukaryotic communities found on

tombstone surfaces, with specific taxa consistently

enriched on either granite or limestone tombstones (Fig.

4). This finding is in line with previous work (Mauck and

Roberts, 2007; Hutchens et al., 2010), although it remains

unclear whether the rock type effects are driven more by

differences in the physical characteristics (e.g., porosity,

surface roughness) or chemical characteristics (e.g.,

chemical composition, pH) of the rock surfaces (Porca

et al., 2012).

In addition to the taxonomic shifts in the microbial com-

munities on the two rock types (Fig. 4), we also identified

specific gene categories and phenotypic traits (Fig. 5 and

Supporting Information Figure S2) that were consistently

enriched on either granite or limestone tombstones across

the subset of samples analysed. Granite surfaces, for

example, were enriched in acidophilic bacteria and gene

pathways linked to both acid resistance (sequiterpenoid/tri-

terpenoids synthesis, Schmerk et al., 2011; Chen et al.,

2015; Wu et al., 2014) and acid production (glyoxylate/

dicarboxylate metabolism). Granite communities were also

enriched in genes related to cell movement (flagellar

assembly), and substrate transport (ABC transporters,

bacterial secretion systems), suggesting that bacteria on

granite surfaces are more likely to be motile, better able to

escape growth-limiting local conditions, and more likely to

be free-living (Alexandre, 2015) than their limestone coun-

terparts. Granite communities also had more genes linked

to metabolically expensive amino acid synthesis (trypto-

phan, phenylalanine and methionine are some of the most

energy intensive amino acids to synthesize, Akashi and

Gojobori, 2002). Auxotrophic taxa reap an energy savings

so long as the amino acid they cannot produce is available

in their environment. The more metabolically expensive an

amino acid, the larger the energy savings. Free-living bac-

teria are predicted to have fewer auxotrophies than

symbiotic taxa (D’Souza et al., 2014), which may explain

why granite bacterial communities are likely more capable

of synthesizing metabolically-expensive amino acids.

Limestone surfaces, in contrast, harboured more radiation

resistant bacteria with neutral to alkaline pH growth ranges,

along with many potentially lichenized microbes. These com-

munities were enriched in carbon-fixing pathways (possibly

derived from the photobiont Chroococcidiopsis, Fig. 4) and

mechanisms of UV resistance (carotenoid synthesis).

Limestone bacterial communities were also enriched in

genes related to vitamin and cofactor synthesis (ubiqui-

none, folate, thiamine and pantothenate synthesis).

Bacteria associated with the lichen symbiosis have been

suggested to synthesize vitamins and cofactors for other

members of the symbiosis (Grube et al., 2015), which

may explain the enrichment of these gene pathways.

Microbial communities and mechanismsof rock weathering

The microbes and functional pathways we identified on

tombstone surfaces may give us some insight into the

impacts of these communities on rock weathering.

For example, one of the major routes to stone deteriora-

tion is biocorrosion, caused by the routine secretion of

acidic metabolites by microbes (Uroz et al., 2015). Signifi-

cant drops in pH can occur over time when microbes

colonize inert, unbuffered media (Liermann et al., 2000).

Specifically, strains of bacteria isolated from granite rocks

have been shown to acidify unbuffered media by secretion

of oxalic acid, leading to dissolution of granite particles

(Frey et al., 2010). Interestingly, a metabolic pathway that

results in oxalic acid byproducts was significantly enriched

in bacterial communities on granite tombstones (glyoxylate

and dicarboxylate metabolism, Supporting Information

Table S3). Pitting of stone surfaces caused by biocorrosion

may create microenvironments capable of shielding

microbes from some environmental stresses, such as UV

radiation. The microbial communities living on granite sur-

faces likely need to tolerate acidic conditions, but may gain

some protection against UV radiation, decreasing the rela-

tive importance of radiation resistance genes (Snider

et al., 2009).

Limestone, on the other hand, is rich in carbonate com-

pounds that can buffer acidic metabolic products routinely

secreted during microbial growth (Warscheid and Braams,

2000). A buffered rock surface would be more resistant to

acidic pitting and have a higher surface pH than an unbuf-

fered rock such as granite, an explanation supported by

the relatively high pH growth ranges of bacteria found on

limestone (Fig. 4 and Supporting Information Figure S2).

Additionally, limestone had a higher proportion of cyano-

bacteria and photosynthesis-linked genes than granite.

Photosynthesis in cyanobacteria raises local pH through

CO2 concentrating mechanisms (B€udel et al., 2004), and

may further elevate the pH of limestone surfaces. If lime-

stone surfaces are less vulnerable to biocorrosion than

granite, there is likely to be less UV protection for those

microbes living in microsites, giving those microbes with

radiation resistance a selective advantage. Limestone sur-

faces were also enriched in lichen-associated taxa. Lichen

are potent weathering agents, and while acidic secretions

may be buffered by limestone chemistry, mechanical

stress from the contraction and expansion of fungal



filaments through wetting and drying cycles can also

weather limestone (Burford et al., 2003).

Conclusions

By sampling 149 tombstones across three continents and

analysing their bacterial and eukaryotic communities via

marker gene sequencing, we were able to demonstrate

that microbes living on rock surfaces have specific adap-

tions to this environment, and vary based on geographic

location and climate. Moreover, we found that rock type

had a strong influence on these microbes, effects that

were also evident when we used shotgun metagenomic

analyses to identify gene categories that were differentially

abundant across two rock types. Further work is needed to

identify the specific factors contributing to the observed

rock type effects. However, we found that different rock

types consistently feature different microbial communities,

which in turn, may alter the biotic processes associated

with weathering, even though the tombstones within a

given cemetery are subjected to the same climatic

conditions.

Experimental procedures

Sample collection and processing

To investigate microbial communities associated with rock sur-

faces, we recruited volunteers to sample tombstone surfaces.

Between December 2014 and March 2015, volunteers wear-

ing sterile nitrile gloves used individually sealed nylon bristle

toothbrushes (packaged by the manufacturer) to brush the

surface of tombstones across a standardized 100 cm2 area in

a non-destructive manner (Supporting Information Figure S4).

Volunteers were asked to sample tombstones that did not

appear to have been recently cleaned. Several characteristics

of each tombstone were recorded: location of the tombstone,

rock type, length of tombstone exposure (date of death), direc-

tion of the sampled face (cardinal direction) and surface

texture (polished/unpolished). Volunteers took pictures of the

front and back of tombstones, which we used to verify tomb-

stone rock type.

Due to the sensitive nature of the sample type, we did not

perform any destructive mineralogical analyses. While we

acknowledge that chemical structure and inclusions can vary

within rock types, we make the assumption that the mineralo-

gies of different types of granite (for example) are more similar

to each other than to the other rock types sampled (limestone,

marble, sandstone and slate). Likewise, we assumed the

effects of any microscopic inclusions or other sources of

micro-heterogeneity in rock composition would be minimized

by sampling from a comparatively large surface area on each

tombstone.

After sampling, toothbrushes were placed in sterile bags

and shipped to the University of Colorado where they were

stored at 2208C upon receipt. The range of time between

sampling and storage at 2208C never exceeded two weeks.

We received a total of 149 samples from cemeteries in

Barcelona (Spain), Dagua (Colombia), Jacksonville (FL,USA), Falmouth (ME, USA), Nederland (CO, USA), Hellerup

(Denmark) and Antwerp (Belgium). Two separate cemeterieswere sampled in Hellerup and Antwerp while single cemeter-ies were sampled everywhere else. Sites were selected by

availability of volunteers and existence of appropriate sam-pling locations. Climate data for all sites were retrieved fromclimate-data.org. Further details on the sampling locations are

provided in Supporting Information Table S1.

We extracted DNA from the toothbrush samples within 1–2months of receipt by cutting half the bristles off each tooth-brush with flame-sterilized scissors and placing them into a

single well of a MoBio PowerSoil-htp 96 well Soil DNA Isola-tion Kit (MoBio, Carlsbad, CA). This work was done in a sterilehood and we extracted one aliquot of DNA per sample. Some

wells were left empty to test for potential contamination.Because the toothbrushes were packaged by the manufac-turer and not necessarily sterile, we also included controls

with ‘blank’ toothbrushes to check for contamination. DNAextraction was conducted following the manufacturer’s instruc-tions (MoBio), with an additional incubation step at 658C for 10

min prior to bead beating.

Amplicon-based analyses

We used two primer pair sets for our amplicon-based studies:515f/907r for 16S rRNA gene sequencing and F1391/REukBrfor 18S rRNA gene sequencing. The 515f/907r primer pair

[515F (50- GTGYCAGCMGCCGCGGTAA 230) and 907R (50-CCGTCAATTCMTTTRAGTTT 230)] targets the V4-V5 regionof the 16S rRNA gene (Klindworth et al., 2013). PCR condi-

tions were: 948C for 3min; 35 cycles of 948C for 45 s, 508C for60 s, 728C for 90 s; 728C for 5 min. This primer set was initiallypredicted to cover all SSU rRNA and be truly universaldomain. However, in practice, the sequences returned from

this reaction were almost entirely from bacteria (>95%).Therefore, we used another primer pair to generate eukaryoticsequences. The F1391/REukBr primer pair [F1391 (50- GTAC

ACCGCCCGTC -30) and REukBr (50 -TGATCCTTCTGCAGGTTCACCTAC -30)] targets the V9 hyper-variable region ofthe eukaryotic 18S rRNA gene (Ramirez et al., 2014). PCR

conditions for this primer pair were: 948C for 3 min; 35 cyclesof 948C for 45 s, 578C for 60 s, 728C for 90 s; 728C for 10 min.All primers were ‘fusion’ primers that included 12 bp barcodes

unique to each primer pair. PCR was performed in triplicateand PCR reactions for all samples were cleaned and normal-ized to 25 ng using 96 well SequalPrep Normalization Plate

Kits (Thermo Fisher Scientific, Waltham, MA). All sampleswere pooled and sequenced on an Illumina MiSeq (2x150paired end sequencing) at the University of Colorado Next

Generation Sequencing Facility.

Sequences were processed as described previously (Emer-son et al., 2015). We used USEARCH7 (Edgar, 2010) toquality filter (-fastq_filter with fastq_maxee 1), dereplicate

reads (-derep_fulllength) and remove singletons (-sortbysize).Neither of the amplicon datasets could be merged due toinsufficient length, so we used only the forward reads in down-

stream analyses. Sequences were assembled into phylotypesat the �97% identity level using UCLUST (-cluster_otus)(Edgar, 2010). We then mapped the raw sequences to the de

novo clustered phylotypes using USEARCH7 command -



usearch_global. Taxonomy was assigned using the QIIME

v.1.9.1 (Caporaso et al., 2012) script assign_taxonomy.py with

the Ribosomal Database Project classifier (Cole et al., 2014)

and either the Greengenes 13_8 database (515/907 primer

set, DeSantis et al., 2006) or the Silva119 database (F1391/

REukBr primer set, Quast et al., 2013). Mitochondrial sequen-

ces were removed (filter_taxa_from_otu_table.py, QIIME) and

both datasets were rarefied to 2300 and 11 000 randomly

selected sequences per sample for the bacterial and eukary-

otic analyses, respectively, using the QIIME command

single_rarefaction.py.

All statistical analyses were conducted in R. Co-occurrence

patterns were determined as described previously (Barber�an

et al., 2012). An upper limit of 0.6 was set for rcorr values and

0.01 for FDR-corrected p-values (generated by Spearman cor-

relations; rcorr command from the Hmic R package - Harrell,

2017). The R package igraph (Cs�ardi and Nepusz, 2006) was

used to generate a .graphml file from these correlations which

was explored and visualized with Gephi version 3. Only phylo-

types that occurred in at least 50% of samples were included

in these analyses as we wanted to focus on the more ubiqui-

tous co-occurrences that were common on rock surfaces

regardless of location.

R packages vegan (Dixon, 2003) and plyr (Wickham, 2011)

were used to perform ecological analyses. Phylotype abun-

dance tables were Hellinger transformed, Bray-Curtis

dissimilarity matrices were generated and PCoA (Principal

Coordinates Analysis) plots were used to visualize differences

in community composition across sample categories. PER-

MANOVA was used to test the importance of the site and

tombstone characteristics on the overall taxonomic structure

of the bacterial and eukaryotic communities. To identify taxa

that specialized on one rock type in paired conditions (lime-

stone versus granite tombstones that were co-located), the

Mann-Whitney test was used with FDR corrected p-values. All

other plots were created using the R package ggplot2 (Wick-

ham, 2009).

Shotgun metagenomic analyses

Eight samples from the same site in Maine, United States

were selected for metagenomic sequencing: four granite

tombstones and four limestone tombstones. These stones

were all within similar age ranges – the youngest was erected

in 1947, the oldest in 1910. DNA was extracted as described

above and metagenomic libraries were generated using the

TruSeq DNA LT library preparation kit (Illumina, San Diego,

CA) according to manufacturer instructions. All samples were

pooled and sequenced on an Illumina MiSeq (2x150 paired

end sequencing) at the University of Colorado Next Genera-

tion Sequencing Facility.

After merging the paired-end metagenomic reads (-fastq_-

mergepairs) and quality filtering (-fastq_filter with fastq_maxee

0.5) with USEARCH, we obtained an average of �650 000

merged sequences per sample. We used Metaxa2 (Bengts-

son-Palme et al., 2015) with default settings to identify taxa

found in the metagenomic libraries from rRNA gene analyses

using only the sequences that successfully merged. We

uploaded the unassembled, merged sequences to MG-RAST

(Meyer et al., 2008) for annotation and used KEGG (Kanehisa

et al., 2012) annotations for all further analyses. Using the

domain assignment output of MG-RAST, we removed sequen-

ces conclusively identified as eukaryotic for the functional

gene analyses. Once eukaryotic sequences were removed,

we ran MicrobeCensus (Nayfach and Pollard, 2015) and nor-

malized our KO matrices by dividing by the genome

equivalents in each sample, then converted each sample to

proportional data. Genome equivalents were determined

using MicrobeCensus with default settings (Nayfach and Pol-

lard, 2015). On average, only 12% of sequences could be

annotated using cutoffs for e-value of e25 and percent identity

of 60%. We used KO matrices of genes identified as bacterial

or eukaryotic by MG-RAST to create dissimilarity matrices

and PCoA plots. As the eukaryotic community did not vary sig-

nificantly based on rock type (detailed in Results and

Supporting Information Figure S3), we focused solely on bac-

terial reads for all further analyses. We next used FDR-

corrected Mann-Whitney tests to determine which KEGG

pathways were significantly enriched on one rock type in bac-

terial communities, considering only the 75 most abundant

pathways.

Data availability

Annotated and raw metagenomic reads have been deposited

and made publicly available through MG-RAST (project id:

mgp15093). In addition, phylotype tables and gene annotation

tables are available on Figshare (https://doi.org/10.6084/m9.

figshare.4604371.v1).

Acknowledgements

We thank the volunteers who collected the samples: Wenke

Smets, Serena Moretti, Leen Van Ham, Albert Barber�an,

Anne Madden, Erin Tripp, Karsten Elmose Vad, James

Brewer, Gwynn Brewer and Kara Gallardo. Funding to support

this work was provided by grants from the National Science

Foundation to N.F. (DEB0953331, EAR1331828).

References

Abdulla, H. (2009) Bioweathering and biotransformation of

granitic rock minerals by Actinomycetes. Microb Ecol 58:

753–761.

Ahmed, E., and Holmstr€om, S.J.M. (2014) Siderophores in

environmental research: roles and applications. Microb Bio-

technol 7: 196–208.Ahn, D.-H., Han, S.-R., Oh, T.-J., and Park, H. (2016) Com-

plete genome sequence of ionizing radiation-resistant

Hymenobacter sp. strain PAMC26628 isolated from an Arc-

tic lichen. J Biotechnol 223: 50–51.Akashi, H., and Gojobori, T. (2002) Metabolic efficiency and

amino acid composition in the proteomes of Escherichia

coli and Bacillus subtilis. Proc Natl Acad Sci USA 99: 1.Albuquerque, L., Sim~oes, C., Nobre, M.F., Pino, N.M.,

Battista, J.R., Silva, M.T., et al. (2005) Truepera radiovictrix

gen. nov., sp. nov., a new radiation resistant species and

the proposal of Trueperaceae fam. nov. FEMS Microbiol

Lett 247: 161–169.

Alexandre, G. (2015) Chemotaxis control of transient cell

aggregation. J Bacteriol 197: 3230–3237.



https://doi.org/10.6084/m9.figshare.4604371.v1

https://doi.org/10.6084/m9.figshare.4604371.v1

Banfield, J.F. (1999) Biological impact on mineral dissolution:

application of the lichen model to understanding mineral

weathering in the rhizosphere. Proc Natl Acad Sci USA 96:

3404–3411.Barber�an, A., Bates, S.T., Casamayor, E.O., and Fierer, N.

(2012) Using network analysis to explore co-occurrence

patterns in soil microbial communities. ISME J 6: 343–351.Bates, S.T., Cropsey, G.W.G., Caporaso, J.G., Knight, R., and

Fierer, N. (2011) Bacterial communities associated with the

lichen symbiosis. Appl Environ Microbiol 77: 1309–1314.Bell, R.A. (1993) Cryptoendolithic algae of hot semiarid lands

and deserts. J Phycol 29: 133–139.Bengtsson-Palme, J., Hartmann, M., Eriksson, K.M., Pal, C.,

Thorell, K., Larsson, D.G.J., et al. (2015) metaxa2:

improved identification and taxonomic classification of small

and large subunit rRNA in metagenomic data. Mol Ecol

Resour 15: 1403–1414.Billi, D., Friedmann, E.I., Hofer, K.G., Caiola, M.G., and

Ocampo-Friedmann, R. (2000) Ionizing-radiation resistance

in the desiccation-tolerant cyanobacterium. Appl Environ

Microbiol 66: 1489–1492.

Black, J., Dean, T., Byfield, G., Foarde, K., and Menetrez, M.

(2013) Determining fungi rRNA copy number by PCR.

J Biomol Tech 24: 32–38.

B€udel, B., Weber, B., K€uhl, M., Pfanz, H., S€ultemeyer, D., and

Wessels, D. (2004) Reshaping of sandstone surfaces by

cryptoendolithic cyanobacteria: bioalkalization causes

chemical weathering in arid landscapes. Geobiology 2:

261–268.Burford, E.P., Fomina, M., and Gadd, G.M. (2003) Fungal

involvement in bioweathering and biotransformation of

rocks and minerals. Mineral Mag 67: 1127–1155.Caporaso, J.G., Lauber, C.L., Walters, W.A., Berg-Lyons, D.,

Huntley, J., Fierer, N., et al. (2012) Ultra-high-throughput

microbial community analysis on the Illumina HiSeq and

MiSeq platforms. ISME J 6: 1621–1624.Chen, J., Blume, H.P., and Beyer, L. (2000) Weathering of

rocks induced by lichen colonization – a review. Catena 39:

121–146.Chen, L.-X., Hu, M., Huang, L.-N., Hua, Z.-S., Kuang, J.-L., Li,

S.-J., et al. (2015) Comparative metagenomic and meta-

transcriptomic analyses of microbial communities in acid

mine drainage. ISME J 9: 1579–1592.Cole, J.R., Wang, Q., Fish, J.A., Chai, B., McGarrell, D.M.,

Sun, Y., et al. (2014) Ribosomal Database Project: data and

tools for high throughput rRNA analysis. Nucleic Acids Res

42: D633–D642.Cs�ardi, G., and Nepusz, T. (2006) The igraph software pack-

age for complex network research. Complex Systems 1695.Dakal, T.C., and Cameotra, S.S. (2012) Microbially

induced deterioration of architectural heritages: routes and

mechanisms involved. Environmental Sciences Europe 24:

1–13.Dartnell, L.R., Hunter, S.J., Lovell, K.V., Coates, A.J., and

Ward, J.M. (2010) Low-temperature ionizing radiation resis-

tance of Deinococcus radiodurans and Antarctic dry valley

bacteria. Astrobiology 10: 717–732.DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie,

E.L., Keller, K., et al. (2006) Greengenes, a chimera-

checked 16S rRNA gene database and workbench compat-

ible with ARB. Appl Environ Microbiol 72: 5069–5072.

Dixon, P. (2003) VEGAN, a package of R functions for com-

munity ecology. J Veg Sci 14: 927–930.D’Souza, G., Waschina, S., Pande, S., Bohl, K., Kaleta, C.,

and Kost, C. (2014) Less is more: selective advantages can

explain the prevalent loss of biosynthetic genes in bacteria.

Evolution 68: 2559–2570.

Edgar, R.C. (2010) Search and clustering orders of magnitude

faster than BLAST. Bioinformatics 26: 2460–2461.Emerson, J.B., Keady, P.B., Brewer, T.E., Clements, N.,

Morgan, E.E., Awerbuch, J., et al. (2015) Impacts of flood

damage on airborne bacteria and fungi in homes after the

2013 Colorado Front Range flood. Environ Sci Technol 49:

2675–2684.Fajardo-Cavazos, P., and Nicholson, W. (2006) Bacillus endo-

spores isolated from granite: close molecular relationships

to globally distributed Bacillus spp. from endolithic and

extreme environments. Appl Environ Microbiol 72: 2856–

2863.Favet, J., Lapanje, A., Giongo, A., Kennedy, S., Aung, Y.-Y.,

Cattaneo, A., et al. (2013) Microbial hitchhikers on intercon-

tinental dust: catching a lift in Chad. ISME J 7: 850–867.Frey, B., Rieder, S.R., Brunner, I., Plotze, M., Koetzsch, S.,

Lapanje, A., et al. (2010) Weathering-associated bacteria

from the Damma glacier forefield: physiological capabilities

and impact on granite dissolution. Appl Environ Microbiol

76: 4788–4796.Gorbushina, A.A. (2007) Life on the rocks. Environ Microbiol

9: 1613–1631.Gorbushina, A.A., and Broughton, W.J. (2009) Microbiology of

the atmosphere-rock interface: how biological interactions

and physical stresses modulate a sophisticated microbial

ecosystem. Annu Rev Microbiol 63: 431–450.Grube, M., Cernava, T., Soh, J., Fuchs, S., Aschenbrenner, I.,

Lassek, C., et al. (2015) Exploring functional contexts of

symbiotic sustain within lichen-associated bacteria by com-

parative omics. ISME J 9: 412–424.Gustavs, L., Schumann, R., Karsten, U., and Lorenz, M.

(2016) Mixotrophy in the terrestrial green alga Apatococcus

lobatus (Trebouxiophyceae, Chlorophyta) Graham L (ed).

J Phycol 52: 311–314.

Harrell, F.E., Jr. (2017). Hmisc: Harrell Miscellaneous.

R package version 4.0-3. https://CRAN.R-project.org/

package=Hmisc.Hutchens, E., Gleeson, D., McDermott, F., Miranda-

CasoLuengo, R., and Clipson, N. (2010) Meter-scale diver-

sity of microbial communities on a weathered pegmatite

granite outcrop in the Wicklow Mountains, Ireland; evidence

for mineral induced selection? Geomicrobiol J 27: 1–14.Kanehisa, M., Goto, S., Sato, Y., Furumichi, M., and Tanabe,

M. (2012) KEGG for integration and interpretation of large-

scale molecular data sets. Nucleic Acids Res 40: D109–

D114.Klindworth, A., Pruesse, E., Schweer, T., Peplies, J., Quast,

C., Horn, M., et al. (2013) Evaluation of general 16S ribo-

somal RNA gene PCR primers for classical and next-

generation sequencing-based diversity studies. Nucleic

Acids Res 41: e1–e1.Lan, W., Li, H., Wang, W.-D., Katayama, Y., and Gu, J.-D.

(2010) Microbial community analysis of fresh and old micro-

bial biofilms on Bayon temple sandstone of Angkor Thom,

Cambodia. Microb Ecol 60: 105–115.



https://CRAN.R-project.org/package=Hmisc

https://CRAN.R-project.org/package=Hmisc

Lapanje, A., Wimmersberger, C., Furrer, G., Brunner, I., and

Frey, B. (2012) Pattern of elemental release during the

granite dissolution can be changed by aerobic heterotrophic

bacterial strains isolated from Damma Glacier (central Alps)

deglaciated granite sand. Microb Ecol 63: 865–882.Lee, J.-J., Srinivasan, S., Lim, S., Joe, M., Im, S., Bae, S.I.,

et al. (2014) Spirosoma radiotolerans sp. nov., a gamma-

radiation-resistant bacterium isolated from gamma ray-

irradiated soil. Curr Microbiol 69: 286–291.Li, Q., Zhang, B., He, Z., and Yang, X. (2016) Distribution and

diversity of bacteria and fungi colonization in stone monu-

ments analyzed by high-throughput sequencing Brusetti L

(ed). PLoS One 11: e0163287–e0163217.Liermann, L.J., Barnes, A.S., Kalinowski, B.E., Zhou, X., and

Brantley, S.L. (2000) Microenvironments of pH in

biofilms grown on dissolving silicate surfaces. Chem Geol

171: 1–16.

Low-D�ecarie, E., Fussmann, G.F., Dumbrell, A.J., and Bell, G.

(2016) Communities that thrive in extreme conditions cap-

tured from a freshwater lake. Biol Lett 12: 20160562–

20160565.Lower, S., Hochella, M., and Beveridge, T. (2001) Bacterial

recognition of mineral surfaces: nanoscale interactions

between Shewanella and -FeOOH. Science 292: 1360–

1363.Macedo, M.F., Miller, A.Z., Dionisio, A., and Saiz-Jimenez, C.

(2009) Biodiversity of cyanobacteria and green algae on

monuments in the Mediterranean Basin: an overview.

Microbiology 155: 3476–3490.Mailloux, B.J., Alexandrova, E., Keimowitz, A.R., Wovkulich,

K., Freyer, G.A., Herron, M., et al. (2009) Microbial mineral

weathering for nutrient acquisition releases arsenic. Appl

Environ Microbiol 75: 2558–2565.

Mapelli, F., Marasco, R., Balloi, A., Rolli, E., Cappitelli, F.,

Daffonchio, D., et al. (2012) Mineral-microbe interactions:

biotechnological potential of bioweathering. J Biotechnol

157: 473–481.Martiny, J.B.H., Jones, S.E., Lennon, J.T., and Martiny, A.C.

(2015) Microbiomes in light of traits: a phylogenetic per-

spective. Science 350: aac9323–aac9323.Mauck, B.S., and Roberts, J.A. (2007) Mineralogic control on

abundance and diversity of surface-adherent microbial com-

munities. Geomicrobiol J 24: 167–177.Meyer, F., Paarmann, D., D’souza, M., Olson, R., Glass, E.M.,

Kubal, M., et al. (2008) The metagenomics RAST server –

a public resource for the automatic phylogenetic and func-

tional analysis of metagenomes. BMC Bioinformatics 9:

386–388.

Muggia, L., Vancurova, L., �Skaloud, P., Peksa, O., Wedin, M.,

and Grube, M. (2013) The symbiotic playground of lichen

thalli – a highly flexible photobiont association in rock-

inhabiting lichens. FEMS Microbiol Ecol 85: 313–323.Mushegian, A.A., Peterson, C.N., Baker, C.C.M., and Pringle,

A. (2011) Bacterial diversity across individual lichens. Appl

Environ Microbiol 77: 4249–4252.Nayfach, S., and Pollard, K.S. (2015) Average genome size

estimation improves comparative metagenomics and sheds

light on the functional ecology of the human microbiome.

Genome Biol 16: 59–18.Noack-Sch€onmann, S., Bus, T., Banasiak, R., Knabe, N.,

Broughton, W.J., Dulk-Ras Den, H., et al. (2014) Genetic

transformation of Knufia petricola A95 – a model organism

for biofilm-material interactions. AMB Express 4: 80.Northup, D.E., Snider, J.R., Spilde, M.N., Porter, M.L., van de

Kamp, J.L., Boston, P.J., et al. (2010) Diversity of rock var-

nish bacterial communities from Black Canyon, New Mex-

ico. J Geophys Res 115: 19.Nwodo, U., Green, E., and Okoh, A. (2012) Bacterial exopoly-

saccharides: functionality and prospects. IJMS 13: 14002–

14015.Pohl, W., and Schneider, J. (2002) Impact of endolithic bio-

films on carbonate rock surfaces. Geol Soc Lond 205: 177–

194.Porca, E., Jurado, V., �Zgur-Bertok, D., Saiz-Jimenez, C., and

Pa�sic, L. (2012) Comparative analysis of yellow microbial

communities growing on the walls of geographically distinct

caves indicates a common core of microorganisms involved

in their formation. FEMS Microbiol Ecol 81: 255–266.Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T.,

Yarza, P., et al. (2013) The SILVA ribosomal RNA gene

database project: improved data processing and web-

based tools. Nucleic Acids Res 41: D590–D596.Ramirez, K.S., Leff, J.W., Barberan, A., Bates, S.T., Betley, J.,

Crowther, T.W., et al. (2014) Biogeographic patterns in

below-ground diversity in New York City’s Central Park are

similar to those observed globally. Proc R Soc B Biol Sci

281: 20141988–20141988.Rivasseau, C., Farhi, E., Compagnon, E., de Gouvion Saint

Cyr, D., van Lis, R., Falconet, D., et al. (2016) Coccomyxa

actinabiotis sp. nov. (Trebouxiophyceae, Chlorophyta), a

new green microalga living in the spent fuel cooling pool of

a nuclear reactor M€uller K (ed). J Phycol 52: 689–703.

Ruibal, C., Platas, G., and Bills, G.F. (2008) High diversity and

morphological convergence among melanised fungi from

rock formations in the Central Mountain System of Spain.

Pers - Int Mycol J 21: 93–110.Sanders, W.B., P�erez-Ortega, S., Nelsen, M.P., L€ucking, R.,

and de los Rıos, A. (2016) Heveochlorella (Trebouxiophy-

ceae): a little-known genus of unicellular green algae out-

side the Trebouxiales emerges unexpectedly as a major

clade of lichen photobionts in foliicolous communities

DeClerck O (ed). J Phycol 52: 840–853.Schmerk, C.L., Bernards, M.A., and Valvano, M.A. (2011)

Hopanoid production is required for low-ph tolerance, anti-

microbial resistance, and motility in Burkholderia cenocepa-

cia. J Bacteriol 193: 6712–6723.Snider, J.R., Goin, C., Miller, R.V., Boston, P.J., and Northup,

D.E. (2009) Ultraviolet radiation sensitivity in cave bacteria:

evidence of adaptation to the subsurface?. Int J Speleol 38:

11–22.Spribille, T., Tuovinen, V., Resl, P., Vanderpool, D., Wolinski,

H., Aime, M.C., et al. (2016) Basidiomycete yeasts in the

cortex of ascomycete macrolichens. Science 353: 488–491.Uroz, S., Kelly, L.C., Turpault, M.-P., Lepleux, C., and Frey-

Klett, P. (2015) The mineralosphere concept: mineralogical

control of the distribution and function of mineral-associated

bacterial communities. Trends Microbiol 23: 751–762.

V�azquez-Nion, D., Rodrıguez-Castro, J., L�opez-Rodrıguez,

M.C., Fern�andez-Silva, I., and Prieto, B. (2016) Subaerial

biofilms on granitic historic buildings: microbial diversity and

development of phototrophic multi-species cultures. Biofoul-

ing 32: 1–13.



Villa, F., Vasanthakumar, A., Mitchell, R., and Cappitelli, F.(2015) RNA-based molecular survey of biodiversity of lime-stone tombstone microbiota in response to atmospheric sul-

phur pollution. Lett Appl Microbiol 60: 92–102.Walker, J.J., and Pace, N.R. (2007) Endolithic microbial eco-

systems. Annu Rev Microbiol 61: 331–347.Warnecke, F., Sommaruga, R., Sekar, R., Hofer, J.S., and

Pernthaler, J. (2005) Abundances, identity, and growth state

of Actinobacteria in mountain lakes of different UV transpar-ency. Appl Environ Microbiol 71: 5551–5559.

Warscheid, T., and Braams, J. (2000) Biodeterioration ofstone: a review. Int Biodeterior Biodegradation 46: 343–368.

Wickham, H. (2009). ggplot2: Elegant Graphics for Data Anal-

ysis (Use R). New York, NY: Springer.Wickham, H. (2011) The split-apply-combine strategy for data

analysis. J Stat Softw 40:Wu, J., Li, Y., Cai, Z., and Jin, Y. (2014) Pyruvate-associated

acid resistance in bacteria. Appl Environ Microbiol 80:

4108–4113.

Supporting information

Additional Supporting Information may be found in the

online version of this article at the publisher’s web-site:

Fig. S1. Network diagram highlighting significant positiveco-occurrences between individual eukaryotic and bacterialphylotypes (defined at the �97% sequence similarity level).We used relatively strict Spearman correlations (q>0.60,p< 0.01) on phylotypes that were present in at least 50% of

samples to identify general co-occurrence patterns thatexist across multiple rock types and locations. Each phylo-type is labelled either by phylogenetic order or the lowestresolvable phylogenetic level. Several phylotypes that could

not be classified to the class level of resolution are simplylabelled as ‘Fungi’. The size of each circle (phylotype) cor-responds to its average abundance across all samples.

Many of these taxa are lichen associated; see SupportingInformation Table S2 for additional information. The numberassociated with each phylotype indicates its position in Sup-porting Information Table S2.

Fig. S2. Isolates from granite-specializing bacterial familieshave significantly lower pH limits for growth than isolatesfrom limestone-specializing families (Lower limit p<0.0001;Higher limit p<0.05; Mann-Whitney test). We lookedthrough the literature for isolates that belonged to bacterialfamilies enriched on limestone or granite, and found thatgranite enriched families consistently had lower pH limitsfor growth. On the left we show the lower pH limit forgrowth listed for isolates, and on the right the upper pH limitfor growth. We attempted to replicate this approach for theeukaryotic rock-specialists, but we were constrained by alack of consistent phenotypic data for closely-relatedisolates.Fig. S3. The functional composition of bacterial communi-ties differs significantly based on rock type. PCoA plotshowing the differences in functional dissimilarity matricesbetween rock types in a single Maine cemetery. Rock typewas a significant factor describing differences in KEGGfunctional potential for bacterial communities (p<0.05,R2 5 0.34; PERMANOVA), but not for eukaryotic communi-ties (p> 0.05; PERMANOVA).Fig. S4. Sampling set-up for tombstone surfaces. A 100cm2

template was affixed to tombstone surfaces and this areawas scrubbed vigorously with individually wrapped tooth-brushes to collect sufficient biomass. Pictures were taken ofthe front and back of tombstones to assess rock type. Thename of the deceased has been blurred.Table S1. Site Details.Table S2. Co-occurrence matrix OUT details.

Table S3. KEGG pathways significantly enriched on selectrock types.



tales from the tomb: the microbial ecology of exposed rock...

Documents