**3. Results**

#### *3.1. Endolithic Bacterial and Phototrophic Growth*

Visual inspection of the colonized tiles showed a marked increase in both pigmentation and erosion with time (Figure 2c–f). The tiles sustained both nonphototrophic and phototrophic bacterial growth over the 9-month exposure period. Bacterial biomass increased at an average rate of 3 × 10<sup>10</sup> 16S rRNA gene copies per m<sup>−</sup><sup>2</sup> month−1, reaching a mean value of 1.1 × 10<sup>11</sup> 16S rRNA gene copies per m −2 at 9 months (Figure 3a). As expected, phototroph colonization followed a similar trend with photopigment content increasing at a rate of 2.3 mg m<sup>−</sup><sup>2</sup> month−1, reaching an average value of 7.22 mg m −2 by 9 months. (Figure 3b), at which point 16S rRNA gene counts were not significantly different from those found in steady-state climax communities described by Couradeau et al. [14] and Roush et al. [15] (Student's *t*-test, *p* < 0.05). While total chlorophyll pigment concentrations were not significantly different between 9 months and steady-state climax communities either, cyanobacteria-specific counts were actually higher at 9 months than in steady-state climax communities.

**Figure 2.** Endolithic colonization of travertine tiles. (**a**) Areal concentration of 16S rRNA gene copies. Each bar is an independent replicate. Error bars are from biological replicates. (**b**) Areal concentration of total photosynthetic chlorins (chlorophylls plus bacteriochlorophylls). Single determinations were carried out for each replicate tile. (**<sup>c</sup>**–**f**) Photographic evidence of colonization after removal of epilithic biomass by brushing. (**c**) Initial, virgin tile. Excising squares were samples used for analyses.

#### *3.2. Incidence of Anoxygenic Phototrophs*

APB abundance measured by bacteriochlorophylls increased with time but trailed in concentration by some two orders of magnitude to cyanobacterial abundance measured by chlorophylls during the colonization period. This situation obviously changed significantly later during succession, as bacteriochlorophylls were statistically as abundant as chlorophylls when compared to steady-state climax communities (Figure 3b). The magnitude of the difference between APB and cyanobacteria was less marked, but still very significant, when measured by 16S rRNA gene abundance (Figure 3a). By using this metric it was obvious that although APB trailed cyanobacteria during the colonization period, they eventually matched and even exceeded cyanobacteria in steady-state climax communities. We found very differing dynamics between populations of relevant APB groups: while Chloroflexales were only present in very small quantities during early phases (Figure 3c) and reached only 6.2 × 10<sup>6</sup> 16s rRNA gene copies per m<sup>−</sup><sup>2</sup> after 9 months, *Erythrobacter* abundance was stable throughout the colonization, with an average of 1.2 × 10<sup>9</sup> 16S rRNA gene copies per m-2 at 9 months. In comparison, the situation was reversed in steady-state mature communities, where *Erythrobacter* sp. decreased to some 1.6 × 10<sup>8</sup> 16s rRNA gene copies per m<sup>−</sup><sup>2</sup> in steady-state climax communities, but Chloroflexales increased to populations close to those of cyanobacteria (Figure 3c). The apparent differences in trends between bacteriochlorophyll and 16S rRNA genes as proxies for population size can be explained by the relatively low bacteriochlorophyll content of *Erythrobacter* spp. compared to members of the Chloroflexales [51,52], which essentially made the total content of bacteriochlorophyll be very sensitive to the population size of the latter.

**Figure 3.** Time series of bacterial biomass proxies detected in colonized tiles and steady-state climax communities by guild or taxon. (**a**) Areal concentrations of 16S rRNA gene copies based on quantitative PCR and high-throughput sequencing phylogenetic assignments (**b**) Areal photosynthetic chlorins as biomarkers for oxygenic phototrophs (total chlorophylls) or APB (total bacteriochlorophylls) (**c**) areal population size of APB clades *Erythrobacter* spp. and Chloroflexales based on quantitative PCR and high-throughput sequencing phylogenetic assignments. (**d**) Endolithic colonization dynamics of specific microboring cyanobacterial clades, based on qPCR and bioinformatic placement of high-throughput environmental sequences using the Cydrasil cyanobacterial reference tree and database. Error bars are for biological sample triplicates.

#### *3.3. Cyanobacterial Succession: Diversity and Composition*

Unexpectedly, cyanobacterial richness gauged by the number of observed amplicon sequence variants (ASVs) was not significantly different across time points and when compared to steady-state climax communities (Kruskal–Wallis, *p* = 0.33; Table 1), whereas ASV evenness (measured as Pielou's Evenness) decreased significantly (Kruskal–Wallis, *p* = 0.04) with time. Pairwise Kruskal–Wallis comparisons indicated that the difference was driven by a drop in evenness between early (3 and 6 months) and late succession communities (9 month and steady-state climax) (adjusted *p* = 0.07 for all four comparisons). Shannon's diversity also followed the evenness trend, with significant differences with time (Kruskal–Wallis, *p* = 0.02) where late succession samples were less diverse than early succession samples (adjusted *p* = 0.06 for all four comparisons). Regarding cyanobacterial community composition (beta-diversity), all time points and steady-state climax communities were significantly different from each other (PERMANOVA, *p* < 0.05, pairwise Kruskal–Wallis *p* < 0.05), a result also supported statistically by a PCOA (principal coordinates ordination analysis; Weighted UniFrac metric; Figure S2).


**Table 1.** Alpha diversity metrics of cyanobacterial endolithic communities in tiles placed in the intertidal zone of Isla de Mona and metrics from geographically similar natural substrate communities on Isla de Mona described by Roush et al. [15].

Community composition of steady-state climax communities was taken from calcite samples published in Roush 2018. Lower-case letters denote samples not significantly different (α = 0.1). ASVs, amplicon sequence variants.

#### *3.4. Identification of Endolithic Cyanobacteria Clades*

In nonporous virgin substrates, only euendolithic organisms can colonize and grow to large abundance. Since we removed all epilithic biomass before sequencing, those organisms found to be abundant early on can be deemed to be bona fide euendoliths since they must have been able to excavate the substrate. Therefore, to identify pioneer euendolithic cyanobacterial clades, the most abundant cyanobacterial ASVs from the 3-month-old tiles were placed using the RAxML Evolutionary Placement Algorithm into the Cydrasil reference cyanobacterial 16S rRNA gene tree containing 980 curated cyanobacterial sequences, which includes all full-length 16S rRNA gene sequences traceable to known euendolithic cyanobacteria (Figure 3d and Figure S3). Euendolithic sequences that were not full length were included in the query sequence list and checked for correlation with known clades. In order to pare down the dataset for placement, we ranked each sample's cyanobacterial ASVs in order of abundance until cumulative counts reached 95% of the total abundance in each sample, yielding 213 unique ASVs across all tile samples and steady-state climax communities. We then placed the resulting pared ASV dataset into the Cydrasil reference tree. Of the 213 initial ASVs, 139 were placed with high confidence and clustered onto four distinct tree nodes. Two of the nodes contained known euendolithic species: Cluster 2 (containing 37 unique ASVs) encompassed endolithic members in the Pleurocapsales, and Cluster 3 (27 ASVs) contained *Mastigocoleus testarum*. The other two did not align with known euendoliths: one contained *Leptolyngbya* species (Cluster 1; 60 ASVs) and the other was a novel clade that contained only environmental sequences lacking taxonomic assignment and only distantly related (<95.2% similarity) to *Stanieria cyanosphaera*. We named this clade UBC (15 ASVs), for "unknown boring cluster".

#### *3.5. Colonization Dynamics of Euendolithic Cyanobacterial Clades*

To quantify colonization dynamics, qPCR-normalized abundances of the euendolithic clusters were plotted over time (Figure 3d and Figure S3). Members of the UBC were double to an order of magnitude more abundant than the other groups after 3 months of exposure, with an average biomass of 9.1 × 10<sup>9</sup> 16S rRNA gene copies per m-2. UBC abundance remained stable throughout the experiment and was not significantly di fferent when compared to steady-state climax communities. Cluster 1 (*Leptolyngbya-*like) population size lagged that of UBC, reaching a maximum after 6 months (1.1 × 10<sup>9</sup> 16s rRNA gene copies per m<sup>−</sup>2). Clusters 2 (Pleurocapsalean) and 3 (*Mastigocoleus-*like) colonized substrate at the slowest rate, reaching maximum populations after 9 months (8.7 × 10<sup>10</sup> and 9.9 × 10<sup>9</sup> 16S rRNA gene copies per m<sup>−</sup>2, respectively). Clusters 2 (Pleurocapsalean) and 3 (*Mastigocoleus-*like) also decreased in abundance in steady-state climax communities.

#### *3.6. Di*ff*erential Abundance Analysis*

In order to identify which cyanobacterial colonizers were driving compositional di fferences between early (3-month) and late (9-month) tiles, we conducted a di fferential abundance analysis. The most abundant and significant ASVs at both time points were members of the four clades delineated

above (Figure S4). At 3 months, representatives of the UBC were three of the four most abundant cyanobacteria (*p* < 0.05), both in total sequence count and in di fferential relative abundance (fold change) with respect to 9-month communities. The fourth ASV was a member of Cluster 1, allied to *Leptolyngbya*. At 9 months, Cluster 2 (Pleurocapsalean) and Cluster 3 (*Mastigocoleus*-like) sequences were found to be the most di fferentially abundant with respect to 3-month communities.

#### *3.7. New Pioneer Euendolith Clade*

Both qPCR-adjusted relative abundance and di fferential abundance analysis revealed that the previously unknown UBC clade played a significant role in early colonization of hard intertidal carbonates. In order to better constrain its identification, we conducted a maximum-likelihood phylogenetic reconstruction of 395 sequences, largely from cyanobacterial isolates (Figure 4), but including those of the most di fferentially abundant UBC and the seven sequences most similar to UBC that we could find by BLAST analyses. As before (i.e., Figure 3d), UBC members were only distantly related (<5.2% similar) to cultured cyanobacteria, the nearest being *Stanieria cyanosphaera* (formerly *Chroococcidiopsis cyanosphaera*), an epilithic freshwater unicellular cyanobacterium [53]. UBC was distant from the canonical euendolithic groups, with the Cluster 2 (Pleurocapsalean) being the closest. However, UBC members were phylogenetically close to environmental sequences obtained from marine carbonate microbialites, a habitat not dissimilar from the interior of hard carbonates and containing known euendoliths [54].
