Unraveling the microbial processes of black band disease in corals through integrated genomics

Coral disease outbreaks contribute to the ongoing degradation of reef ecosystems, however, microbial mechanisms underlying the onset and progression of most coral diseases are poorly understood. Black band disease (BBD) manifests as a cyanobacterial-dominated microbial mat that destroys coral tissues as it rapidly spreads over coral colonies. To elucidate BBD pathogenesis, we apply a comparative metagenomic and metatranscriptomic approach to identify taxonomic and functional changes within microbial lesions during in-situ development of BBD from a comparatively benign stage termed cyanobacterial patches. Results suggest that photosynthetic CO2-fixation in Cyanobacteria substantially enhances productivity of organic matter within the lesion during disease development. Photosynthates appear to subsequently promote sulfide-production by Deltaproteobacteria, facilitating the major virulence factor of BBD. Interestingly, our metagenome-enabled transcriptomic analysis reveals that BBD-associated cyanobacteria have a putative mechanism that enables them to adapt to higher levels of hydrogen sulfide within lesions, underpinning the pivotal roles of the dominant cyanobacterium within the polymicrobial lesions during the onset of BBD. The current study presents sequence-based evidence derived from whole microbial communities that unravel the mechanism of development and progression of BBD.


Bacterial community structures of microbial lesions associated with cyanobacterial patches and black band disease
In both metagenomic and metatranscriptomic datasets derived from polymicrobial lesions of cyanobacterial patches (CP) and black band disease (BBD), sequences assigned to Cyanobacteria, Gammaproteobacteria, Epsilonproteobacteria and Deltaproteobacteria were more abundant in BBD than in CP, while those assigned as Alphaproteobacteria, Cytophaga, were highly consistent with those in previous PCR-based microbial profiling studies , Sato et al. 2013. Specifically, a shift in the dominant cyanobacterial types and increase in relative abundance of sequences associated with Epsilonproteobacteria (dominated by Arcobacter spp. ) and Deltaproteobacteria (dominated by Desulfovibrio spp.) are common features during transition from CP to the BBD stage of pathogenesis. Overall consistency in the taxonomic compositions between metagenomic and metatranscriptomic libraries suggests that taxonomic compositions present as genomic materials (i.e. DNA) in the polymicrobial lesion specimens closely represent the compositions of taxa that are actively transcribing functional genes in the community and vice versa. Interestingly, within the CP libraries, cyanobacterial sequences were under-represented and alphaproteobacterial sequences over-represented in the metagenomic data compared to the metatranscriptomic data. Although these patterns suggest that the presence of DNA may not always accurately reflect relative functional importance for certain taxonomic groups in a complex microbial community, a detailed reason for why only these taxa displayed such patterns remains unclear.
3 Cyanobacterial sequences dominating the microbial communities of CP and BBD showed a clear shift from a Trichodesmium-affiliated cyanobacterium in CP to a cyanobacterium affiliated to Roseofilum reptotaenium in BBD, consistent with previous studies , Sato et al. 2013. Importantly, complete 16S rRNA gene sequences recovered from the assembled genomic bins of these cyanobacteria (i.e. Cya1 in CP and Cya2 in BBD) confirmed their phylogenetic positions closely matching with cyanobacterial sequences previously retrieved from CP-and BBD-associated cyanobacteria on the GBR Suppl. Figure 5). The sequence of the BBD-cyanobacteria obtained in this study was also positioned within a tightly clustered group including R. reptotaenium and other cyanobacterial sequences retrieved from BBD lesions collected from a wide geographic range including the Caribbean, Rea Sea and Indo-Pacific regions (Supp. Figure 5). There are other divergent cyanobacterial sequences that have been recovered from BBD lesions although some of them constitute only minor components within BBD-associated microbial communities , Sato et al. 2013. Interestingly, a few of these sequences (e.g. AY123040, DQ446127) form a monophyletic cluster with the CP-dominating cyanobacterial sequences (Suppl. Figure 5). This suggests that a similar successional development of BBD derived from CP-like lesions also occurs in other geographical locations although CP lesions have thus far been reported only from the GBR.
Gammaproteobacterial sequences were also present as an abundant group in both CP and BBD though relatively more abundant in BBD. The compositions of gammaproteobacterial sequences were dominated by those that were not clearly assigned to any identified order, but also including Alteromonadales-and Oceanospirillales-associated sequences in both CP and BBD lesions (Suppl. Data 1 and 2), displaying a pattern also observed in previous sequence-based studies , Sato et al. 2013).
Gammaproteobacterial-assigned transcriptomes indicated that Gammaproteobacteria in BBD 4 lesions utilize diverse metabolic substrates including proteins, carbohydrates, amino acids and fatty acids (Suppl. Figure 3). This versatile resource utilization capability within this group may explain the consistent presence of Gammaproteobacteria during the transition from CP to BBD without dramatic changes in the taxonomic compositions while available organic materials and microenvironmental conditions change.

Other taxonomic groups including Archaea, Eukaryota, and viruses within the lesions
Eukaryotic sequences accounted for between 1 and 3% of the total taxonomicallyassigned sequences in CP and BBD metagenomes and metatranscriptomes, based on the search against the Universally Conserved Proteins (UCP) database (Figure 2a, Suppl. Data 1 and 2). CP and BBD sequences in metagenomes and metatranscriptomes were assigned to diverse eukaryotic taxa, including Bacillariophyta, Chordata, Streptophyta and Ciliophora (Suppl. Figure 6). Large proportions of these sequences, however, especially in metagenomes, could not be confidently assigned to a defined phylum or had weak support (bit scores < 51) in searches against the UCP database. Archaeal-affiliated sequences represented between 0.1 and 0.2 % of the total taxonomically-assigned sequences across the datasets. Sequences assigned to an unassigned archaeal class, unclassified Thaumarchaeota (including Nitrosopumilales), and the classes Methanopyri and Thermoprotei were more abundant in CP than BBD metagenomes, though archaeal metatranscriptomic sequences displayed inconsistent relative abundance patterns, with an unassigned archaeal class being relatively more abundant in BBD than CP metatranscriptomes (Suppl. Figure 7). In addition, these sequence abundance patterns were not consistent with previous 16S rRNA gene profiling of BBD and CP archaeal communities, which detected a dominant uncultured archaeon belonging to an unique lineage remotely associated with Euryarchaeota (Sato et al. 2013).
Overall, the results of eukaryotic and archaeal sequence compositions indicate a potential shortcoming of meta-'omics' approaches for profiling rare members of these microbial communities, due to a lack of sequence information in the publically available sequence databases that represents numerous understudied and unidentified environmental microorganisms (Rinke et al. 2013).
CP and BBD datasets contained between 0.02 and 0.04% of sequences taxonomically assigned to viruses (Suppl. Data 4 and 5). Viral-associated metagenomic sequences were mostly dsDNA viruses in both CP and BBD, while viral metatranscriptomic sequences included dsDNA, ssDNA, and positive-sense ssRNA viruses in CP, as well as unclassified environmental viruses that accounted for approximately 67% of BBD-viral sequences (Suppl. Figure 8). These observations reflect the difference in the nucleic acid types that were extracted and sequenced (i.e. DNA vs. RNA), but also highlight that since viruses can consist of dsDNA, ssDNA, dsRNA and ssRNA genomes, substantial biases can be introduced when investigating environmental samples by adopting different ribonucleotide targeted approaches , Wood-Charlson et al. 2015. Sequences assigned to the family Caudovirales accounted for the largest proportions in the viral-associated sequences of the CP-derived metagenome and metatranscriptome as well as BBD-derived viral metagenomic sequences, while the viral sequences in the BBD metatranscriptome consisted of a high relative abundance of uncultured cyanophage-associated sequences. Interestingly, these cyanophage-assigned sequences shared 99% sequence identity (over 104 nt) with 2 genomic regions coding the PSII D1-protein within the Cya2 genomic bin that represented the dominant cyanobacteria in BBD. We thus hypothesize that the dominant cyanobacterium in BBD has a host-phage association with a novel type of cyanophage, which encodes the PSII gene likely acquired from the host (Lindell et al. 2004, Sullivan et al. 2006, Chénard and Suttle 2008. Furthermore, the BBD metagenome had relatively more viral sequences than 6 CP that were associated with Cellulophaga phage phiST, which infects members of the Bacteroidetes (Holmfeldt et al. 2013;Suppl. Figure 8), a bacterial group that was relatively abundant in CP but not well-represented in BBD (Figure 2a). This may indicate that these phages removed Bacteroidetes in CP lesions and remained highly abundant in the BBD lesions. Characterization of viral communities associated with coral diseases is becoming increasingly important. Viruses can be considered not only as disease-causing agents affecting a coral host and its symbionts (e.g. symbiotic algae Symbiodinium spp.) (Pollock et al. 2014, Soffer et al. 2014, Correa et al. 2016, or phenotypes in bacterial pathogens thereby indirectly triggering coral disease , but also as therapeutic agents to control coral diseases by lysing pathogenic bacteria (Atad et al. 2012, Cohen et al. 2013.
The presence of the dominant cyanophage-associated sequences in the BBD metatranscriptome highlights a potential option for phage-therapy. However, further investigations are required to characterize the role of viruses in the disease dynamics before we can consider such therapy strategies as a potential tool for mitigating impacts of BBD.

A potential adaptive mechanism within the BBD-dominating cyanobacterium
Data presented here indicate that the BBD-dominating cyanobacterium may perform anoxygenic photosynthesis as an additional mechanism to adapt to increased sulfide concentrations in the BBD lesion. Gene expression patterns of the genomic bin Cya2, representing the BBD-dominating cyanobacteria, suggested the presence of a putative sulfide-adaptive mechanism in their photosynthetic regulation, where genes coding photosystem I (PSI) were highly expressed over those coding photosystem II (PSII; Figure   4a). An alternative explanation for this observed pattern is that the BBD-dominating cyanobacterium possibly performs PSI-driven anoxygenic photosynthesis using sulfide as an 7 electron donor (Cohen et al. 1986, Miller andBebout 2004), and thus PSII appears to be downregulated in comparison to the expression of PSI reaction center. Interestingly, sequences of a sulfide-quinone reductase (sqr), which drives anoxygenic photosynthesis in cyanobacteria (Arieli et al. 1994), is present in the Cya2 genomic bin and transcriptomes mapped to the sqr sequence were approximately 17-fold more abundant within BBD than in CP (data not shown). Previous experimental work has failed to demonstrate that the BBDdominating cyanobacterium, R. reptotaenium, performs sulfide-dependent anoxygenic photosynthesis since a non-axenic culture of R. reptotaenium did not survive in the presence of dichlorophenyldimethyl urea (DCMU, an inhibitor of electron flow in PSII (Trebst 1980)) and sulfide under light conditions (Myers and Richardson 2009). However, it has also been observed that DCMU does not inhibit the experimentally-induced development of BBD or the progression of active BBD lesions dominated by R. reptotaenium on coral colonies ). Assuming that DCMU effectively inactivates PSII of R.
reptotaenium, these results indicate two possibilities: (1) The cyanobacterium can grow heterotrophically, or (2) it performs photosynthesis independent of the oxygen-evolving PSII, e.g. anoxygenic photosynthesis using sulfide as an electron-donor. The former is not wellsupported since a recent culture-based incubation experiment demonstrated that R.
reptotaenium does not survive heterotrophically without light (Stanic et al. 2011), whereas the latter is more supported given the presence sulfide in deeper anoxic areas of the BBD lesion (Carlton andRichardson 1995, Glas et al. 2012) and the sequence-based evidence of sulfide-quinone reductase expression presented in this study. Furthermore, a field study observed sulfide oxidizing bacteria (SOB)-like organisms on the surface of the BBD lesion under high-light conditions, and it was proposed that the phenomenon was due to occasional non-oxygenation of the mat surface under light because these SOB are expected to be at the oxygen-sulfide interphase (Viehman and Richardson 2002). This field observation thus 8 provides additional support for the capability of the BBD cyanobacterium to perform anoxygenic photosynthesis without relying on PSII.
The isiA gene, coding a chlorophyll-binding protein IsiA, also appeared to be highly expressed in BBD (Table 2). This protein is induced in iron-deficient conditions and protects cells from light-induced damage by increasing the size of a photo-harvesting antenna attached to PSI, thereby converting light-energy to heat (Wilson et al. 2007). Iron-deficiency is especially critical for cyanobacteria as their photosynthetic apparatus requires between 22 and 23 molecules of iron per compound (Ferreira and Straus 1994). Cyanobacteria, like other Bacteria, adapt to iron-limiting environments by secreting iron-scavenging siderophores and substituting ferredoxin with flavodoxin (Ferreira and Straus 1994). In this study, Cya2-genes coding a siderophore transporter and flavodoxin were upregulated and 3 out of 4 [2Fe-2S] ferredoxins were downregulated in BBD relative to CP, although these differences were not statistically significant (data not shown). These observations suggest that the BBD cyanobacterium has strategies to mitigate the adverse-effects of iron-deficiency and grow at a high-density. Within the BBD lesion, cyanobacterial filaments are found to be neatly aligned (Miller et al. 2012), which may be responsible for intensifying vertical stratification of microenvironmental conditions in BBD lesions compared to CP lesions (Glas et al. 2012).
These sequence-based results provide further support for the role of tightly-aggregated cyanobacterial biomass in BBD pathogenesis through retention of anoxic and sulfidic conditions in the BBD lesion.

Transcriptomic patterns of the gammaproteobacterial bin Oce
Profiles of transcriptomes mapped to the metagenomic bin Oce, sharing 98% sequence identity with the 5S rRNA gene sequence of Thalassolituus oleivorans R6-15 9 (GenBank Acc. CP006829; belonging to the order Oceanospirillales), indicated differential gene expression patterns between CP and BBD libraries in a wide range of functions (Suppl. Table 2). Sequence regions relatively more expressed in BBD than in CP included those coding fatty acid metabolism, glycerol metabolism, ATP synthesis and virulence factor, while sequences relatively more expressed in CP than in BBD included genes associated with virulence regulation, stress response, oxidative stress resistance, polysaccharide synthesis and secretion (Suppl. Table 2). Notably, isocitrate lyase (icl)-and aerobic glycerol-3-phosphate dehydrogenase (glpD)-coding genes were relatively more expressed in BBD than CP.
Isocitrate lyase is an enzyme essential to the glyoxylate bypass, an anaplerotic route for replenishing the TCA cycle through assimilation of acetyl coenzyme A (Kornberg and Krebs 1957). Beta-oxidation of fatty acids produces substantial acetyl coenzyme A and therefore the glyoxylate bypass is important for heterotrophic growth of bacteria utilizing fatty acid substrates (Kornberg andKrebs 1957, McKinney et al. 2000). An aerobic glycerol-3phosphate dehydrogenase is an essential enzyme for aerobic (i.e. oxygen-consuming) growth of bacteria utilizing glycerol (Cozzarelli et al. 1965). Therefore, greater expression of these enzymes in BBD together suggests that Oce bacteria more actively utilize lipids as growth substrates and consumes oxygen in BBD than in CP. Fatty acids utilized by Oce may be at least partially sourced from degraded cyanobacterial fatty acids, of which higher synthetic gene abundance and expression were observed in BBD than in CP (Figure 3; labelled as 'mycolic acid synthesis'; also see the main document). Another possible source of lipids is degrading coral tissues since (1) rates of disease progression are faster in BBD than CP , providing more coral-derived organic matter to the BBD lesion-associated microbial communities (Sato et al. 2015), and (2) 10~30% of the dry weight of coral tissue consists of lipids, an important energy storage for the animal, in the form of phospholipids, diacylglycerol, free fatty acids, and triacylglycerol (Harland et al. 1993, Grottoli et al. 2004). Furthermore, bacteria represented by Oce may perform robust energy production potentially with this increased lipid substrates in BBD. This is supported by significantly increased relative abundance of transcriptomes coding F0F1-type ATP synthase subunits in BBD (atpA, atpA2, LOR_35c02990; Suppl. Table 2), and 11 out of 13 other genomic regions coding subunits of F0F1-type ATP synthase were also more expressed in BBD than in CP (data not shown). Oce in CP displayed higher expressions of genes coding a transcriptional regulatory CpxR protein and RseC protein, both indicative of envelope stress responses (Missiakas et al. 1997, Bianchi and Baneyx 1999, Buelow and Raivio 2005, and an alkyl hydroperoxide reductase subunit F (ahpF) indicative of oxidative stress (Poole et al. 2000).
These gene-expression patterns indicative of stress suggest that microenvironmental conditions within CP lesions are not preferred conditions for Oce. In addition, an increased expression of a gene in CP coding a RNA polymerase sigma-H factor (algU), an initiator of polysaccharide biosynthesis in the form of alginate (Martin et al. 1993), may indicate enhanced bacterial adaptation to oxidative stress by energy investment in mucoid surface production (Mathee et al. 1999, Sabra et al. 2002. Interestingly, Oce in BBD indicated a greater gene expression of a virulence factor (sigL) and a lower expression of a virulence down-regulator (cdpA) than within CP. Extracytoplasmic function sigma factor SigL has an important role in pathogenesis as it controls the pathogen's persistence through surface interaction with infected host (Hahn et al. 2005, Dainese et al. 2006, whereas the downregulation of a 3',5'-cyclic adenosine monophosphate phosphodiesterase CdpA elevates the intercellular concentration of cyclic adenosine monophosphate (cAMP), thereby enhancing the expression of a number of cAMP-dependent virulence genes such as exotoxin-, type 3 secretion system-and effector protein-coding genes (Fuchs et al. 2010). In summary, observed patterns in the mapped transcriptomes in Oce collectively suggest that (1) an increased influx of total lipids derived from degrading BBD-cyanobacteria and/or coral tissues at the BBD stage enhances energy-generating metabolism of Oce bacteria, (2) the enhanced metabolism may confer further virulence to the Oce bacteria within BBD lesions, potentially further damaging coral tissue through host-microbial interactions, and (3) increased oxygen-consuming heterotrophy on lipids observed in BBD will further contribute to the formation of anoxia within the BBD lesion, one of the major virulent biogeochemical factors in BBD (Glas et al. 2012).
Given an indication of high genetic contamination in the Oce bin (Table 1), the above interpretations of the observed gene expression require extra caution as they may not be derived from a single species population. However, even if genomic sequences in Oce were derived from more than one bacterial species, biological indications inferred from the gene expression profiles of Oce-associated organism(s) provide insights into their role in the overall lesion functioning of the BBD-associated community as a collective taxonomic group. This is because Oce was assembled with sequences sharing similar taxonomic positions, coverage ratio in CP and BBD, and nucleotide signatures such as GC% and tetranucleotidefrequency. To resolve gene expression patterns of Oce at a single-species level, the examination of these hypotheses requires further investigations using more focused approaches, such as single-cell genomics.

Little sequence-based evidence for involvement of toxin-related genes in BBD pathogenesis
Pathogenicity of the BBD microbial mat has also been attributed to other microbiologically-mediated virulence factors including the biosynthesis of cyanotoxins ) and Vibrio-toxins (Barneah et al. 2007, Arotsker et al. 2009).
Production of nodularin and microcystin has been demonstrated from BBD lesions and BBD-12 derived cyanobacterial cultures from the Caribbean , Stanic et al. 2011 although these toxins were not detected in BBD found on the GBR (Glas et al. 2010). We searched BBD and CP datasets against a set of publically-available nucleotide sequences that code biosynthesis pathways of nodularin (ndaABCDEFGHI) and microcystin (mcyABCDEGJ) using tBLASTx (Camacho et al. 2009;e-value ≤ 10 -5 ). Sequences associated with nodularin-synthesizing genes were relatively more abundant in CP than BBD in both metagenomes and metatranscriptomes (Suppl. Figure 9), suggesting that nodularin is not linked to the increase of lesion-virulence during the onset of BBD from CP ). The relative abundance of microcystin-synthesizing genes were 1.5-fold lower in the metagenome of BBD than that of CP, although the metatranscriptomes indicated that they were 1.2-times more highly represented in BBD than in CP (Suppl. Figure 9). Given that the relative sequence abundance of the Cya2 bin representing the BBD-dominating cyanobacteria was approximately 291-times higher in the BBD metagenome than CP metagenome (Table 1), it is unlikely that these microcystin synthesizing genes are harbored by the Cya2 cyanobacterium. Instead, the presence of other cyanobacterial types that represent minor constituents of BBD and CP lesions have been demonstrated in the GBR region , Sato et al. 2013, and thus these may be the host of the microcystin and nodularin biosynthesis sequences observed in the datasets. Gene-prediction and functional annotation of sequences within the BBD-dominating Cya2 genomic bin also indicated that microcystin production genes are not present in the Cya2 sequences. The Cya2 bin demonstrated a high genome coverage (99%; Table 1) and therefore the inability to observe microcystin production genes in Cya2 is unlikely due to missing gene reads or a lack of completeness in this bin, but more likely due to the region-specific variation of BBD-associated cyanobacterial genotypes among the Caribbean, Red Sea and Info-Pacific coral reefs , Glas et al. 2010. One gene-coding region (7,056-nt) in Cya2 was 13 annotated as an amino acid adenylation enzyme/thioester reductase family protein and shared 48% amino acid sequence identity with a puwainaphycin-production gene puwA, which is responsible for the final peptidyl elongation prior to the cyclization of this cytotoxic lipopeptide (Mareš et al. 2014). This may indicate the presence of a potential cyanotoxinassociated gene sequence in BBD-dominating cyanobacterial sequences obtained from the GBR. However, transcriptomes that were mapped against this sequence represented extremely low proportions (0.002% and 0.003% of the total number of transcriptomes mapped against Cya2 in CP and BBD libraries, respectively), and the gene sequence arrangement in the Cya2 around this candidate puwA region does not indicate the domain structure required for the synthesis of puwainaphycin (Mareš et al. 2014). Therefore, its potential function and contribution to the overall BBD pathogenicity remain unclear and subject to further investigations. Our finding contrasts the previous metatranscriptomic study by Arotsker et al. (2016), which has proposed that the most transcribed gene in the BBD lesion, namely cyanobacterial adenosylhomocysteinase, is involved in cyanotoxin production.
However, a link between the adenosylhomocysteinase and cyanotoxin production has not been established, as adenosylhomocysteinase is one of the most wide-spread, highly conserved proteins among prokaryotes and eukaryotes and plays a critical role in the metabolism of sulfur-containing amino acids (Sganga et al. 1992).
Metagenomic and metatranscriptomic sequences that were affiliated to the virulence factors of Vibrio spp. (i.e. all genes included in SEED classifications of 'Vibrio pathogenicity island' and 'Cytolysin and lipase operon in Vibrio') were also compared between CP and BBD datasets. While metagenomic sequences coding Vibrio-virulence factors were relatively more abundant in BBD than in CP, those in metatranscriptomic datasets were more abundant in CP than BBD (Suppl. Figure 9). These results suggest that even though functional genes 14 associated with the pathogenicity of Vibrio were present in BBD metagenomes, those sequences are not actively transcribed in the BBD lesion compared to the CP lesion.
Sequences associated with Vibrio zinc-metalloproteases, a known gene associated with coral tissue lysis (Sussman et al. 2008, Sussman et al. 2009), were also only confirmed at low relative abundance within the CP-metagenome and the BBD-transcriptomes (representing only 0.0005% and 0.00003% of functionally annotated sequences in each dataset, respectively). Together, these observations suggest that Vibrio-related toxins are probably not a major factor involved in the development of BBD pathogenicity during the onset of disease originated from CP, and thus Vibrio spp. within the BBD lesions are likely secondary opportunists among other heterotrophic organisms utilizing available organic substrates (Saeed 1995, Wendling andWegner 2013). 15

Supplementary figures and tables
Supplementary Figure 1 Differential coverage and GC-content of metagenomic assembly recovered from microbial lesions of cyanobacterial patches (CP) and black band disease (BBD). Contigs larger than 5,000-bp with more than 0.1 mean coverage in both CP and BBD metagenomes are shown. Crosses indicate the mean coverage of the resulting genomic bins shown in Table 1. All proportions shown are calculated as the relative abundance of sequences within the indicated taxa divided by the total number of sequences that were matched with Universally Conserved Protein sequences (E-value < 1.0E-5; left, relative proportions; right, differences between proportions (negative value indicates BBD > CP) with 99.9% confidence intervals shown with error bars (most of them are not visible due to small ranges)). All differences in proportions were statistically significant (p-value < 1e-100).  Figure 6 Taxonomic compositions of eukaryotic annotated sequences in the metagenomes (a) and metatranscriptomes (b) derived from microbial lesions of cyanobacterial patches (CP) and black band disease (BBD) at the phylum level. All proportions are calculated as the relative abundance of sequences within the indicated taxa divided by the total number of sequences that were matched with Universally Conserved Protein sequences (E-value < 1.0E-5; asterisks denote taxa that have not been officially classified at the phylum level (see Supplementary Data 1 and 2)).  Figure 7 Statistical comparisons of taxonomically annotated archaeal sequences at the class level in the metagenomes (a) and metatranscriptomes (b) derived from microbial lesions of cyanobacterial patches (CP) and black band disease (BBD). All proportions are calculated as the relative abundance of sequences within the indicated taxa divided by the total number of sequences that were matched with Universally Conserved Protein sequences (E-value < 1.0E-5; left, relative proportions; right, differences between proportions (negative value indicates BBD > CP) with 99.9% confidence intervals shown with error bars). Asterisk denotes a taxon that has not been officially classified at the phylum level (see Supplementary Data 1).   Comparisons of the relative abundance of sequences affiliated with genes involved in the production of nodularin and microcystin and genes associated with Vibrio-associated virulence factors in the metagenomes (a) and metatranscriptomes (b) recovered from microbial lesions of cyanobacterial patches (CP) and black band disease (BBD). Graphs indicate relative proportions in comparison between CP and BBD, with whichever higher being set as 1.0. Actual proportions of the higher value in CP or BBD were indicated in brackets, calculated as the relative sequence abundance within all the sequences that were assigned to SEED Subsystems terms. Asterisks, p < 0.001; n.s., non-significant; ppm, parts per million.