Biological markers, which are measurable indicators of a pathogenic process or response to an exposure or intervention [1, 2], can provide rapid and cost-effective means to inform decision-making regarding subsequent interventions. Biomarkers can be diagnostic when used to provide information on a current condition, or predictive if they are correlated with a future outcome. Further, biomarkers can be utilized in both dynamic and static manners. A dynamic biomarker continuously incorporates information about changes in risk factors by repeatedly taking measurements of an indicator in the same individual over time [3, 4], while a static biomarker is the result of a single baseline measurement of a given biological indicator [4]. Biomarkers have the potential to act in these diagnostic and predictive roles simultaneously by diagnosing whether an individual is diseased or healthy, and predicting future health outcomes based on that condition. One such example is how healthcare providers use blood sugar as an indicator of type 2 diabetes [5]. In a diagnostic marker context, a diabetes diagnosis is based on whether blood sugar levels exceed a predetermined threshold. In a predictive context, if an individual’s blood sugar is below the threshold, but elevated relative to a known baseline, that patient could be at risk for diabetes (prediabetes) and the healthcare provider would likely recommend lifestyle changes, such as increased exercise or dietary changes [6]. Similarly, blood sugar could be considered a dynamic biomarker for diabetes, in which the blood sugar of individuals diagnosed with or at-risk of diabetes is continuously monitored through time. Whereas a static biomarker would be the use of marker genes to determine an individual’s genetic risk for diabetes [7]. Although biomarkers are most commonly used in medical contexts, they are being increasingly applied in conservation and restoration decision making [8–11].

Coral reefs are in rapid decline in response to warming waters across the world [12], and consequently, coral restoration efforts have increased in scale, highlighting the need for evidence-based coral reef restoration informed by science [13, 14]. Biomarkers have received significant attention given their potential application in guiding coral restoration decision making regarding which species or genets should be outplanted, and understanding when they may be in a state of physiological stress [8, 11]. In particular, a number of studies highlight the potential of gene expression as a dynamic biomarker in the coral restoration context. For example, Kenkel et al. [15]) developed quantitative PCR (qPCR)-based diagnostic gene expression biomarkers for heat stress of the stony coral Porites astreoides where expression of heat shock protein genes were upregulated in response to heat stress, and scaled with levels of temperature stress. Later, Bay and Palumbi [16]) found higher expression of a cluster of genes related to heat stress response (such as Caspase, Rab- and Ras-related proteins) was correlated with lower survival in Acropora hyacinthus. Wright et al. [17]) exposed fragments of Acropora millepora to Vibrio owensii, a putative pathogenic bacteria, and found that corals with higher mortality rates possessed higher baseline antioxidant or cytotoxic activities prior to the immune challenge. Additionally, they found that survival could be accurately predicted with baseline expression of only two genes, as a subsequent validation experiment used these genes to identify corals which succumbed to disease with a 73% accuracy [17].

Although gene expression biomarkers have received the greatest attention to date [8], there has been similar work to test whether aspects of a coral’s microbiome can be diagnostic or predictive of disease, heat tolerance, and general reef health [17–22]. For example, Candidatus Aquarickettsia rohweri is an abundant alphaproteobacterium in the branching coral Acropora cervicornis [23, 24]. Increased abundances of A. rohweri have been observed in response to nutrient enrichment with subsequent growth trade-offs [25], as well as in A. cervicornis genets susceptible to white band disease [20]. Interestingly, white band susceptible genets also experienced a drastic reduction in the abundance of A. rohweri in response to bleaching [20]. Thus, it is possible that the abundance of A. rowheri could be a predictor of disease risk, as well as a diagnostic biomarker of bleaching severity. However, due to the transient nature of the coral microbiome and stochastic compositional changes in response to stressors [26], identification of specific microbial indicators can be challenging and questions remain as to their utility [8].

Under the assumption that performance metrics like growth and thermal tolerance are fixed properties of a genet [27], genomic biomarkers may be the most promising as a static predictive biomarker [8, 11]. Prioritization of resilient individuals to maximize the success of coral reef restoration outcomes remains a top priority [13]; the most common way to do this is asexually, in which genets that appear to be the most resilient are propagated, or sexual reproduction (e.g. selective breeding) of resilient genets, which has increased in popularity more recently [8, 28–30]. A genomic indicator (e.g. identification of loci associated with resilience) could facilitate rapid identification of resilient broodstock, bypassing the need to undertake more time-consuming experimentation and phenotyping [31]. However, recent studies have identified genotype by environment interactions in key performance traits, in which a genet’s bleaching, growth form, and survival vary by site [32, 33], indicating that performance is not necessarily genetically “hard-coded.” Thus, a dynamic gene expression or microbial biomarker that could be deployed in a cost effective and environment-informed manner could be a useful tool to assess an individual’s current physiological state, and assist in determining the readiness of an individual coral for ecosystem restoration at any given point in time.

Here, we build on results from a multi-genet, multi-site transplant experiment done in 2018 [33, 34] by examining whether gene expression and/or microbial community composition prior to outplanting (T0) can be predictive of the future performance of A. cervicornis following outplanting to restoration sites in the lower Florida Keys, FL, USA. Performance metrics included growth, survival, and “restoration value” at twelve months (T12) post outplanting. Here, we define restoration value as the sum of the interstitial space (space in between branches) (Fig. 1) created by each surviving coral genet at T12, a measure analogous to the ‘habitat production’ metric defined by Ladd et al. [35]. This metric gives weight not only to survival, but also to growth and branch complexity – two traits that vary independently [33]. Additionally, it is important to note that while growth and survival are important metrics in terms of the ecological impacts of restoration, they do not capture the full range of ecosystem-level impacts nor the social and cultural value restoration programs also provide [36]. We also examined whether T0 expression modules were preserved through T12 in order to understand if gene expression pathways were perturbed after transplantation to a different environment. Lastly, to further elucidate whether genotype survival rankings are dynamic or fixed, we compared survival rankings from the initial transplant to a more recent deployment of the same coral genets to a subset of the same field sites.

We found that host gene expression in the nursery is correlated with future performance: genets that ultimately exhibited low restoration value upregulated key immune genes in the nursery, and continued to upregulate them after twelve months, potentially indicating a chronic condition that could both predict and explain subsequent performance differences in the field. Further, survival rankings changed between the initial experiment and a more recent deployment of the same coral genets, indicating that survival may not be fixed, but rather related to the condition of a coral when outplanted, including direct impacts or interactions with local biotic and/or abiotic environmental conditions.

Using WGCNA’s modulePreservation function, modules identified in the T0 samples were imposed onto T12 samples to determine the degree to which expression patterns were maintained [37]. We observed high levels of module preservation, defined as a Zsummary score of > 10 [37]. All modules of interest were above this threshold, with the pink module Zsummary = 33.15, tan = 28, greenyellow = 25.22, and magenta = 16.91, similar to the majority of other modules (Table S5). In terms of the magnitude and direction of expression, the magenta and tan modules exhibited similar expression patterns to T0 at T12, with high restoration value genets continuing to downregulate the magenta module eigengene, and low restoration value genets upregulating the magenta module eigengene (Fig. 2A, E). Genets 41, 62, and 13 (bottom 50% percentile restoration value) continued to significantly upregulate the magenta module eigengene compared to genets 50 and 3 (Tukey’s HSD < 0.0005, Table S6, Fig. 2E). Additionally, the worst surviving genet (G41) also continued to strongly upregulate the tan module eigengenes twelve months later (Tukey’s HSD < 0.0005, Table S6, Fig. 2B, F). Following twelve months post-outplant, all genets were expressing the pink and greenyellow modules at a similar level (ANOVA, p = 0.0.467, 0.282, respectively) losing the pattern of upregulation of expression by high restoration value genets and downregulation of expression by low restoration value genets that was observed at T0 (Fig. 2C, D, G, H), despite high levels of module preservation.

Fisher’s Exact Tests showed 36 significantly enriched gene ontology (GO) terms among annotated isogroups in the magenta module (10% False Discovery Rate (FDR), Fig. S3), while 127 terms were significantly enriched among isogroups in the tan module (10% FDR, Fig. 3A, Fig. S4). The anticorrelated modules, pink and greenyellow, were enriched for 10 and 19 GO terms respectively (10% FDR, Fig. 3A, Figure S5-6). An additional Mann-Whitney rank-based enrichment test, which gives weight to the strength of an isogroup’s module membership (kME) in addition to its module assignment, showed 29 significantly enriched GO terms for the T0 magenta module (Fig. 3B). The top enrichments for ‘biological process’ GO terms in the magenta module were cell-cell adhesion (GO:0098609, p < 0.0005), biological adhesion (GO:0007155;GO:0022610, p < 0.0005), and cellular response to gamma radiation (GO:0071480, p < 0.0005, Fig. 3B, Fig. S3). Some of the top tan module parent term enrichments included terms like response to stimulus (GO:0050896, p < 0.0005), immune system process (GO: 0002376, p < 0.0005), cellular process (GO:0009987, p < 0.0005), and regulation of macromolecule biosynthetic process (GO:0010556, p < 0.0005, Fig. 3A, Fig. S4) in addition to “regulation of receptor signaling pathway via STAT” (GO:1904893, p < 0.0005, Fig. 3A, Fig. S4).

The anticorrelatory modules were generally enriched for metabolic and developmental GO terms. Specifically, the greenyellow module was enriched for GO terms related to metabolic process (GO:0008152, p < 0.0005), regulation of biological process (GO:0050789, p < 0.0005), viral genome replication (GO:0019079, P < 0.0005), immune system process (GO:0002376, p < 0.0005) and weakly enriched for response to stimulus (GO:0050896, p = 0.0416, Fig. 3A, Fig. S5). Lastly, the pink module was strongly enriched for several GO terms related to metabolism and mitochondrial function, including protein localization to mitochondrion (GO:0072655;GO:0070585, p < 0.0005), mitochondrial transport (GO: 0006839, p = 0.01), protein-containing complex assembly (GO:0065003, p < 0.0005), and metabolic process (GO:0008152, p = 0.0167, Fig. 3A, Fig. S6).

We further investigated differential regulation of specific isogroups among module-specific ontology enrichments, in addition to hub genes within each module with the aim of identifying candidate biomarkers predictive of future restoration value that were strongly and consistently regulated among ramets of a genet. Hub genes, or genes with many interactions with other genes [38], were identified for the four primary modules of interest (magenta, tan, pink, and greenyellow). The magenta module hub gene was annotated as interferon regulatory factor 2 (isogroupDN72717_c4_g1), and the tan module hub gene was annotated as guanylate-binding protein 7 (isogroupDN53976_c1_g1), while the pink (isogroupDN4562_c0_g1) and greenyellow (isogroupDN68057_c0_g1) module hub genes were unannotated.

Criteria for determining which specific isogroups could be utilized as biomarkers included isogroups that were annotated, had a high kME (90% percentile and above), and had significant differential expression patterns between high and low restoration value genets. Five isogroups were identified that fit these parameters (Table S5-9, Fig. 3C): Interferon regulatory factor 2 (isogroupDN72717_c4_g1), E-selectin (isogroupDN74671_c2_g1), tyrosine protein kinase receptor tie-1 (isogroupDN74674_c1_g3), and allograft inflammatory factor (isogroupDN57704_c0_g1). Interferon regulatory factor 2 (isogroupDN72717_c4_g1) showed significantly higher expression between the three lowest restoration value genets (G41, G62, and G13), and the three highest restoration value genets (G50, G3, G1, Tukey HSD, p < 0.05, Table S8-9, Fig. 3C). Further, genets 13 and 41 were significantly different from all genets in the top 50% of restoration value genets (Tukey HSD, p < 0.05, Table S8-9, Fig. 3C). E-selectin was significantly upregulated in bottom 50% restoration value ( G36, G41, G62, and G13), compared to the top 50% performers (G50, G3, G1, G31, G44) (Tukey HSD, p < 0.0005, Table S10, Fig. 3C). Additionally, allograft inflammatory factor 1 was differentially expressed between genets, with the three lowest restoration value genets ( G41, G62, G13) significantly upregulating the gene compared to the top three restoration value genets (G50, G3, G1, Tukey HSD, p < 0.005. Table S11). Similarly, tyrosine protein kinase receptor tie-1 (isogroupDN74674_c1_g3) was also significantly upregulated in the three lowest performing genets compared to the three top performing genets (G50, G3, G1, Tukey HSD, p < 0.005, Table S12).

Lastly, in the tan module, three isogroups were significantly enriched for the GO term “negative regulation of receptor signaling pathway via JAK-STAT”: interferon regulatory factor 1 (isogroupDN73115_c4_g1), guanylate-binding protein 7 (isogroupDN53976_c1_g1), and protein mono-ADP-ribosyltransferase (isogroupDN72552_c1_g1). Notably, genet 41, the worst survivor in 2018, showed strong signatures of upregulation for all three genes relative to other genets in the top 50% percentile of survival (Tukey HSD < 0.05, Table S12-14, Fig. S7).

We reconfirmed the findings of [34], which identified Synechococcus CC9902 and MD3-55, a Rickettsiales symbiont (sensu Candidatus A. rohweri) as the two most abundant microbial genera in nursery ramets which also varied among genets. Overall, the top five most abundant microbial genera in T0 ramets were Synechococcus CC9902, MD3-55, NS5 marine group, Candidatus Actinomarina, and HIMB11 [34]. There were no clear patterns between the relative abundances of these genera and future genet survival or growth (Fig. S8). Similarly, there were no clear patterns between microbial abundances of the top 100 most abundant phyla nor between the abundance of the two most abundant microbes (MD3-55 and Synechococcus CC902) and genet ordered by restoration value ranking (Fig. S9, S10).

Overall alpha diversity (Shannon’s) of the microbiome differed among genets; however, a post-hoc Tukey’s HSD test revealed limited pairwise differences between genets in the top 50% of restoration value, and the bottom 50%. Specifically, both genets 13 and 62 (lower restoration value) had significantly higher alpha diversity from genets 31 and 1 (higher restoration value, Table S16). It is worth noting that when comparing the bottom and top 50% genets, alpha diversity in genet 41 (the worst survivor) was only significantly higher than in genet 1 – the second best survivor – but not significantly different from any other top performing genet (Tukey’s HSD = 0.049, Table S16). Differences in beta diversity among genets were visualized by principal coordinate analysis (PCoA). No clustering was observed with respect to genet restoration value or survival ranking, although the adonis PERMANOVA indicated that overall, genets significantly differed in beta diversity (p = 0.001, Fig. S11, Table S17). Further, the pairwise comparisons revealed that for genets that differed from each other, statistical significance was weak (p = 0.045) but various pairwise differences existed for genets regardless of restoration value (Table S17).

Genets of A. cervicornis that ultimately had lower restoration value upregulated key immune genes in the nursery prior to being outplanted to nine different sites across the Florida Keys, despite an outward appearance of health (i.e. no signs of tissue sloughing, bleaching or disease). These genes remained upregulated within low restoration value individuals that were still alive 12 months later at their respective outplant sites (T12), indicating these coral may have been suffering from a visibly undetectable health condition which could be the result of a chronic infection or attributable to a genetic condition. A subsequent outplant experiment [39] using the same ten genets at a subset of two of the same sites showed that the relative “winners” and “losers” changed: genets that were poor survivors in 2018 climbed the ranks as high survivors in 2022, and vice versa. If a temporary health condition was the reason for poor survival, either directly or indirectly through increased susceptibility to biotic or abiotic factors, it is possible that the condition was resolved by 2022. However, no gene expression data are available for the 2022 experiment, thus, further biomarker validation experiments are necessary to determine causative links.

We did not find evidence for strong correlations between initial host microbiome, such as the abundance of A. rohweri, and future restoration success, suggesting that if the condition that certain genets were suffering from was a chronic infection, it was not of bacterial origin. Overall, these results suggest that gene expression in the nursery prior to outplanting may diagnose health state and could potentially inform restoration success in the future, supporting their possible application as a dynamic biomarker. However, further validation studies are necessary in order to determine how consistently these genes can predict future restoration value. With additional validation, field trials and development of protocols [8], such a biomarker could be deployed by restoration practitioners to vet genets prior to outplanting as part of a routine “coral health check” to help determine which coral to outplant, and when to outplant them to maximize restoration success.

Corals have a robust innate immune system which plays an essential role in symbiotic homeostasis and their response to environmental perturbations such as climate change and disease [40–42]. Elevated expression of immune genes in subsequently poor-performing genets suggests they were immunocompromised prior to outplanting. It is important to note that ramets were only outplanted in the 2018 experiment if they appeared visually healthy, mirroring the regulatory requirement for coral restoration programs in Florida [43]. Thus, gene expression was the only indicator of a potentially immunocompromised state and may hint at an underlying causative agent. Interferon regulatory factor 2 (IRF2), tyrosine protein kinase receptor tie-1, E-selectin, and allograft inflammatory factor-1 were strongly and consistently upregulated in low restoration value genets in the nursery prior to outplanting, as well as after one year at nine different sites. These genes have been repeatedly implicated in response to pathogens in coral and other organisms.

IRF2 is part of the interferon family of cytokines and can be expressed in response to viruses [44]. In vertebrates, the IRF cytokine family modulates multiple parts of the immune system [45], and can respond to viral infections [44, 46]. In coral, cytokine expression is triggered by the Toll-like receptor (TLR)-to-NF-kb signaling pathway, a highly conserved immune pathway that responds to microbe associated molecular patterns (MAMPs) [40, 47]. Importantly, use of single cell RNA sequencing technology in the stony coral Stylophora pistillata revealed that the coral had two distinct cell types with signatures of immune function – both of which expressed IRF2 and IRF1, indicating that these genes are central parts of the coral immune system [48]. Further, Kozlovski et al. [49] found that IRF2 was upregulated in immune cells both at a basal state, and in response to poly(I:C) in embryos of the sea anemone Nematostella vectensis. Lastly, van de Water et al. [50] found that when the coral Montipora aequituberculata was exposed to heat stress, both IRF1 and 2 were differentially expressed, despite the corals appearing visually healthy. Further, they saw no change in the microbial communities of heat stressed corals [50]. Here, expression of IRF2 could point towards an elevated antiviral response in low restoration value genets as we did not find any specific bacterial signatures associated with coral performance metrics. Viruses are ubiquitous in reef building corals [51, 52], especially in the coral host mucus [53], and impact coral host health [54, 55]. Interestingly, IRF2 (isogroupDN72717_c4_g1) was the hub gene for the magenta module, indicating high connectivity with other genes in the magenta module [38], and thus, is likely to have functional similarity with other genes in the module. Further work is needed to determine whether these corals were afflicted by a viral infection, or if IRF2 could be indicative of a general immune response to an environmental stressor.

Expression of tyrosine protein kinase receptor tie-1 (PTK) was also upregulated in low restoration value genets. PTKs are proinflammatory cytokine promoters, an essential function of the innate immune system [56]. Previous work in the mountainous star coral coral Orbicella faveolata identified significant expression of PTK in response to immune challenge via exposure to bacterial lipopolysaccharides (LPS) [57]. Similarly, expression of PTK was also increased in corals exposed to white plague disease, and was proposed to be a mediator of the development of disease lesions [58]. It is worth noting that the tan module, which was also strongly upregulated by our worst performing genet, was significantly enriched for the GO terms: “negative regulation of receptor signaling pathway via JAK-STAT,” as well as “regulation of receptor signaling pathway via STAT” (p < 0.005) (Fig. S4). Janus kinases (JAKS) are a type of PTK [59], and also play an important role in cytokine signalling [60]. In fact, evidence for JAKs in cytokine signalling emerged from work that revealed that the JAK family is essential for interferon signaling [61–63]. Specific isogroups annotated with these ontology terms include guanylate binding protein 7 (GBP7) as well as interferon regulatory factor 1 (IRF1, Fig. S7). Similar to IRF2, IRF1 is also a regulator of the interferon cytokine system and responds to viruses [64]. Interestingly, in vertebrates, IRF2 is believed to act as an antagonist to IRF1, as they compete for the same promoter elements of type I and II interferon inducible genes – thus, it has been suggested that IRF2 suppresses IRF1 function in some contexts [44, 65]. As noted in [48], IRF1, along with IRF2, is expressed in an immune cell type in a stony coral species. Emery et al. [66] also found IRF1 to be upregulated in Cassiopea xamachana, a facultatively symbiotic jellyfish, in response to exposure to the cnidarian pathogen S. marcescens. Further, Additionally, GBP7 has previously been found to be upregulated in response to a viral infection in vertebrates, and can suppress innate immunity via the NF-kB and JAK-STAT pathways, leading to reduced interferon and cytokine expression [67]. In larvae of the Indo-Pacific reef building coral Acropora digitifera experimentally infected with an algal symbiont, GBP7 was downregulated, indicating a suppression of the immune response post symbiont infection, recognized as a non-self particle [68]. In contrast, the sea anemone Exaiptasia pallida upregulated GBP5 in response to starvation, interpreted as an upregulation of the immune system [69]. Taken together, the upregulation of these genes by the worst surviving genet may implicate a role in the response to a putative asymptomatic (i.e., not visible to the eye) infection, in addition to genes related to cytokine production in the magenta module. The final two magenta module genes that showed significantly upregulated expression in low restoration value genets were E-selectin and Allograft inflammatory factor 1 (AIF-1), which have previously been characterized as calcium binding proteins in cnidarians [70, 71]. However, previous work has proposed that coral biomineralization and immunity are biological functions that are tightly linked [72–74]. Although their roles are poorly understood in invertebrates, it is possible that their upregulation may be a part of or a response to the cytokine production/IRF signalling indicated by elevated expression of IRF2 and PTK in the magenta module, and IRF1 in the tan module. Thus, these two Ca² + genes may play a role in both calcification and immune response, similar to work that has been found in other invertebrates [73, 75, 76]. E-selectin, a Ca² + dependent lectin, is involved in cell-adhesion, immunity, and likely calcium signalling [70, 77]. E-selectin is upregulated in response to bacterial LPS and interferon-γ – which is a gene product of IRF2 – in cell cultures of the human cerebral endothelium [78]. Notably, interferon-γ is believed to induce an antiviral state [79], and is induced by inflammatory cytokines [80]. In corals, E-selectin was found to be upregulated in bases of colonies of Acropora colonies, implicating a role in calcium signalling [70]. AIF-1 is a cytokine, and an interferon inducible Ca² + binding protein [81, 82]. While there is similarly limited knowledge on AIF-1 in invertebrates, the sea anemone Anemonia viridis has been found to upregulate AIF-1 in response to environmental stressors including warming temperatures, heavy metal exposure, and a bacterial immune challenge [71]. Lastly, it has been found to be upregulated in larvae of O. faveolata exposed to high UV levels [83]. The interferon activity by E-selectin and AIF-1 further points towards their tight connectivity to the magenta module’s hub gene, IRF2, an interferon, as well as to their potential function in response to viral activity [46, 84].

Overall, upregulation of this suite of immune genes in low restoration value genets prior to poor growth and survival in the field indicates that fragments of A. cervicornis may have been immunocompromised or undergoing a cellular stress response prior to outplanting, despite appearing visually healthy. Given the role of the genes highlighted above in antiviral activity, and the lack of relationship between microbial abundance and a genet’s restoration value (Fig. S5, S6), it is possible that the immune system was responding to a viral infection – which could have manifested asymptomatically [85, 86] – rather than a particular pathogenic bacterial infection. Corals have a diverse virome which can shift in response to environmental stressors, either by infecting the coral hosts or their microbiota [51, 52]. Low restoration value genets continued to upregulate these putatively antiviral genes after twelve months in different environments (Fig. 3) – if these genes were in fact a response to a viral infection, this could have been a factor contributing to poor restoration value of immunologically active genotypes. Million et al. [33] determined that abiotic environmental parameters between outplant sites were not significantly different; thus, it is unlikely that hosts were suffering from an abiotic stress, such as temperature. Another possibility is that specific genets were responding to a biotic stressor such as reduced food availability or increased predation, and was mounting a general, non-specific immune response to this stressor that was not resolved after being transplanted to a new environment. However, we would expect to see visual evidence of predation pressure in the form of feeding scars, and a lack of heterotrophic food sources should have impacted all coral genets similarly as they were reared in a common garden nursery prior to outplanting. Another possibility for the continued upregulation of these genes could be that some genets have a higher baseline expression of immune genes. Young et al. [87] found that baseline immune gene expression in Acropora palmata was the main driver of variation in expression between genets. Interestingly, they found that season also played a role in driving immune expression. Due to the lack of gene expression data from our subsequent 2022 experiment, it is difficult to distinguish whether low restoration value genets were suffering from a subclinical infection in the 2018 outplant study or simply have a genetic basis for chronically elevated expression of immune genes. Further work elucidating the gene expression patterns of these same genotypes in different seasons across different years is necessary to determine whether the gene expression patterns found here are the result of a temporary condition or variation in baseline immune expression.

Here, we found that neither microbiome structure nor abundance of MD3-55 (Ca. Aquarickettsia rohweri) were correlated with future restoration success of a genet. This is in contrast to previous studies, which have determined that abundances of MD3-55 are associated with genet susceptibility to environmental stressors. For example, [20] found that disease susceptible but visually healthy genets of A. cervicornis harbored 36-fold higher relative abundances of A. rohweri compared to disease resistant genets. In contrast, [88] reported that A. cervicornis genets that harbored high abundances of MD3-55 had high survivorship in response to nutrient enrichment. A. rohweri abundance in A. cervicornis significantly increased in response to phosphate enrichment, however, there was no significant increase in response to nitrate enrichment [89]. No change in MD3-55 abundance was observed after field transplantation in this experiment [34], and we see no differences in MD3-55 abundance among genets based on their subsequent restoration value. It is possible that MD3-55 as an indicator taxa may be stress-type specific, where its predictive ability is limited to specific types of acute stressors, but not in response to an ambient environment as we investigate here.

Overall, [34] found that the host-specific epibiomes investigated for their correlation with holobiont performance here did not change after being outplanted to the nine sites, indicating microbiome stability despite environmental changes. Microbiome stability is often species-specific [47, 90, 91], with A. cervicornis exhibiting a particularly stable microbiome despite environmental perturbations or site transplantation [92]. For example, [92] exposed three Caribbean corals, including A. cervicornis, to a multi-spectrum antibiotic disturbance prior to outplanting them to a field site. They found that the A. cervicornis host microbiome exhibited the same high degree of microbiome stability in both perturbed and non-perturbed outplants. Similarly, [91] found that A. cervicornis microbial communities in the coral mucus did not change after infection with disease. Whether microbiome composition can be used as a biomarker for future restoration success, especially in A. cervicornis, remains unclear, as the apparently high degree of stability in response to perturbations may not allow for the diagnosis of distinct microbial signatures of “healthy” and “stressed” individuals. However, it is possible that this stability is context and stressor specific; thus, further investigation into the role of the microbiome as a biomarker for A. cervicornis may be warranted, especially in the context of restoration operations.

Here, we tested the utility of gene expression patterns to diagnose the health state and predict restoration success of A. cervicornis. We identify five genes which could serve as putative biomarkers for restoration success: visually healthy genets with subsequent low restoration value resulting from low growth and/or survival were strongly upregulating these genes in the nursery prior to being outplanted, and some continued to upregulate the genes after twelve months in the field. In tandem, WGCNA co-expression modules were highly conserved after one year, indicating the biological pathways that were expressed in the nursery were not perturbed, despite being outplanted to nine different sites. Upregulation of these immune genes prior to outplanting not only provides us with information about the health state of the coral, but also provides a better understanding of what may be afflicting the coral health state. Here, it appears that visually healthy corals may have been responding to an asymptomatic condition by upregulating genes involved in IRF pathways and cytokine production. While chronic conditions can be driven by underlying genetic differences, differences in genet performance rankings in a 2022 outplant, combined with the context dependent genotypic success described by [96] suggests that a genet’s survival ability is likely not “hard-coded,” but rather, transient, dynamic, and partially dependent on the health state of a coral individual. We advocate for further investigation to understand the role of these four putative biomarkers in coral response to other abiotic and biotic stressors, and whether these genes are only upregulated as stress responses when individuals are apparently healthy, or if similar patterns persist when under environmental stressors. In addition, validation of a dynamic biomarker would allow restoration practitioners to assess the health of a coral prior to outplanting, enhancing restoration yield by facilitating the process of determining which individuals are in top condition for restoration cohorts.