Potential adaptations of the microbiome to seasonal anoxia.
In the Labhra Cliff sponges, microbiomes were largely stable within individual species under different oxygen conditions, despite substantial changes in microbial populations in the water column. This stability suggests possible adaptations to deoxygenation within the microbiome. Similar stabilities of sponge microbiomes have also been noted in other studies where sponges have been exposed to sediment loading (
36), thermal stress (
74), and food shortage (
75). Sponge microbiomes are significantly disrupted only when sponges experience physiological stress such as bleaching (
76,
77), necrosis (
34,
78), disease (
79), or mortality (
36). Therefore, the stability in the microbiomes we observed in Labhra Cliff sponges might indicate that the holobionts were “healthy” and adapted to periods of anoxia.
Although microbiomes were mostly host species specific across the whole data set (
Fig. 7), some symbiont strategies were shared across sponge taxa in anoxia-tolerant species. In the four sponge species
Eurypon sp. 2,
H. stellifera,
Mycale sp., and
Raspaciona sp. (
Fig. 7), three symbiont combinations were identified. These combinations were characterized by their most abundant OTUs, as follows: combination i, OTU1 and OTU2; combination ii, OTU3 and OTU7; and combination iii, OTU3 and OTU10 (
Fig. 4 and
7). Combination i was exhibited by
Eurypon sp. 2 and
Mycale sp., and
H. stellifera and
Raspaciona sp. had combinations ii and iii, respectively. All combinations included high abundances of
Thaumarchaeota (either OTU1 or OTU3). Combinations i and ii are also dominated by large populations of the
Gammaproteobacteria OTU2 and OTU7, respectively. In addition to the
Thaumarchaeota OTU3, combination iii was characterized by high abundances of the unknown OTU10. A summary of the various combinations is shown in
Fig. 8 along with aerobic and anaerobic metabolic pathways that might be present in the holobionts, which are discussed in detail below.
While
Gammaproteobacteria and
Thaumarchaeota are common sponge symbionts and often cooccur in one host (
40), the combinations of these specific OTUs at their high relative abundances may be unique to anoxia-tolerant species (
Fig. 5 and
6). There was strong evidence of convergence toward combination i, since it was acquired only by
H. stellifera under anoxia and was shared across a large host phylogenetic distance, i.e., between the poecilosclerid
Mycale species and the axinellid
Eurypon sp. 2. Although many emergent properties of sponge microbial communities, e.g., community complexity and interactions, are conserved across Porifera, it is rare that specific OTUs are shared across large host phylogenetic distances (
40). This exceptional symbiont commonality as well as the acquisition of combination i by
H. stellifera indicated that this combination may be better adapted to seasonal anoxia than combination ii. Combination iii, conversely, may represent a strategy just as successful as combination i, given its stability in anoxia within
Raspaciona spp. (
Fig. 7).
Both
Thaumarcheota OTUs were part of the
Nitrosopumilaceae family, but OTU1 is present only in sponges, making it sponge specific, whereas OTU3 was also present in sediment and water samples, making it a generalist. The OTU3 is part of a clade that contains more free-living
Thaumarchaeota members, including
Nitrosopumilus maritimus (
Fig. 5), than symbionts. Conversely, OTU1 forms a clade that is almost exclusively sponge or coral associated (
Fig. 5). Based on the genomes of their close relatives, both OTU1 and OTU3 are likely AOA and could therefore oxidize sponge-derived ammonia, detoxifying the holobiont and potentially providing dissolved organic carbon (DOC) for the host that is ultimately sourced from chemolithotrophic carbon fixation (
80,
81), making them integral parts of holobiont metabolism under normoxia (
Fig. 8).
It has been reported that both ammonia oxidation rates and carbon fixation rates by an AOA symbiont are positively correlated within the sponge
Ianthella basta (
44). Similarly,
Thaumarchaeota are the main drivers of nitrification in four cold-water sponges (
41,
42). The AOA symbionts of a glass sponge living under mild hypoxia also possess elements of a facultatively anaerobic metabolism, including fermentation and fumarate, nitrite, and sulfite respiration (
19). Although other AOA within the
Thaumarchaeota do not generally include the aforementioned anaerobic elements, a terrestrial AOA does have the capacity for aromatic amino acid fermentation (
82), and it is possible that DOC transfer between symbiont and host continues via fermentation under anoxia (
Fig. 8). The microbes themselves could also be a food source for the sponge (
49,
83,
84).
Thus, it is possible that hypoxic environmental conditions are beneficial for the holobiont, given the low-oxygen requirements of sponges (
16,
17) and the high abundance of
N. maritimus in marine OMZs (
85). Accordingly, the relative abundance of OTU3 significantly increased in the water (but not in any sponge species) during hypoxia but was not significantly different between anoxia and normoxia. Despite this increased abundance in the environment under hypoxia, populations of OTU3 were not significantly increased in
Eurypon sp. 2 under the same conditions. Hypoxia, however, is not the “typical” condition between 25 and 30 m during the summer in Lough Hyne; instead, anoxia is typical in summer (
68). Assuming that the sponges are active and pumping (see “Potential adaptations of the sponge host to seasonal anoxia”) and given that ammonium oxidation requires oxygen in
Archaea (
48), holobiont metabolisms may be very different under anoxia. Furthermore,
Thaumarchaeota are functionally diverse (for examples, see reference
44), so the actual metabolisms and symbiotic functions of
Thaumarchaeota in Lough Hyne sponges need to be verified under their respective oxygen conditions.
Like
Thaumarcheota,
Gammaproteobacteria symbionts may contribute key functions to their holobionts, including some that provide tolerance to deoxygenation. Although no single
Gammaproteobacteria OTU occurred in relative abundances greater than 5.2% in
Raspaciona spp., symbiont combination i contained high relative abundances of the
Gammaproteobacteria OTU2 and symbiont combination ii contained relatively high abundances of OTU7. Unlike the
Thaumarchaeota, both OTU2 and OTU7 were sponge specific. Within the data set of focus taxa (
Fig. 4), the significant decrease in OTU7 in
H. stellifera during anoxia, compared to other oxygen conditions, may correspond to the appearance of OTU2, if both occupy the same niche. The same might be true of another gammaproteobacterium, OTU1075, which significantly increased in relative abundance in anoxia in
Eurypon sp. 2 and was more closely related to OTU2 than OTU7 (
Fig. 6).
The facultative anaerobe
Thioalkalispira microaerophila, which can use sulfide as an election donor and grows in micro-oxic conditions (
86), as well as
Thiohalophilis thiocyanatoxydans, which can grow anaerobically using thiosulfate as an electron donor and nitrite as an electron acceptor (
86), are in the same clade as the sponge-specific
Gammaproteobacteria (
Fig. 6). Therefore, it is possible that the sponge-specific
Gammaproteobacteria in our samples possessed both aerobic and anaerobic capacities and could remove exogenous, toxic sulfide from the holobiont under anoxia (
Fig. 8). This latter process, however, would be dependent on an electron acceptor, probably nitrite, that could come from the environment under nitrogenous conditions or through unknown pathways within the holobiont under sulfidic conditions.
In addition to the two most abundant OTUs in combination ii, a
Nitrospira (OTU17) was found in high relative abundances in
H. stellifera. Although it was absent or significantly less abundant in other Labhra Cliff sponges (
Fig. 7), OTU17 may perform important metabolic functions within the
H. stellifera holobiont. The OTU17 is closely related to a
Nitrospira (CcNi) that is associated with the sponge
Cymbastela concentrica, and even though some
Nitrospira spp. can completely oxidize ammonia to nitrate (commamox), OTU17 may only oxidize ammonium to nitrite, as was predicted for CcNi (
87). Although it was not significant, relative abundances of OTU17 decreased during anoxia in
H. stellifera, which could be due to a lack of oxygen inhibiting the metabolism and growth of
Nitrospira species.
Although
Nitrospira spp. are also conspicuously absent (
Fig. 3) (see also references
40 and
44) or inactive (
42) in some sponge holobionts in general, CcNi symbionts in
C. concentrica form close metabolic associations with the host and other microbes, including a member of the
Thaumarchaeota, namely, CcThau (
87). Coincidentally, CcThau is more closely related to OTU3 than to OTU1 (
Fig. 5), and therefore the cooccurrence of relatively large populations of OTU3 and OTU17 might indicate a coevolution between these two OTUs. The acquisition of OTU1 by
H. stellifera could therefore disrupt these partnerships under anoxia (
88), but this remains to be tested. It is also possible that OTU17, like some of its congenerics, could perform comammox and/or hydrogen oxidation coupled to sulfur reduction under anaerobic conditions, making it well adapted for low-oxygen stress (
89). In either case, the
H. stellifera holobiont likely employed a separate metabolic strategy under normoxia than under anoxia and compared to hosts with symbiont combination i or iii.
The holobionts from the genus
Raspaciona likely employ different metabolic strategies in response to anoxia. Like
H. stellifera, they host large, stable populations of OTU3, but
Raspaciona spp. do not acquire more OTU1 or any OTU2 populations under anoxia. Instead,
Raspaciona spp. harbored large, stable populations of the unidentified OTU10 (
Fig. 7), which was completely absent from all other samples except for one
Eurypon cf.
cinctum, taken under hypoxia. The unassigned OTU10 has an unknown metabolism; however, its stability in
Raspaciona spp. through anoxia and normoxia, and its presence in
Eurypon cf.
cinctum, indicates that it may confer some degree of deoxygenation tolerance (
Fig. 8).
Microbes associated with sulfate reduction and anammox were conspicuously absent or present in very low abundances in Labhra Cliff sponges. Probable sulfate-reducing OTUs were present in the anoxic water at much higher abundances than in any sponge species under the same conditions, but they were absent in all sponges under normoxic and hypoxic conditions (
Fig. 3C and
6). The
Planctomycetes as a phylum, which contains anammox bacteria, were present at low levels in all samples and decreased in relative abundances in anoxic water in
Eurypon sp. 2 and
H. stellifera (see
Fig. S6 in the supplemental material). In
Raspaciona spp., conversely, the relative abundance of this phylum increases in anoxia from 2.4% to 4.7% (
Fig. S6). The low signals of
Planctomycetes in the sponges, other than
Rapasciona spp., were likely contamination from microbes in the sponge water canal system and are usually bioinformatically filtered out of analyses of symbionts (for an example, see reference
90).
Curiously, both sulfate reduction and anammox bacteria have been confirmed in the holobiont
G. barretti, which experiences internal anoxia in its tissues (
49,
50). The difference between the microbial communities in
G. barretti and those in anoxia-tolerant species from Lough Hyne might be due to differences in morphology, as
G. barretti is a massive species, or environment, since it occurs more under constant oxygenation than under seasonal anoxia. Thus, the “anoxic microecosystems” observed in
G. barretti (
49) may result from its morphology, and the thin, encrusting sponges of Lough Hyne could be comparatively more oxygenated most of the year, even if pumping ceases (
91). Periods of pervasive oxygenation would restrict symbioses with obligate anaerobes and favor microbes with flexible metabolic strategies.
Notwithstanding these potential anaerobic processes, the three symbiont combinations outlined above do not universally confer hypoxic tolerance to sponges in general but may be necessary for full anoxic tolerance. For example,
Thaumarchaeota are effectively absent in the hypoxia-tolerant species
H. panicea (
92). Moreover,
H. panicea contains high abundances (>75%) of an alphaproteobacterium as does its congeneric species
Halichondria bowerbankii (
Fig. 7), and
Alphaproteobacteria were effectively absent from anoxia-tolerant Lough Hyne sponges. Although the microbial community within the hypoxia-tolerant
T. wilhelma has not yet been investigated in detail, only two bacterial genomes have been identified from genomic sequencing of
T. wilhelma, and both were likely
Alphaproteobacteria (
93). It is therefore unclear if
T. wilhelma contains
Thaumarcheota OTUs in high abundance, although its congeneric species
T. citrina does (
Fig. 7). An alphaproteobacterium cultured from the Lough Hyne sponge
Axinella dissimilis was able to grow anaerobically via fermentation and denitrification (
94). Therefore, many microbiome structures may be capable of coping with hypoxia. Nevertheless, neither the hypoxia-tolerant
H. panicea (
17) nor
T. wilhelma (
16) tolerates prolonged anoxia, so the Labhra Cliff holobionts may be uniquely tolerant to anoxia even in comparison to other poriferans. These sponge hosts may also be adapted to tolerate deoxygenation directly, or the ability to survive anoxia may depend on metabolic shutdown of either the host, symbionts, or both.
Potential adaptations of the sponge host to seasonal anoxia.
It is unlikely that Lough Hyne sponges die off
en masse during anoxia and recolonize the area during normoxia, because no dead tissue or discoloration was present around anoxic sponges. It is possible that these sponges decrease or cease metabolic activity during anoxia (
25,
95), but this requires further investigation. For some marine invertebrates, environmental anoxia can trigger a switch to a fermentation-based metabolism, which results in a considerably decreased metabolic rate; however, the by-products of fermentation still need to be eliminated into the environment (reviewed in reference
95). If this elimination cannot be achieved by diffusion alone, it is possible that the sponges continue to pump during anoxia, albeit at a potentially decreased rate, and thereby still provide dissolved organics, ammonia, CO
2, and other metabolic products to their symbionts (
Fig. 8).
Nonetheless, Labhra Cliff sponges were definitely pumping under hypoxic conditions (
Fig. 1A), which is consistent with normal transcription activity in
T. wilhelma under hypoxia (
16) and with observations of sponges inhabiting consistently hypoxic environments (
96,
97). Pumping was unsurprising considering that mobile fish and crabs were also observed under hypoxia at Lough Hyne (B. W. Strehlow, personal observation), and oxygen levels were above lethal and sublethal thresholds of many fish and invertebrates (
12). Nevertheless, in a separate study, a single individual of
G. barretti drastically reduced its pumping rate following
ex situ oxygen depletion (20% air saturation) (
98), and feeding rates in one individual (out of three) of
H. panicea were reduced in low-oxygen concentrations (3% air saturation) (
17). The sublethal, physiological impacts of deoxygenation, therefore, also need to be considered in the future.
Despite deoxygenation tolerance in some species, sponge diversity and abundance are overall likely limited by seasonal anoxia in Lough Hyne, and even
Eurypon species and
H. stellifera decreased significantly in abundance below the thermocline (
20). Growth and reproduction may be impacted by seasonal anoxia because collagen synthesis is oxygen dependent; however, it is still possible in very low oxygen concentrations (see reference
99). Sponge larvae may also require elevated oxygen due to their motility. Also, elevated oxygen may be needed during settlement and early development in sponges; however, the specific oxygen requirements for these life stages remain unknown. Moreover, a combination of factors could restrict settlement and growth to fewer species. Although sedimentation rates in the anoxic region are equivalent to that of other sites in Lough Hyne (
100), the combination of sediment and anoxic stress might restrict sponge distributions. Indeed, most encrusting species in Lough Hyne are more abundant on vertical than inclined (40°) surfaces due to high sedimentation on the latter (
20). Additionally, seasonally decreased metabolic activity in the holobiont could limit growth or the production of secondary metabolites under anoxia, leading to increased spongivory in normoxic months when mobile predators return.
It is also possible that, like the sponges, some or all of the microbiome becomes dormant under anoxia. The capacity to become dormant is common and phylogenetically widespread in microorganisms (reviewed in reference
101) and may occur in response to environmental stress including hypoxia (
102). For instance, pelagic
Thaumarchaeota are present in sulfidic (anoxic) zones, but they exhibit lower expression levels of genes involved in ammonia oxidation and may be inactive (
103). However, as stated in the previous section,
Thaumarchaeota in symbiosis with sponges may possess elements of anaerobic metabolism (
19) that could aid the sponge holobionts under anoxia if they are active. The major
Gammaproteobacteria OTUs could similarly be inactive during anoxia in Lough Hyne, although there is also strong evidence of anaerobic metabolisms within the clade formed by these OTUs within the anoxia-tolerant sponge holobionts. Moreover, even dormant microbes require some maintenance of proton motive force (
102) or DNA repair (
104). So, there may be some activity within a dormant microbiome under anoxia, which could perhaps be linked to the host’s decreased pumping rate suggested above for waste elimination. Dormancy could also leave the holobiont vulnerable to predation by protists, but the dynamics of this potential interaction are unknown and warrant further study. If microbial dormancy occurs, it could still convey some resilience to the holobiont, stabilizing the microbiome under deoxygenation stress and allowing for rapid recovery following reoxygenation. However, this type of dormancy-recovery dynamic might be best suited for surviving seasonal anoxia, e.g., in enclosed or eutrophicated systems, whereas anaerobic metabolisms would be necessary for long-term survival in the functionally anoxic core of permanent OMZs.
Implications for early animal evolution and future oceans.
These results have important implications for early animal evolution and the state of future oceans. Under the lower oxygen concentrations of the Neoproterozoic Era, early metazoans likely tolerated both widespread hypoxia and transient anoxia (
105–107). The Labhra Cliff sponges demonstrated that this ancient tolerance is possible in sponges and that it could have involved the microbiome, but how analogous are these holobionts to early metazoans? The ancestral sponge evolved in a microbial world and may have consequently formed close associations with many symbionts like modern sponges. Modern symbioses with both
Gammaproteobacteria and
Thaumarchaeota were present in all 81 species assessed by Thomas et al. (
40). It is therefore conceivable that the ancestral sponge contained either
Gammaproteobacteria,
Thaumarchaeota, both, or their respective ancestral forms. The loss or decreased abundances in these symbiont groups in sponge lineages that evolved into our “outgroup” samples (
Fig. 7) could then correspond with the absence of this group in the seasonally anoxic site. The first metazoans could therefore have been very similar to the Labhra Cliff sponges in their symbiont composition and morphology, and even if it is only through convergent evolution, the seasonally anoxic sponges of Lough Hyne are an important model system for studying early animal evolution.
Considering the past also yields clues about the future. Some previous mass extinction events were probably driven by acidification, warming, and deoxygenation of the oceans following extensive volcanism (
12,
108–111). Similar stressors are facing modern oceans as a result of anthropogenic CO
2 release. These emissions cause ocean warming, acidification, and the expansion of OMZs, and local deoxygenation can also be caused by anthropogenic nutrient runoff (
1–5). For organisms like sponges and corals, benthic anoxia could stress adult life stages, while pelagic deoxygenation could threaten larval survival and distribution.
However, given the anoxic tolerances observed in sponges in the present study, could sponges outcompete corals in future scenarios (for examples, see references
112 and
113)? Sponge abundance has recently increased on some coral reefs, due in part to a decrease in coral abundance (
114–116), but the future may be more nuanced. Recent experiments show that the necrosis and bleaching caused by thermal stress was ameliorated by increased CO
2 in two phototrophic sponge species under scenarios equivalent to the worst-case warming predictions, i.e., RCP8.5 (representative concentration pathway) (
117), but necrosis was exacerbated by these two stressors in two heterotrophic species (
118). Moreover, not all phototrophic sponges have this advantage, since at least one species died under these conditions (
119), and most experiments do not include oxygen as a factor, which is predicted to decrease by up to 3.7% under RCP8.5 (
120).
According to a meta-analysis across marine benthic organisms, the combination of thermal stress and deoxygenation reduced survival times by 74%, compared to each stressor in isolation, and increased the lethal concentration of oxygen by 16% on average (
121). As with increased temperature and CO
2, responses to deoxygenation are likely species specific, as suggested by this study. Some coral (
122,
123) and sponge (
11) species may be tolerant to deoxygenation, while others are not, and if anoxic tolerance in sponges is limited to seasonal exposure, the expansion of permanent OMZs could still cause mortality. The effects of the combination of all three stressors, however, are virtually unknown and thus require extensive research in the future.