Differential gene expression indicates modulated responses to chronic and intermittent hypoxia in corallivorous fireworms (Hermodice carunculata)

Climate models predict an increase in extent, frequency, and duration of marine hypoxia events in the twenty first century. A better understanding of organismal responses to hypoxia in individual species is a crucial step for predicting ecosystem responses. We experimentally subjected a common invertebrate, the bearded fireworm (Hermodice carunculata) to two levels of chronic hypoxia and, in a separate experiment, to intermittent hypoxia. We found components of the conserved hypoxia-inducible factor (HIF) pathway and show a modulated response to hypoxia depending on the severity of hypoxic stress: under mild hypoxia, only the HIF-1α subunit is upregulated, while expression of the other subunit, aryl hydrocarbon nuclear translator, only increases significantly at more severe hypoxia levels. The chronic trials revealed down-regulation of genes related to cell adhesion, transport, development and heme-binding, and up-regulation of genes related to glycolysis, oxygen binding, cell differentiation, digestive and reproductive function. The intermittent hypoxia trials revealed an upregulation of heme transporter activity during hypoxia, and our time series analysis characterized nine clusters of genes with similar expression patterns. Our findings suggest that H. carunculata is likely to tolerate, and be resilient to, predicted future hypoxia conditions.

. (a) HIF-1 pathway under normoxic (left) and hypoxic conditions (right). HSP90 is a heat shock protein that stabilizes HIF1-α until (under normoxic conditions) O2 facilitates the hydroxylation of HIF-1α at two proline residues with proline hydroxylase (PHD) and one asparagine residue with factor inhibiting HIF-1α hydroxylation (FIH) (created and modified from Liu and Semenza 2007). (b) Anterior end of the bearded fireworm (Hermodice carunculata) showing the primary tissues for oxygen-uptake which are the red branching branchia (similar to gills or lungs). Photo by CJ Grimes. www.nature.com/scientificreports/ differences in gene expression patterns as well especially with regards to oxidative stress 61,62 . When organisms undergo intermittent hypoxia, they must be able to handle times of low oxygen and high oxygen when the reactive oxygen species increase which threaten cell health and structure which can result from the formation of reactive oxygen species under hypoxia or upon reoxygenation 61,62 . For intermittent hypoxia, we predict an up-regulation of aerobic metabolism genes after normoxic timepoints and behavioral response genes (such as TRP1A) as in Capitella teleta 17 after hypoxic timepoints. Hypoxia trials. Throughout the trials, the worms were not fed to reduce ammonia accumulation and decomposition. DO levels were controlled by the influx of nitrogen gas (Radnor nitrogen from Airgas Welding Supplies, TX, USA) and controlled with Neptune System's Apex Jr. Controller. See 33 (Fig. 3b). The DEFs between organisms in laboratory trials and those from the field largely consisted of metabolism-associated features and unclassified sequences. Anaerobic metabolism genes associated with glycolysis were up-regulated in the chronic hypoxia treatments (Tables 2, 3) as opposed to those associated with cell adhesion, transport, and development which were downregulated. There was also a significant up-regulation of HIF-1α in the Mild DO treatment compared to Normal and Moderate DO treatments, and an up-regulation of ARNT in the Moderate DO treatment compared to Normal (Tables 3, 4, Fig. 5). Heme-binding protein (CP3AD) was down-regulated in the Moderate treatment while oxygen-binding (extracellular hemoglobin) was up-regulated. However, there was a decrease in the expression of metabolic genes and an increase in those associated with cell differentiation, digestive tract, and reproductive system development.

Methods
Intermittent hypoxia trials. The inclusion of timepoints in the pairwise differential expression resulted in 0 DEGs and 65 DETs up-regulated compared with 3 DEGs and 67 DETs down-regulated (Fig. 4). There was a distinction between hypoxia and normoxia conditions, but timepoints within treatments were not separated in the dendrogram (Fig. 4, top). The majority of DEFs between conditions serve metabolic functions, nucleotide binding, GTPase, and polymerase activity. However, heme-transporter activity (proton-coupled folate transporter) was down-regulated at the hypoxic timepoints and protein ubiquitination was down-regulated at the normoxic timepoints. www.nature.com/scientificreports/ Our timepoint analysis for intermittent hypoxia revealed 9 DEF clusters with characteristic expression profiles (Table 5). In clusters 3, 4, and 7, expression patterns followed similar trends for the normoxic and hypoxic timepoints (line trends in Table 5); however, DEFs in cluster 4 had higher expression levels under normoxia, whereas DEFs in clusters 3 and 7 were more expressed under normoxia. Cluster 1 contained genes associated with epithelial cell morphogenesis and metabolism and were expressed at higher levels after the first timepoint of hypoxia but decreased to around normal levels after the second 6-h hypoxia block (timepoint 3). Cluster 2 genes, consisting primarily of heme-binding, metabolism, and translation features, increased after both hypoxic and normoxic timepoints; however, the increase at hypoxia timepoints was more drastic. Genes in cluster 3 were expressed at higher levels in hypoxic timepoints consisting of increased heme-binding affinity and increased production of red blood cells. We did not see any differential expression of HIFs in intermittent hypoxia; however, we did see an initial decrease in expression of Usp19 (Ubiquitin carboxyl-terminal hydrolase 19) after hypoxia timepoint 1 compared to normoxia timepoint 2 but equally expressed after timepoints 3 and 4 (Cluster 5). Genes associated with embryonic development, cell proliferation, and angiogenesis were up-regulated during hypoxia (Cluster 7). The largest slope increase and decrease in expression trends occurred in cluster 8 for hypoxic and normoxic timepoints, respectively. The large positive slope associated with cluster 8 under hypoxia, nearly doubles www.nature.com/scientificreports/ in expression between to the two timepoints. Cluster 8 genes had inverse expression trends between hypoxic and normoxic timepoints and contained genes associated with epidermal growth factor-, actin-, and Notch-binding.

Discussion
Our chronic trials revealed down-regulation of genes related to cell adhesion, transport, development and hemebinding, and up-regulation of genes related to glycolysis, oxygen binding, cell differentiation, digestive, and reproductive function under hypoxic conditions. The overall up-regulation of DEFs associated with metabolic  www.nature.com/scientificreports/  www.nature.com/scientificreports/ www.nature.com/scientificreports/ functions in the Moderate DO fireworms indicates a stress response through shifted metabolism in response to chronic hypoxia from aerobic to anaerobic (Tables 2, 3). We found differential expression of metabolic genes between Mild and Moderate treatments indicating modulated responses (e.g. down-regulation of aerobic metabolic pathways to hypoxic exposure based on level of severity) in Table 4. For example, we saw a down-regulation of ATP synthase and cytochrome c oxidase, key members of the electron transport (Table 4). This response has been shown in several invertebrate species and shows that fireworms have a similar physiological response to hypoxia as the Pacific oyster, Crassostrea gigas, and the water flea, Daphnia magna 42,48,51 .
The two subunits of the HIF transcription factor, HIF-1α and ARNT, are constitutively expressed, even under normoxic conditions, but a functional transcription factor only forms under hypoxia when the oxygendependent degradation of HIF-1α is disrupted. Once dimerized, HIF triggers downstream hypoxia responses in the organism 21,[30][31][32] . However, few studies have examined the differential expression of the two HIF subunits in response to different levels of hypoxia. We found that both HIF-1α and ARNT are upregulated under Mild hypoxia (Fig. 5), although the upregulation was only significant for HIF-1α. Under Moderate hypoxia, only ARNT is significantly upregulated whereas HIF-1α expression decreases. The increase in ARNT expression from Normal to Mild to Moderate chronic hypoxia (Fig. 5) further indicates a modulated response to chronic hypoxia exposure dependent upon DO level. This implies that downstream hypoxia responses may not only be triggered through oxygen-dependent degradation of HIF-1α, but possibly also through a modulation of expression levels depending on DO levels. The higher expression of HIF-1α after Mild chronic hypoxia but return to normal expression at moderate levels could indicate yet unknown underlying acclimatory responses to this chronic condition.
As in Paralvinella sp. 52,53 , extracellular hemoglobin was up-regulated in the fireworms after Moderate chronic hypoxia compared to Mild hypoxia (Table 4). Another invertebrate blood pigment, hemocyanin, has putatively been found in the transcriptomes of a closely related fireworm, Paramphinome jeffreysii 72 , and was down-regulated in the bearded fireworm after Moderate chronic hypoxia compared to normoxic conditions. The increased production of extracellular hemoglobin and decreased production of hemocyanin in Moderate hypoxia suggests that the two pigments are optimized for different DO levels in the blood; however, a more targeted study is necessary to understand this relationship.
Our control DO levels were similar to prevailing conditions around the time of capture of the field-collected specimens; consequently, the fewest DEFs were found between the Normal treatment and the field-collected worms. Also, the down-regulation of certain metabolic genes associated with aerobic metabolism when comparing the animals exposed to lowered DO conditions to those of their natural habitat indicates the overall metabolic depression of worms exposed to chronic hypoxia. These results therefore support a previous study demonstrating lowered oxygen uptake in chronic hypoxic fireworms 27 .
When analyzing the top DEFs during the intermittent hypoxia trials, no clear separation of timepoints within the hypoxic treatment was obvious, but there was a distinction between the normoxic timepoints (Fig. 4, top left and right, respectively). While most of the differentially expressed genes in the Fig. 4 remained unclassified, a few in the upper right section (expressed more after normoxic timepoints) were putatively characterized as genes related to aerobic metabolism and cell adhesion, such as Cytochrome c oxidase subunit 1 (COX1) and periostin. However, the time course cluster analysis revealed some expression trends depending on time points ( Table 5). The worms used in this study had likely never been exposed to hypoxia before, so it is not surprising that we see different expression trends at timepoint 1 and 3. The increase in expression of genes related to epithelial morphogenesis after 6 h (Table 5) may be the first sign of increasing branchial filament formation in this species to cope with hypoxic conditions, which occurs both under intermittent 38 and chronic hypoxia 39 . Although fireworms most likely cannot increase their branchial surface area within six hours, this short time may be sufficient to trigger the onset of branchial tissue formation.
The consistent higher expression of hematopoietic prostaglandin D synthase (Cluster 3) after the hypoxic timepoints suggests physiological responses related to vasodilation and inhibition of clotting so that blood transporting oxygen can flow more freely 73 . This short-term response to intermittent hypoxic stress may increase blood flow and oxygen transport to maintain necessary metabolic functions during times of lowered oxygen conditions.
Ubiquitin-specific protease 19 (Usp19) protects HIF-1α from degradation and is required for typical hypoxia response in vertebrates 74,75 . If Usp19 is absent under oxygen depletion, HIF-1α will follow an oxygen-independent pathway to degradation not shown in Fig. 1a 75,76 . Usp19 interacts with the N-terminus of HIF-1α which is also where the chaperone, Heat shock protein 90 (Hsp90), associates to stabilize HIF-1α (Fig. 1a). In our experiment, Usp19 was initially less expressed after hypoxic timepoint 1 compared to normoxia (Cluster 5 in Table 5), possibly indicating the destabilization of the HIF-1α chaperone which may allow it to dimerize with ARNT more rapidly. www.nature.com/scientificreports/ However, Usp19's similar expression level at timepoints 3 and 4 suggests it has other roles under normoxic conditions as well. Typically, under hypoxic conditions, HIF-1α is protected from proteasomal degradation by Usp19, but the expression trends (higher expression after first hypoxic exposure) here suggest there may be another protective mechanism for HIF-1α in the fireworm and potentially other animals. Upon comparison of the intermittent and chronic studies, we found consistent regulation of a TRP1A homolog across all hypoxia and normoxic conditions and timepoints supporting its role in oxygen sensing in annelids 17 . Heme-binding proteins were more highly expressed after the hypoxia timepoints during intermittent hypoxia and portrayed a positive slope as time progressed indicating increased expression after repeated hypoxic exposure. Compared with chronic hypoxia, heme-binding proteins were up-regulated after Mild hypoxic exposure compared with Normal, followed by a down-regulation in severe compared with Mild. Such a result may entail several stages of hypoxic response in this species depending on DO level, exposure time, and frequency of exposure. It may be necessary to reach a certain level of lowered DO (4.5 mg O 2 L −1 ) before heme-binding protein genes are up-regulated in response to hypoxia, but then they may be down-regulated if the production of the protein is too energetically expensive. However, more research would need to be done to further support this suggestion.
Filamin A (FLNA) is a structural protein associated with actin-binding that has been described as important for cancer tumor growth in the lungs 77 , and we saw it up-regulated in the Moderate chronic hypoxia treatment compared to Normal. In addition, in the intermittent trials, FLNA was consistently more expressed after hypoxic timepoints indicating a similar response to chronic vs intermittent hypoxia. Since this protein is involved with new growth of cancerous tumors and other cell types (such as blood vessels 78 ), the higher expression after chronic Moderate and intermittent hypoxia supports the idea that lowered oxygen levels triggers growth in specific cell types.
The different expression patterns of hematopoietic prostaglandin D synthase during intermittent hypoxia, but absence during chronic hypoxia indicates that this may be purely a short-term reaction to hypoxic stress. After the initial stress, the organism seems to shift to morphological and longer-term responses (such as metabolic depression and development of new blood vessels) to cope with hypoxia. H. carunculata has previously been shown to exhibit plasticity in its branchial morphology to cope with hypoxic conditions 38,39,41 , but the underlying molecular mechanisms for these changes have not previously been described. Our study shows that hypoxia responses in H. carunculata can be modulated depending on the type (intermittent vs. chronic) and the severity of hypoxia exposure.
Our chronic hypoxia experiments showed that Notch-binding was up-regulated in the Moderate treatment compared to the Mild but was down-regulated in the Mild treatment compared to Normal indicating a differential response that is dependent on DO levels. We also saw differential expression of Notch associated proteins between the chronic and intermittent hypoxia trials. The up-regulation of Notch in Moderate hypoxia compared to Mild was similar to the drastic increase in regulation between timepoints 1 and 3 in the intermittent trials. The up-regulation of Notch proteins, associated with development (especially that of the heart) under hypoxia conditions has been described in zebrafish, Danio rerio 32 . Likewise, hypoxia-triggered up-regulation of Notch-binding proteins has been shown to increase tracheal formation in insects 79 . Our findings suggest that these three divergent phyla, representing the three major branches of the Bilateria (Deuterostomia, Ecdysozoa, Spiralia/Lophotrochozoa) may utilize similar signaling pathways to alter their vastly different oxygen-delivery systems. The role of Notch-binding in oxygen delivery in lophotrochozoans requires further study. In addition, Notch signaling has been shown to increase expression of HIF-1α in hypoxic conditions 80 , so its importance in development and response to hypoxia has been exemplified here again.
Fibropellin-1, related to Notch protein-binding, was recently shown to have drastically increased in expression post chronic hypoxia in the clam Ruditapes philippinarum 81 . In the current study, we see a similar expression pattern in the intermittent but not chronic hypoxia trial. We saw an increase in expression of Fibropellin-1 in the bearded fireworm after the second hypoxic timepoint in the intermittent hypoxia trial (over 18 h after initial hypoxic exposure) which corresponds well with the clam's drastic increase in this gene after 2 days of chronic hypoxia. However, it is important to note that a significant decrease in expression of Fibropellin-1 was seen after 5 and 8 days in the clam 81 . During our chronic hypoxia exposure experiment, we were only able to sample after 7 days, so we may have missed this spike in expression, but our intermittent experiment allows us to see a similar response.
Although our biological replicates were limited to 2-3 individuals per treatment, our gene expression analyses discerned statistically significant differences and trends between groups exposed to different types (chronic vs. intermittent) and levels of hypoxia. Many of the gene expression responses that we observed are consistent with previous observations. For example, the higher expression levels of genes related to epithelial morphogenesis are consistent with the increase in surface branchial surface area 39 ; likewise, expression patterns of metabolismrelated genes correspond with our observation of oxygen uptake 27 . Other observed trends, such as the expression patterns of genes in the HIF or Notch pathways, can guide future biochemical and physiological studies in H. carunculata or other species.
This study showed that six hours of hypoxia exposure is sufficient to initiate molecular pathways for epithelial morphogenesis and blood vessel remodeling, providing further support that morphological changes of respiratory structures are not reliable taxonomic characters in amphinomids (or even possibly in annelids in general). Our data strongly support the notion that H. carunculata can swiftly initiate physiological and morphological responses to hypoxia 38,41 even if a behavioral response (avoidance) is not an option. This study supports our previous prediction 33 that H. carunculata will thrive under changing ocean conditions and may experience population increases that could potentially have damaging effects on coral reefs and other sensitive marine habitats in its distribution range. www.nature.com/scientificreports/