Global stabilization of the mitochondrial proteome is associated with extreme anoxia tolerance in Austrofundulus limnaeus WS40NE cells

Abstract

Background

Austrofundulus limnaeus is an extremophile vertebrate native to small temporary ponds of Venezuela. Embryos of A. limnaeus must survive variable and often extreme conditions, including long periods of anoxia. Neuroepithelial cells derived from these embryos, WS40NE cells, provide a unique tool to understand how the proteome changes in response to anoxic stress.

Results

Using label-free proteomics, 19,604 peptides and 3487 proteins were quantified in normoxic, 4d anoxic, and 24 h recovery WS40NE cells. Of these, 2612 proteins (74.9%) were statistically significantly differentially abundant in at least one comparison: 1988 comparing normoxia to 4 days anoxia (57.0%), 923 comparing 4 days anoxia to 24 h recovery (26.5%), and 1814 comparing normoxia to 24 h recovery (52.0%). Further, interaction networks of proteins with similar expression patterns suggest that relative mitochondrial capacity may increase during anoxia, including upregulation and/or preferred stabilization of proteins involved in mitochondrial metabolism and mitochondrial transcription and translation. This is in sharp contrast to trends in proteins that support cytoplasmic translation.

Conclusions

These data support an active role for mitochondria in mediating the survival of the anoxia-tolerant WS40NE cell line and highlight the value of this non-traditional vertebrate model for uncovering novel mechanisms of cellular resilience.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12915-026-02674-9.

Background

While nearly all organisms have mechanisms to modify gene expression in response to environmental challenges [1–3], environmental stressors can often lead to aberrant gene expression, especially when organisms experience extreme conditions. Anoxia is usually deadly to vertebrates, most of which require a constant supply of oxygen for normal brain and heart function. These organs fail because anoxic stress disrupts regulation of cellular pathways, induces mitochondrial dysfunction, and ultimately results in cell death via regulated and unregulated pathways [4–6]. Most animals that can survive long bouts of anoxia do so by entering a profound state of dormancy called anoxia-induced quiescence [7–9]. In this state, cells downregulate unnecessary cellular processes and only expend adenosine triphosphate (ATP) on pathways that increase survival [10, 11]. Survival often includes coordinated downregulation of metabolic and biosynthetic pathways to maintain cellular ATP levels while correctly modulating gene expression, maintaining cellular integrity, and preventing dysregulation [12–16]. Gene expression is regulated by a variety of mechanisms, including epigenetic, transcriptional, posttranscriptional, translational, and posttranslational controls [17–22], that can drastically shift the proteomic landscape [23–25]. Recently, a continuous cell line that can survive seven weeks of anoxia has been isolated from embryos of the annual killifish, Austrofundulus limnaeus [26]. This novel feature of the embryonic A. limnaeus cell line makes it an excellent model for studying the cellular mechanisms that can support extreme anoxia tolerance in vertebrates.

Across the vertebrate classes, anoxia tolerance is highest just after fertilization and decreases steadily during development [27]. In contrast to this pattern, anoxia tolerance in embryos of A. limnaeus first increases substantially during development before returning to levels comparable to other vertebrates at the completion of development [27]. Anoxia is a normal part of the developmental environment of annual killifishes, and embryos of A. limnaeus can survive over 150 days without oxygen despite being composed mainly of neural and cardiac tissues [28–30]. Anoxia tolerance increases during early development and is associated with the embryos entering into a programmed metabolic depression known as diapause [27]. However, peak anoxia tolerance in this species occurs in actively developing, post-diapause embryos that can enter a profound state of anoxia-induced quiescence in response to oxygen deprivation [27, 29, 30]. In addition, at the transcriptome level, diapause and anoxia-induced quiescence are distinct states in embryos of A. limnaeus [31]. Multiple genes associated with mitochondrial function and heat shock proteins are believed to be under positive selection in this species, suggesting that mitochondrial and protein chaperone pathways may also contribute significantly to their anoxia tolerance [32, 33]. This is further supported by a lack of mitochondrial DNA degradation and differential expression of mitochondrial small noncoding RNAs (mitosRNAs) during anoxia and reoxygenation in developing embryos [34, 35]. Additionally, diapausing and developing embryos have a high small-molecule antioxidant capacity that may protect against oxidative damage during transitions into and out of anoxia [36]. Thus, embryos of A. limnaeus are poised to survive without oxygen and appear to have evolved a very different developmental physiology from other vertebrates that supports their extreme tolerance.

PSU-AL-WS40NE, (WS40NE), is a neuroepithelial cell line that was isolated from a late-stage (Wourms’ Stage 40) embryonic A. limnaeus tissue explant [26]. Wourms’ stage 40 embryos are actively developing, have functional and differentiated nervous and circulatory systems, and have an intermediate anoxia tolerance (LT₅₀ ~ 7 days at 25 °C) compared to earlier stages of development [37]. The WS40NE cell line provides a resource for interrogating or altering cell components to test for their contribution to cellular anoxia tolerance. Notably, these cells continue to slowly divide during initial exposure to anoxia, suggesting a different response to anoxia compared to other vertebrate cell lines [26]. Understanding how these cells can survive and successfully navigate through the cell cycle in the face of severe stress may lead to new insights on the molecular mechanisms that support anoxic survival. In particular, proteomic profiling of WS40NE cells allows direct observation of shifts in the functional proteins and pathways that sustain survival under oxygen deprivation. However, the proteomic landscape of this cell line in response to anoxia and subsequent reoxygenation has never been explored. In this study, we use a label-free proteomics approach to identify proteins that change in response to anoxia and aerobic recovery from anoxia to identify the key cellular pathways that support anoxia tolerance in A. limnaeus. Our results suggest that WS40NE cells have a unique proteomic response to anoxia compared to other vertebrate models that may highlight new mechanisms to support vertebrate anoxia tolerance.

Results

Protein abundance

A total of 19,604 peptides and 3487 proteins were identified and quantified across the three experimental conditions in this study. Protein sequence coverage ranged from 1 to 78%. Of the identified proteins, 2612 (74.9%) were significantly differentially expressed in at least one comparison with principal component analysis indicating clear differences in protein expression in all 3 experimental conditions (Fig. 1A). Twenty-eight of these statistically significantly expressed proteins are uncharacterized and may represent novel players in the vertebrate response to anoxia (Additional file 1: Table S1). Protein expression differed greatly across experimental conditions: 1279 increased and 709 decreased in relative abundance after 4 days of anoxia compared to normoxic cells (Fig. 1B, C), and 359 were increased in relative abundance and 564 decreased in relative abundance after 1 day of recovery compared to anoxic cells (Additional file 2: Table S2, Fig. 1B, D). Additionally, 1029 increased in relative abundance and 785 decreased in relative abundance when comparing 1 day of aerobic recovery to normoxic cells (Additional file 2: Table S2, Fig. 1B,E). These differences between normoxia and anoxia/anoxic recovery are reflected in condition-dependent shifts in global protein abundance profiles across samples (Fig. 1F). A large proportion of the differentially abundant proteins are shared between 4 days of anoxia and 24 h of recovery from anoxia (Fig. 1G). Filtering for highly differentially abundant proteins (log₂ fold-change of ± 1 or greater) resulted in 33 proteins increased in relative abundance and 123 decreased in relative abundance comparing normoxia to 4 days of anoxia, 49 proteins increased in relative abundance and 19 decreased in relative abundance comparing 4 days anoxia to 1 day of aerobic recovery, and 81 proteins increased in relative abundance and 90 decreased in relative abundance comparing normoxia to 1 day of aerobic recovery (Fig. 1, Additional file 3: Supplemental Workbook 1).

Fig. 1.

Changes in the proteomic landscape during anoxia and recovery in WS40NE cells. A Principal component analysis of protein expression patterns reveals distinct patterns of expression during anoxia and aerobic recovery compared to normoxic samples. Anoxic (green), normoxic (blue), and recovery (purple) samples are plotted according to PC1 and PC2 with a 95% confidence ellipse surrounding each cluster. B Pie charts representing the percentage of differentially expressed proteins in each comparison. A large proportion of the detected proteome is differentially expressed in response to anoxia and aerobic recovery. Volcano plots illustrating the changes in protein abundance in C normoxia compared to 4 days anoxia, D recovery compared to 4 days anoxia, and E normoxia compared to recovery. Proteins are colored according to their significance in terms of corrected p value (blue), fold change > 1.0 (green), both corrected p value and fold change (red), or neither (gray). Adjusted p values less than 0.05 were considered significant. F Bar graphs depicting raw protein abundance for each condition. Bars (means ± s.d., n = 12) with different letters are statistically significant (ANOVA, p, 0.05). Raw protein abundance across treatment groups suggests relatively comparable amounts of peptides were generated for each treatment group. G Venn diagram illustrating the large number of shared differentially expressed proteins in anoxia and after 1 day of aerobic recovery. H Clust identified 11 protein clusters (C0–C10), which included 1408 proteins (38). Of the 11 clusters, six co-expression patterns were observed: protein abundance that increases in anoxia and remains high during recovery (C0), proteins that increase during anoxia and continue to increase during recovery (C1, C2), proteins that increase in anoxia and but decrease during recovery (C3, C4, C5), proteins that remain stable in anoxia and decrease during recovery (C6, C7), proteins that decrease in anoxia and remain low during recovery (C8, C9), and proteins that decrease in anoxia but increase during recovery (C10). The highlighted yellow cell components beneath each panel represent the Cellular Component Gene Ontology terms that are highly enriched in each cluster. Created in BioRender. Hughes, C. (2025) https://BioRender.com/4lnj3z2

The top 25 proteins that increased and decreased in relative abundance in anoxia and recovery relative to normoxia were determined by comparing log₂ fold changes. In anoxia, the top two proteins with the greatest increase in relative abundance were heat shock proteins of the 30 kDa class—the only two heat shock proteins in the top 25. In general, proteins with increased abundance were from a variety of pathways (Table 1). Of the top 25 proteins that decreased in relative abundance in anoxia, 56% were cytoplasmic ribosomal proteins (Table 1). In terms of aerobic recovery, the top 25 proteins with the greatest increase in relative abundance are dominated (24%) by heat shock or molecular chaperone proteins (Table 2). Of those most decreased in recovery, 52% were ribosomal proteins (Table 2). Seven of the 25 most increased and eleven of the most decreased proteins were shared between both anoxia and recovery.

Table 1.

The top 25 most increased and decreased proteins in WS40NE cells after 4 days of anoxia compared to normoxic controls

Protein identifier	Protein annotation	log2FC
Increased
XP_013857981.1	Heat shock protein 30-like*	3.2632
XP_013857979.1	Heat shock protein 30-like isoform X2*	2.2715
XP_013881880.1	Glia-derived nexin isoform X2	1.9921
XP_013872870.1	Zinc finger protein 638-like	1.9771
XP_013864068.1	C-terminal-binding protein 1-like isoform X2*	1.8001
XP_013887191.1	Structural maintenance of chromosomes protein 6-like	1.7335
XP_013880212.1	Procollagen-lysine,2-oxoglutarate 5-dioxygenase 1	1.7241
XP_013864250.1	Tumor necrosis factor receptor superfamily member 16-like*	1.5770
XP_013887549.1	5-phosphohydroxy-L-lysine phospho-lyase isoform X3*	1.5304
XP_013855979.1	Prolyl 4-hydroxylase subunit alpha-1 isoform X2	1.5114
XP_013888884.1	WD repeat-containing protein 37-like	1.4643
XP_013861108.1	Protein NDRG1-like isoform X1	1.4589
XP_013885977.1	Calcium/calmodulin-dependent protein kinase type 1-like isoform X2	1.4381
XP_013876809.1	Dedicator of cytokinesis protein 5	1.3967
XP_013876234.1	Golgi membrane protein 1 isoform X3*	1.3275
XP_013870563.1	Peptidyl-glycine alpha-amidating monooxygenase	1.2545
XP_013883686.1	Serine/threonine-protein phosphatase CPPED1	1.2467
XP_013861109.1	Protein NDRG1-like isoform X2	1.2368
XP_013867749.1	Nuclear factor of activated T-cells 5-like isoform X2	1.2326
XP_013874967.1	Cellular tumor antigen p53	1.2298
XP_013871888.1	Chromosome-associated kinesin KIF4A	1.1964
XP_013875799.1	Spindle and kinetochore-associated protein 2-like	1.1679
XP_013879464.1	Calmodulin-regulated spectrin-associated protein 1 isoform X3*	1.1479
XP_013885589.1	THO complex subunit 3	1.1062
XP_013876354.1	Beta-galactoside-binding lectin-like	1.1052
Decreased
XP_013884949.1	cathepsin L1-like*	− 3.7639
XP_013877271.1	H/ACA ribonucleoprotein complex subunit 3*	− 2.9810
XP_013864703.1	Lanosterol 14-alpha demethylase*	− 2.5092
XP_013867249.1	60S ribosomal protein L6	− 2.2678
XP_013859853.1	60S ribosomal protein L34*	− 2.2009
XP_013878441.1	PCNA-associated factor	− 2.0810
XP_013862047.1	60S ribosomal protein L7a	− 2.0463
XP_013885731.1	Hematological and neurological expressed 1-like protein	− 2.0332
XP_013865417.1	60S ribosomal protein L35*	− 1.9723
XP_013874616.1	40S ribosomal protein S23*	− 1.9706
XP_013872903.1	Probable ATP-dependent RNA helicase DDX5*	− 1.9201
XP_013855239.1	Prothymosin alpha-B-like	− 1.9077
XP_013884441.1	60S ribosomal protein L18a	− 1.9050
XP_013869136.1	60S ribosomal protein L4	− 1.9020
XP_013863085.1	60S ribosomal protein L19-like	− 1.8958
XP_013883098.1	60S ribosomal protein L8	− 1.8767
XP_013882727.1	60S ribosomal protein L13	− 1.8575
XP_013880442.1	coiled-coil-helix-coiled-coil-helix domain-containing protein 5*	− 1.8458
XP_013870436.1	60S ribosomal protein L13a	− 1.8195
XP_013872998.1	60S ribosomal protein L24	− 1.7647
XP_013858448.1	Probable ATP-dependent RNA helicase DDX5*	− 1.7190
XP_013889413.1	Tumor necrosis factor-inducible gene 6 protein-like	− 1.7087
XP_013863523.1	60S ribosomal protein L36*	− 1.6915
XP_013857786.1	60S ribosomal protein L7	− 1.6785
XP_013870138.1	60S ribosomal protein L26*	− 1.6758

Log₂FC Log₂ fold-change

Proteins with an asterisk are shared with differentially expressed proteins during aerobic recovery when compared to normoxic controls

Table 2.

The 25 proteins that most increased and decreased in relative abundance in WS40NE cells during aerobic recovery compared to normoxic controls

Protein identifier	Protein annotation	log2FC
Increased
XP_013857981.1	Heat shock protein 30-like*	7.7807
XP_013857979.1	Heat shock protein 30-like isoform X2*	6.3969
XP_013882746.1	Pre-mRNA-splicing factor ATP-dependent RNA helicase PRP16	6.2235
XP_013864068.1	C-terminal-binding protein 1-like isoform X2*	4.1771
XP_013855715.1	Keratin, type I cytoskeletal 13-like	3.9186
XP_013875446.1	Heat shock 70 kDa protein	3.6674
XP_013871083.1	Protein virilizer homolog isoform X6	2.9394
XP_013879745.1	Retinoid-inducible serine carboxypeptidase	2.8332
XP_013879464.1	Calmodulin-regulated spectrin-associated protein 1 isoform X3*	2.7698
XP_013870878.1	Protein farnesyltransferase subunit beta-like	2.4700
XP_013867865.1	Heat shock cognate 71 kDa protein-like	2.4237
XP_013887218.1	Histone H3-like	2.3948
XP_013869932.1	Neural Wiskott-Aldrich syndrome protein-like	2.2515
XP_013875366.1	BAG family molecular chaperone regulator 3-like	2.2173
XP_013889266.1	Putative deoxyribonuclease TATDN3	2.2112
XP_013876234.1	Golgi membrane protein 1 isoform X3*	2.1123
XP_013887549.1	5-phosphohydroxy-L-lysine phospho-lyase isoform X3*	2.0462
XP_013864250.1	Tumor necrosis factor receptor superfamily member 16-like*	2.0380
XP_013871017.1	Heat shock 70 kDa protein-like, partial	1.9587
XP_013877056.1	dynactin subunit 3	1.8728
XP_013861071.1	CCR4-NOT transcription complex subunit 2	1.8644
XP_013881446.1	dnaJ homolog subfamily B member 1	1.8006
XP_013880015.1	nestin	1.7485
XP_013884401.1	SPARC	1.6828
XP_013859080.1	dnaJ homolog subfamily A member 4-like isoform X2	1.6306
Decreased
XP_013884949.1	Cathepsin L1-like*	− 3.2899
XP_013864703.1	Lanosterol 14-alpha demethylase*	− 2.8677
XP_013877271.1	H/ACA ribonucleoprotein complex subunit 3*	− 2.7874
XP_013855714.1	40S ribosomal protein S14-like	− 2.6510
XP_013870499.1	40S ribosomal protein S25	− 2.3391
XP_013863523.1	60S ribosomal protein L36*	− 2.2900
XP_013869581.1	High mobility group protein B1-like	− 2.2705
XP_013864436.1	60S ribosomal protein L37a	− 2.1544
XP_013859853.1	60S ribosomal protein L34*	− 2.0449
XP_013870138.1	60S ribosomal protein L26*	− 2.0093
XP_013865417.1	60S ribosomal protein L35*	− 2.0089
XP_013874251.1	60S ribosomal protein L27	− 1.9553
XP_013874616.1	40S ribosomal protein S23*	− 1.9475
XP_013865390.1	40S ribosomal protein S29	− 1.9469
XP_013880442.1	Coiled-coil-helix-coiled-coil-helix domain-containing protein 5*	− 1.8694
XP_013858532.1	60S ribosomal protein L36a	− 1.8184
XP_013877707.1	Cationic amino acid transporter 3-like	− 1.7996
XP_013860561.1	40S ribosomal protein S24 isoform X2	− 1.7982
XP_013857155.1	ATP synthase subunit epsilon, mitochondrial	− 1.7965
XP_013879561.1	Histone H2B 1/2	− 1.7832
XP_013858448.1	Probable ATP-dependent RNA helicase DDX5*	− 1.7657
XP_013862691.1	CD81 antigen	− 1.7463
XP_013872903.1	Probable ATP-dependent RNA helicase DDX5*	− 1.5918
XP_013874037.1	28S ribosomal protein S15, mitochondrial	− 1.5892
XP_013887446.1	Glypican-1-like	− 1.5462

Log₂FC Log₂ fold-change

Proteins with an asterisk are shared with differentially expressed proteins during anoxia when compared to normoxic controls

Protein co-expression patterns

Because we limited the Clust analysis to proteins with human orthologs, 3044 proteins were included in this analysis [38]. Clust identified 11 co-expression clusters (C0–C10), which included 1408 proteins, while 1636 were not sorted into any clusters (Additional file 4: Supplemental Workbook 2). Six major co-expression patterns were observed across the 11 clusters (Fig. 1H): proteins that increase in anoxia and remain high during recovery (C0), proteins that increase during anoxia and continue to increase during recovery (C1, C2), proteins that increase in response to anoxia and decrease during recovery (C3, C4, C5), proteins that remain relatively stable during anoxia and then decrease during recovery (C6, C7), proteins that decrease in anoxia and remain low in recovery (C8, C9), and proteins that decrease in response to anoxia but increase during recovery (C10). Multiple notable pathways were enriched within each cluster as discussed in the following sections (Additional file 5: Supplemental Workbook 3).

Mitochondrial proteins increase during anoxia

Nuclear-encoded proteins involved in mitochondrial function are concentrated in co-expression clusters that increase during exposure to anoxia (C0, C1, C3, C4) (Figs. 2, 3, and 4). Proteins enriched in cluster C0 are involved in the TCA cycle and respiratory electron transport have higher relative abundance after 4 days of anoxia and remain high during 1 day of aerobic recovery (Fig. 2). Conversely, the rate of daily lactate accumulation decreases during anoxia and recovery as compared to normoxia (Fig. 2C). Other ontology terms enriched in this cluster include Golgi-ER retrograde transport, mitochondrial tRNA aminoacylation, metabolism, and mitotic anaphase (Fig. 2). Proteins in cluster C1 that increase during anoxia and recovery are involved in branched-chain amino acid metabolism and mitochondrial gene expression and protein synthesis (Fig. 3). Clusters C3 (Fig. 4) and C5 (Fig. 5), which contain proteins that increase in response to anoxia and then decrease during aerobic recovery, are enriched for proteins involved in mitochondrial gene expression and transport. Thus, many proteins encoded in the nucleus but resident in the mitochondrial compartment increase in abundance in response to 4 days of anoxia.

Fig. 2.

Gene ontology enrichment pathways for proteins that increase in response to anoxia and remain elevated during aerobic recovery in cluster C0. Notable enriched pathways that are co-expressed include the TCA cycle, mitotic anaphase, membrane trafficking, and Golgi-to-ER retrograde transport. The most enriched ontology pathways were identified for A biological process, which notably included many proteins involved in the B TCA cycle and D cellular respiration. Panel C shows the rate of extracellular lactate accumulation per day in anoxic, anoxia-recovered, and normoxic WS40NE cells. Different letters indicate statistically significant differences between lactate accumulation (one-way ANOVA with Tukey’s HSD test p > 0.05, n = 6) Please see Supplemental Methods for more details. E Reactome pathway enrichment analysis also includes proteins involved in the TCA cycle. Reactome and gene ontology terms are sorted by signal, the weighted harmonic mean between the observed/expected ratio and -log(FDR). The color gradient of the lollipop plot denotes the false discovery rate (FDR). For all interaction networks, network nodes represent proteins; filled nodes show the known or predicted 3D structure of each protein. Protein names without asterisks appear in this cluster in the dataset. Protein names with asterisks represent predicted functional partners: partners that did not appear in the dataset are marked with a black asterisk, while partners that appeared in the dataset but are not present in the cluster are marked with a teal asterisk. Predicted partners that appeared in the dataset but not the represented cluster and were also statistically significantly differentially expressed in at least one comparison are marked with a blue asterisk. Edges represent protein–protein associations and are color coded according to the available evidence to suggest interactions: known interactions from curated databases (teal), experimentally determined known interactions (purple), co-expression (black), protein homology (pale blue), text mining (light green), predicted interactions based on gene neighborhood (green), predicted interactions based on gene fusions (red), and predicted interactions based on co-occurrence (blue)

Fig. 3.

Gene ontology enrichment pathways of proteins that increase in anoxia and continue to increase during recovery in cluster C1. The most enriched ontology pathways were identified for A Biological process and C Reactome pathways. Notable enriched pathways include multiple metabolic processes, transcriptional regulation of small RNAs, mitochondrial gene expression and translation, and SUMOylation of DNA damage response and repair proteins. B Network nodes illustrate co-expressed proteins that increase in response to anoxia and continue to increase during aerobic recovery, including those involved in mitochondrial translation (D). Please see the caption of Fig. 2 for a detailed definition of signal in the context of the enrichment analysis and a key to interpreting the lollipop plots and protein networks

Fig. 4.

Gene ontology enrichment pathways for proteins that increase in response to anoxia and decrease during aerobic recovery in cluster C3. The most enriched ontology pathways were identified for A biological process and B Reactome pathways. Proteins included in C3 rise during anoxia and begin to decrease during recovery, nearing normoxic levels. Notable enriched pathways include: mitotic cell cycle, mitochondrial gene expression, cellular biosynthetic process, and membrane trafficking. C Network analysis illustrates the number of proteins involved in mitochondrial gene expression and highlights their connection to other processes enriched in this cluster, including sister chromatid cohesion and mRNA stabilization. Please see the caption of Fig. 2 for a detailed definition of signal in the context of the enrichment analysis and a key to interpreting the lollipop plots and protein networks

Fig. 5.

Gene ontology enrichment pathways for proteins that increase in response to anoxia and decrease during aerobic recovery in cluster C5. The most enriched A biological process and C Reactome pathways include intracellular transport, protein localization to the membrane, mitochondrial protein import, and metabolism of RNA. B Protein networks illustrate the large number of proteins involved in targeting proteins to the mitochondrial compartment and their connection to vesicular trafficking through SIL1, a nucleotide exchange factor for HSPA5 in the ER lumen. Also of note is the increase in proteins involved in ER co-translational protein import. Please see the caption of Fig. 2 for a detailed definition of signal in the context of the enrichment analysis and a key to interpreting the lollipop plots and protein networks

mRNA and miRNA processing proteins increase during anoxia and recovery from anoxia

Cluster C2 includes 141 proteins that increase in expression during anoxia and continue to increase during aerobic recovery; this cluster is enriched for proteins involved in mRNA and miRNA processing (Fig. 6). Also included in this cluster are proteins involved in transcription-coupled DNA repair, HSP90 chaperoning of steroid hormone receptors, and mitochondrial respiratory chain proteins (Fig. 6).

Fig. 6.

Gene ontology enrichment pathways for proteins that increase in response to anoxia and further increase during aerobic recovery in cluster C2. The 15 most enriched ontology pathways were identified for A biological process and C Reactome pathways. B Proteins that increase during both anoxia and anoxic recovery are associated with mRNA processing, RNA polymerase II transcription termination, and double-stranded RNA binding. Please see the caption of Fig. 2 for a detailed definition of signal in the context of the enrichment analysis and a key to interpreting the lollipop plots and protein networks

Proteins associated with transcriptional regulation by small RNAs increase during anoxia and continue to increase in recovery (Fig. 7). These include five nuclear pore proteins and two RNA polymerase II subunits (Fig. 7A). Similarly, six proteins (NUP98, NUP58, POLR2I, POLR2A, PRKRA, and TARBP2) are associated with RNA gene silencing (Fig. 7B). Multiple proteins involved in the metabolism of RNA are also co-expressed in C5 and C6, increasing in anoxia but decreasing during recovery (Additional file 6: Fig. S1). Nineteen proteins are in C5, including nine proteins involved in the processing of capped intron-containing pre-mRNA. Thirteen proteins are in C6, including functional partners DEAD-box helicase 41 (DDX41), splicing factor 3B subunit 1 (SF3B1), and serine/arginine-rich splicing factor 2 (SRSF2).

Fig. 7.

Co-expressed proteins involved in the regulation of small RNAs that increase in response to anoxia and continue to increase during aerobic recovery in clusters C1 and C2. Protein interactions networks show proteins associated with A regulation of small RNAs and B gene silencing by RNA. Please see the caption of Fig. 2 for a key to interpreting the protein networks

Proteins involved in lipid metabolism and Golgi-to-ER retrograde transport increase during anoxia

Cluster C4 includes proteins that increase in response to anoxia and decline during aerobic recovery (Fig. 8). A variety of lipid biosynthetic pathways are included in this cluster, including proteins involved in fatty acid metabolism and phospholipid biosynthesis (Fig. 8).

Fig. 8.

Gene ontology enrichment pathways of proteins that increase in anoxia and decrease in recovery in cluster C4. The most enriched ontology pathways were identified for A biological process and C Reactome pathways. Notable enriched pathways include metabolism of lipids, cellular biosynthetic process, Golgi-to-ER retrograde transport, and intracellular transport. B Proteins involved in lipid and membrane biosynthesis and D metabolism of lipids increase during anoxia and decrease during recovery. Please see the caption of Fig. 2 for a detailed definition of signal in the context of the enrichment analysis and a key to interpreting the lollipop plots and protein networks

Most proteins associated with protein trafficking in the endomembrane system follow similar co-expression trends, increasing in anoxia but increasing or decreasing during recovery. Golgi-to-ER retrograde transport proteins are present in C0, C4, and C6 and include proteins related to both COPI-dependent and independent Golgi-to-ER retrograde traffic and COPII-mediated vesicle transport (Fig. 8 and S2) [39]. Twenty-two proteins associated with post-translational protein modifications in the endoplasmic reticulum and Golgi, specifically geranylgeranylation and N-glycan biosynthesis, are enriched in C6 (Additional file 7: Fig. S2C).

Proteins involved in cytoplasmic translation decrease during anoxia

Proteins associated with cytoplasmic protein synthesis are highly enriched in co-expression clusters C7–C10 where protein abundance decreases during anoxia (Fig. 9). Most of these proteins either continue to decrease during 1 day of aerobic recovery or remain low (Fig. 9A–C). While others decrease dramatically during anoxia and then increase again during recovery (Fig. 9D).

Fig. 9.

Co-expressed proteins that decrease in response to anoxia in clusters C7–C10 are dominated by proteins involved in cytoplasmic translation. Proteins involved in A ribosomal scanning and start codon recognition, B ribosomal components, C translation termination, and D translational initiation all decrease during anoxia. Please see the caption of Fig. 2 for a key to interpreting the protein networks

Stress response pathways have a variable response to anoxia and aerobic recovery

The GO term “cellular response to stress” was identified in two clusters: C2 and C10 (Fig. 10A, B). Proteins associated with the heat shock factor (HSF-1) mediated heat shock response increase moderately in anoxia and dramatically in recovery as observed in cluster C2 (Fig. 10A). Conversely, proteins associated with the heat shock protein 90 (HSP90) chaperone cycle for steroid hormone receptors decrease during anoxia in and increase moderately during recovery in cluster 10 (Fig. 10B).

Fig. 10.

Co-expressed proteins involved in stress responsive pathways exhibit a variety of expression patterns in response to anoxia and aerobic recovery from anoxia. Proteins involved in various stress pathways were identified in both anoxia and recovery, including A, B cellular response to stress, C SUMOlyation of DNA damage response and repair proteins, D cellular response to heat stress, and E stress granule assembly. Please see the caption of Fig. 2 for a key to interpreting the protein networks

Eight SUMOylation of DNA damage response and repair proteins (STAG2, CETN2, XRCC4, TPR, NUPS4, NUP160, NUP88, NUP188) increase in abundance during anoxia and recovery (Fig. 10C). Proteins associated with cellular responses to starvation (mostly ribosomal proteins) decrease in response to anoxia and are enriched in clusters C8 and C9 (Additional files 8–9: Fig. S3–S4). Proteins associated with heat stress decrease during anoxia but increase during recovery (Fig. 10D). Interestingly, proteins involved in stress granule assembly (EIF2S1, RPS23, PUM2, G3BP1, and TIA1) also decrease in response to anoxia and remain low during recovery (Fig. 10E).

Protein co-expression patterns involved in cell cycle progression

Numerous proteins involved in progression through the cell cycle changed in abundance in response to anoxia. For example, proteins involved in postmitotic nuclear pore complex reformation increase in relative abundance during anoxia (cluster C0, Fig. 11A). Additional proteins associated with mitotic cell cycle also increase in abundance during anoxia (cluster C3, Fig. 11B). These 13 proteins include protein kinase C alpha type (PRKCA), Cullin-2 (Cul2), and multiple microtubule components.

Fig. 11.

Co-expressed proteins involved in cell cycle progression increase in response to anoxia. A Proteins involved in postmitotic nuclear pore complex reformation increased in abundance during anoxia. B Multiple microtubule components follow a similar trend, including proteins involved in posttranslational modification of tubulin and tubulin folding pathways. Gene ontology enrichment analyses are presented in Fig. 2 for cluster C0 and Fig. 4 for cluster C3. Please see the caption of Fig. 2 for a key to interpreting the protein networks

Recovery is associated with protein refolding

Multiple heat shock and chaperone proteins associated with protein folding and refolding were identified as having significantly increased abundance during recovery from anoxia (Fig. 12). Proteins highly increased in recovery (log₂ foldchange of 1 or greater) included HSPB1, HSA6, and HSA8 (Fig. 12).

Fig. 12.

Protein pathways of the most proteins that most highly increase in relative abundance in recovery compared to normoxia. Proteins that are highly increased in relative abundance during recovery are predominately associated with the unfolded protein response including three heat shock proteins (HSP): HSPB1, HSPA8, and HSASPA6. Other proteins are associated with nucleosome binding of histones, such as topoisomerase I (TOP1). Please see the caption of Fig. 2 for a detailed definition of signal in the context of the enrichment analysis and a key to interpreting the A protein network and B lollipop plot

Discussion

This study is the first exploration of the proteome of any annual killifish cell type entering into anoxia-induced quiescence, and provides rare insight into vertebrate cellular adaptation to extreme oxygen deprivation. Large-scale restructuring of the proteome is apparent with almost 2000 anoxia-responsive proteins, the majority of which increased in abundance under anoxic conditions. This provides possible evidence that reducing protein degradation and very likely some de novo protein synthesis are both parts of the initial response to anoxia. Past explorations of anoxia tolerance, especially among the few vertebrate species that can survive anoxia, focus on the importance of global downregulation of protein synthesis within hours of exposure [40–46]. For example, in zebrafish embryos, 1 h of anoxia is associated with a global stabilization of the proteome [47]. However, the WS40NE proteome changes substantially in response to anoxia, with most proteins increasing in abundance. It is unknown if this response occurs across all A. limnaeus cell types or at the organismal level in embryos. Future studies using single cell proteomics are certainly warranted to explore if the responses detailed here are indicative of the species, or are limited to this isolated cell line.

In contrast to other vertebrate systems, A. limnaeus embryos do not defend their ATP levels during the initial transition into anoxia, with ATP levels decreasing by 50% over the first 12–14 h of anoxia [30]. It is possible that this decline in ATP is used to fuel synthesis of proteins required for long-term survival, and perhaps more importantly, subsequent recovery from anoxia. Increases in protein abundance during anoxia occurs in the face of reduced abundances of proteins involved in cytoplasmic translation. We would expect this change in protein synthetic machinery during anoxia to cause decreased capacity for cytoplasmic translation, and thus it is highly likely that most proteins that increase in relative abundance are differentially stabilized. Additionally, the rate of lactate accumulation decreased during anoxia and only began to increase toward normoxic levels during recovery (Fig. 2C), which would suggest a decrease in the glycolytic metabolism that would be needed to support de novo translation. However, this does not necessarily mean that cytoplasmic translation is not part of the early anoxic response; it is possible that cytoplasmic translation of key stress proteins is completed within the first few hours of anoxia prior to the large-scale reduction in ATP levels as illustrated by the large increases in the HSP30-class heat shock proteins. While our methodology cannot confirm whether the observed increase in protein abundance represents de novo synthesis of proteins or differential stabilization of existing proteins, the 25 proteins that most increase in abundance were increased over twofold, with some increasing more dramatically (Tables 1 and 2).

Proteins associated with Golgi-to-ER retrograde transport proteins tend to increase in abundance during anoxia in WS40NE cells. Golgi-to-ER retrograde transport is critical for collecting misfolded Golgi-resident proteins and recycling export chaperones or export receptors [48]. Cellular stress can cause an accumulation of unfolded proteins in the ER, triggering the conserved unfolded protein response (UPR). The UPR upregulates folding capacity of the ER and downregulates other tasks such as protein synthesis to balance the biosynthetic load [49, 50]. To our knowledge, no prior studies have examined global changes in Golgi-associated protein abundance during anoxia or hypoxia. Additionally, there is very little research on Golgi-to-ER retrograde transport during anoxia or hypoxia, though the UPR is a well-established component of the hypoxic stress response [51]. In mammalian cells, Golgi stress and dysregulation may precede mitochondrial dysfunction and subsequent cell death [52], suggesting a link between Golgi functioning and dysregulation associated with oxygen stress. Additionally, reduced ER-to-Golgi protein trafficking (the reverse and a major biosynthetic pathway) is associated with increased cell death in hypoxia in mouse cells due to inhibition of the UPR [53]. Taken together, these data suggest that Golgi homeostasis and protein trafficking must be maintained during anoxic stress despite a likely increase in unfolded proteins. An increase in Golgi-to-ER retrograde transport proteins in WS40NE cells suggests that returning unfolded proteins to the ER is maintained in this cell line (Figs. 8 and S2) and may successfully contribute to its anoxia tolerance by carefully managing unfolded proteins.

Mitochondrial stabilization and investment during anoxia

Unexpectedly, mitochondrial translational proteins increase in anoxia and in recovery, suggesting a relative increase in mitochondrial translational capacity in anoxia. This is in stark contrast to pathway trends observed in human Ri-1 cells, which when exposed to severe hypoxia have a significant decrease in proteins associated with mitochondrial gene expression, including mitochondrial translation [54]. Multiple proteins implicated in mitochondrial translation elongation increase in anoxia, while other proteins involved in mitochondrial translation increase in abundance in both anoxia and recovery. These are mostly mitochondrial ribosomal small subunit (MRPS) proteins, which in contrast, have been found to be downregulated in both hypoxic HeLa and Ri-1 cells [54, 55]. Additional clusters involving tRNA metabolism even suggest the possibility of active mitochondrial protein synthesis, with proteins involved in tRNA aminoacylation increasing in anoxia (Figs. 2D, E and 8C). Again, this is contrary to other organisms that appear to decrease tRNA-aminoacylation in response to hypoxia [56].

In addition to the protein synthetic machinery, many other classes of mitochondrial proteins increase in abundance during anoxia. For example, proteins involved in the TCA cycle and respiratory electron transport increase in anoxia and recovery. These proteins are representative of multiple parts of the TCA cycle: pyruvate dehydrogenase E1 component subunit beta (PDH3), aconitate hydratase (ACO2), isocitrate dehydrogenase [NAD] subunit gamma (IDH3G), 2-oxoglutarate dehydrogenase (OGDH) and hydroxyacylglutathione hydrolase (HAGH), succinate dehydrogenase [ubiquinone] iron-sulfur subunit (SDHB), and NAD-dependent malic enzyme (ME2) [57]. Three mitochondrial translocases (SLC25A4, SLC25A5, and DNAJC19) involved in intracellular transport are also increased in abundance, further suggesting that even after 4 days of anoxia, the mitochondria remain differentially stabilized compared to other cellular components. This would be consistent with the lack of mitophagy observed in anoxic A. limnaeus embryos, and further supports that A. limnaeus cells achieve a remarkable stabilization of the mitochondria in response to anoxia [34]. Future work should be done to determine mitochondrial function in the cells during anoxia to elucidate the role mitochondria play in cell survival under such conditions.

Capacity for RNA processing and generation of small noncoding RNAs increases in anoxia

While overall capacity for cytoplasmic translation appears to be suppressed in anoxia, proteins involved in nuclear transcription increased in abundance. This may suggest continued RNA processing during anoxia. Engaging in RNA processing during anoxic conditions might seem counterintuitive since RNA/DNA synthesis consume non-trivial amounts of ATP [58]. However, given the decreases in ATP levels associated with anoxia tolerance in this species, it is possible that some limited amount of RNA processing is required for anoxic survival. For example, production of small non-coding RNAs (snRNAs) and mitosRNAs have already been identified in the initial 24 h of anoxic response in embryos of A. limnaeus [28, 35, 59]. While the exact function of small noncoding RNAs remains uncertain, these proteomic data suggest that their synthesis and processing may also exist in this cell line and play a critically important role in anoxia tolerance in WS40NE cells.

Proteins supporting microRNA biogenesis increase in anoxia, including four DNA-directed RNA polymerase subunits (POLR2A, POLR2E, POLR2I, and POLR2J), RISC-loading complex subunit TARBP2 (TARBP2), and interferon-inducible double-stranded RNA-dependent protein kinase activator A (PRKRA) (Fig. 7). The role of TARBP2 in anoxia is particularly of interest because beyond its role in posttranscriptional microRNA processing, it is believed to stabilize the oxygen-sensing HIF-1α protein by preventing its ubiquitylation and subsequent proteasomal degradation [60]. These enrichments further support that small RNAs are critical to the anoxic response and is consistent with previously described trends in A. limnaeus embryos [35].

One notable group of proteins involved in nuclear gene expression are the proteins of the nuclear pore complex. Many of these proteins increase in response to anoxia. One protein in this group of particular interest is NUP98 (Fig. 7A). NUP98 is a multifunctional protein with several roles involved in the regulation of gene expression, including the recruitment of epigenetic modifying enzyme complexes [61]. Beyond its role as a nuclear exporter of small RNAs, NUP98 has been implicated in controlling cell cycle inhibitor p21 mRNA levels and therefore cell cycle progression [62–65]. The role of NUP98 in anoxia tolerance warrants further study, particularly since WS40NE cells can divide for over a week during anoxia and proteins supporting this division are seen increasing in anoxia in this dataset (Fig. 11) [26].

Classic stress responses appear during 24 h recovery but not during anoxia

The proteomic landscape of WS40NE cells remains altered after 24 h recovery from anoxia, with 1814 proteins (52.0%) differentially expressed when compared to normoxic levels. Additionally, proteins associated with multiple pathways are enriched when comparing normoxia and recovery, suggesting the proteomic landscape has not returned to baseline after 24 h recovery. Of the 11 co-expression clusters identified, only two clusters (C3, C10) show protein levels approaching normoxic levels after 24 h of recovery. This suggests multiple pathways remain in flux, and supports a unique recovery response, rather than a simple return to the pre-anoxic state.

Another unexpected finding is the general lack of a coordinated and classical heat shock response in cells exposed to anoxia. While a number of classic heat shock proteins and other molecular chaperones are strongly induced in response to anoxia (Table 1), GO terms specifically related to cellular responses to heat shock, anoxia, or ischemia do not appear in the top 15 Biological Process Gene Ontology or Reactome terms for any clusters. This is, again, in stark contrast to human cell lines, where proteins associated with such pathways are differentially increased [54]. However, 49 proteins highly increased in relative abundance (twofold increase or greater) in recovery from anoxia, suggesting a coordinated recovery response is active (Fig. 12), and many of these proteins include molecular chaperones and heat shock proteins involved in protein folding and refolding. Consistent with the expression pattern of HSF-1 regulated proteins, some of these proteins begin to increase during anoxia, but do not peak in abundance until during recovery (Fig. 10A). These proteins are components of the dynactin complex and are known to be induced by the HSF1-mediated heat shock response. Importantly, the HSF1-mediated heat shock response is also responsible for regulating and interacting with other heat shock proteins such as HSP70 [66]. This is consistent with the upregulation of individual heat shock proteins seen in this data set (Fig. 10A, Tables 1 and 2). However, several proteins involved in HSP90 signaling decrease in response to anoxia: FKBP4, CAPZA2, HSP90AB1, DYNLL2, and PTGES3 (Fig. 10B, D). Downregulation of HSP90 proteins may seem counterintuitive. However, HSP90 interacts with many proteins involved in cell signaling, including 30% of ubiquitin ligases [67]. A decrease in HSP90 proteins such as HSP90AB1 may therefore be related to a downregulation of key cell signaling pathways, including steroid hormone signaling and the ubiquitin proteasomal degradation system during anoxia. As metabolic suppression in other anoxia-tolerant species includes suppression of energy-dependent protein degradation systems, downregulation of certain HSP90 chaperone cycle proteins may be representative of a shift toward suppression of protein degradation to support metabolic depression and should be further studied [68].

Stress granules, which are mRNA-protein complexes that form molecular condensates, are another marker of cellular stress, particularly oxidative stress [69]. During a stressor, these granules are believed to store mRNA transcripts for housekeeping genes, thereby ensuring that stress-related transcripts are prioritized [70]. Unexpectedly, proteins related to stress granules decrease in abundance during anoxia in WS40NE cells, suggesting that stress granule formation may be suppressed (Fig. 10E). A lack of stress granules during anoxia would be predicted from these data, as a lower abundance of TIA1 and PUM2 would make it less likely for stress granules to form [71, 72]. A lack of stress granules could be adaptive given that small RNAs appear to be critical to the WS40NE anoxic response [28], and would be consistent with continued protein synthesis during anoxia. In anoxia-sensitive organisms, oxidation of TIA1 due to ROS is associated with inhibition of stress granule assembly and subsequent apoptosis [71]. Thus, the reduction in stress granule proteins observed here is contrary to a large body of evidence for the role of stress granules in supporting survival of cellular stress [71, 73, 74]. It is worth noting that in zebrafish embryos exposed to heat stress, stress granule formation is dynamic across brain regions [75], and thus there may be some variation in the importance of stress granule formation in stress tolerance. Nevertheless, given that stress granules are believed to store nonessential mRNA transcripts during stressors, it is surprising that proteins associated with their formation are not abundant by 4 days anoxia, and this may represent a unique adaptation of A. limnaeus cells for survival of anoxia.

DNA damage and cell cycle regulation

DNA damage is associated with hypoxic stress [76], and inappropriate cell division is a hallmark of ischemic dysregulation in mammalian cells [77]. Eight proteins associated with SUMOylation of DNA damage response and repair proteins are co-expressed and increase during anoxia and aerobic recovery (Fig. 10C). Though, notably, these are mostly nuclear pore proteins, DNA repair protein X-ray repair cross-complementing 4 (XRCC4) is co-expressed in this cluster with centrin 2 (CETN2). XRCC4 is a protein responsible for non-homologous DNA end joining (NHEJ) while CETN2 is a centrin protein involved in single-stranded DNA repair, specifically nucleotide excision repair (NER) via Xeroderma pigmentosum group C (XPC) [78, 79]. Together, these proteins represent major components of both single-stranded and double-stranded DNA repair. SUMOylation of both CETN2 and XRCC4 lead to nuclear localization, improving DNA repair [80, 81]. The co-expression pattern of these proteins supports an increasing DNA damage response due to reactive oxidation species [82].

The anoxic response to DNA damage may also involve p53, which among its many functions is involved in DNA damage repair and regulation of the cell cycle [83]. Stabilization of p53 leads to its upregulation in anoxia and recovery (log₂ fold-change 1.2 and 0.6, respectively) compared to normoxia. Given p53’s diverse roles, it is possible that its expression during anoxia is regulating cellular metabolism [84] or other vital functions such as cell cycle control. When DNA is damaged, p53 accumulates in the nucleus and can induce cell cycle arrest or, in extreme cases of damage, apoptosis [85]. Cell cycle arrest via p53 activation is mediated through its transcriptional target p21, which can cause G₁ phase arrest, or through direct p53 repression of the cdc25C promoter, which can then cause M or G₂ phase arrest [86, 87]. Given the increased abundance of p53 in our data, we would expect the cell cycle to be arrested in WS40NE cells during anoxia and recovery, yet WS40NE cells have been shown to divide for for several days during anoxia [26]. This claim is supported by increased abundance during anoxia of mitotic anaphase proteins involved in postmitotic nuclear pore complex reformation (LMNA, SEC13, NUP93, NUP107), the post-chaperonin tubulin folding pathway (MYH10, TUBB4B, TUBA1A, and TUBB4A), and sister chromatid cohesion proteins (MRE11, SMC1A, PDS5B) (Fig. 11).

Cell cycle progression is inconsistent with increased p53 abundance, suggesting that inhibition of a downstream effector of p53 may be modulating its role in cell cycle arrest. The p21 protein, though not captured by our methodology, is a probable effector in this pathway. Given that NUP98 stabilizes p21 transcripts and increases in relative abundance in the dataset, we would expect p21 transcripts to also be stabilized in anoxia [62]. Given the increased abundance of p53 and regulators of p21, future work should be done to understand the roles of p53 and p21 in WS40NE cells during anoxia. In particular, it is noted that ischemic preconditioning is marked by a decreased response of p53 and p21 in mammals [88]. A more anoxia-tolerant cell line, therefore, may exhibit unexpected phenotypes regarding p53 and p21. Additionally, p21 can be regulated by posttranslational modifications such as phosphorylation and ubiquitylation, which can effect both its stability and location within the cell [89]. Understanding how p21 is modified, as well as posttranslational modifications throughout the whole proteome at large, may be critical to understanding anoxia tolerance in WS40NE cells.

Caveats and limitations

The WS40NE cell line is a neuroepithelial cell line [26], and therefore it cannot necessarily be inferred that the observed proteomic response is this line is universal in all annual killifish cell lines or primary tissues, particularly as different cell types have been shown to have differential proteomic shifts in response to severe hypoxia [54]. Likewise, all pathway analyses depended on mapping A. limnaeus genes to human orthologs due to limited functional annotation resources for annual killifishes. Though this is standard practice in non-model organisms [90], this can obscure species-specific, divergent protein functions. Additionally, our workflow was optimized for global, unbiased detection of proteins, and therefore subtle changes to protein abundance may have fallen below the threshold of detection. Lastly, some proteins were not detectable using our bottom-up proteomics approach and were therefore not included in this analysis. Future studies should be employed using targeted assays to further confirm the presence or absence of specific subcellular structures, including stress granules.

Conclusions

This study has identified 2612 differentially expressed proteins (74.9%) in WS40NE cells that are responsive to changes in oxygen availability. Co-expression analysis of proteins identified in the dataset show increased abundance of key proteins, including many that suggest at the minimum differential stabilization of the mitochondrial compartment, and suggest the possibility of active mitochondrial metabolism and mitochondrial translation during anoxia. Further, proteins associated with protein refolding, Golgi-to-ER retrograde transport, and mitochondrial function increase in abundance during anoxia, suggesting the integrity of the proteome and avoidance of Golgi and mitochondrial stress are integral to the WS40NE anoxic response. Future work should investigate the activity of various processes that may have been impacted by changing capacity, such as cytoplasmic and mitochondrial translation. Additionally, future work should determine if similar proteomic changes occur at the organismal level with embryos of different developmental stages that vary in their anoxia tolerance.

Methods

WS40NE cell maintenance and anoxic exposure

Cells were grown in 10 ml of L-15 cell culture media (L-15 + 8.5% fetal bovine serum (FBS), 5 mM glucose, 100 U/ml penicillin/streptomycin) in 100 × 20 mm tissue culture dishes at 30 °C [26]. Prior to the experiment, cells were split when confluent using TrypLE Express (Thermo Fisher Scientific) [26]. Dishes (CytoOne™, USA Scientific, Ocala, FL) of cells were seeded at a density of 2.5 × 10⁶ cells/dish. For all experiments, cells were given 24 h to adhere to the dish. Samples that required anoxic exposure were introduced into an anoxic chamber (Bactron EZ, Sheldon Manufacturing, Cornelius, OR, USA) that contained an atmosphere of 95% N₂ and 5% H₂. Normoxic media was immediately replaced with 10 ml of anoxic media in the chamber, and the cells were incubated at 30 °C for the duration of the anoxic exposure. Anoxic media was created by purging normoxic media with N₂ gas for 30 min and then pre-equilibrating to the atmosphere of the anoxic chamber overnight as described previously [26]. Normoxic samples were grown for 4 days in normoxic media in an incubator (Sheldon Laboratories, Cornelius, OR) at 30 °C as previously described [26]. Cells were harvested after 3 experimental conditions: (A) 0 h of anoxia (normoxic), (B) 4 days in anoxia, and (C) after 1 day of aerobic recovery following 4 days of anoxia (n = 12 for each condition). Aerobic recovery was achieved by fully replacing anoxic media with 10 ml of normoxic media and returning the dishes to a normoxic incubator (Sheldon Laboratories, Cornelius, OR) at 30 °C for 24 h. Just prior to cell harvest, 1 ml of media was removed and stored at − 80 °C for later determination of extracellular lactate levels. To harvest cells, a cell scraper was used to gently dissociate cells from the bottom of plate into 0.5 ml of media. For each replicate, two 100 × 20 mm tissue culture dishes were pooled and pelleted by centrifugation in a 1.5 ml microcentrifuge tube at 500 × g for 7 min. Larger culture dishes could not be used due to incubator size constraints in the anaerobic chamber. Supernatants were removed and cell pellets were flash-frozen in liquid nitrogen. All samples were stored at − 80 °C.

Protein digestion

Frozen pellets were lysed and resuspended in 125 μl of 8 M urea/50 mM ammonium bicarbonate buffer. Samples were then centrifuged at 5000 × g for 2 min at room temperature (RT). Dithiothreitol (DTT) was added to a final concentration of 5 mM. Between each of the following steps, samples were briefly vortexed to ensure mixing and then centrifuged to collect the liquid prior to the next step. Samples were incubated for 30 min at 37 °C. Iodoacetamide (IAA) was added to a final concentration of 15 mM and the samples were incubated at RT for 30 min in the dark. To quench free IAA, DTT was added to a final concentration of 5 mM using a 75 mM stock. Protein concentration was determined using a bicinchoninic acid (BCA) assay (cat# 23,250, Thermo Scientific, Waltham, MA, USA) and 50 μg of protein were used for digestion from each sample. LCMS-grade water and 1 M ammonium bicarbonate (pH 8.5) was added to each sample to a final concentration of 50 mM ammonium bicarbonate for a total volume of 1.24 ml. A trypsin/Lys-C digestion (1:50 ratio enzyme to total protein concentration; cat #A41007; Thermo Scientific, Waltham, MA, USA) was used to generate tryptic peptides (< 3 kDa) [91]. Trypsin/Lys-C cleaves proteins at the carboxyl end of both lysine and arginine residues. After a 3-h incubation on a shaker at 37 °C, formic acid (FA) was added to a final concentration of 5% to stop digestion. Pierce C18 Spin Columns (cat #89,873, Thermo Scientific, Waltham, MA, USA) were used for peptide clean-up according to the manufacturer’s instructions. Due to max. cartridge binding capacity, 25 μg peptide recovery was assumed at a concentration of 0.5 μg/μl. The peptide solution was transferred into a new 1.5 ml microcentrifuge tube and dried using a SpeedVac (Thermo Savant, model# ISS110, Thermo Scientific, Waltham, MA, USA). Peptides were reconstituted in 100 μl of 1% formic acid in LCMS-grade water. The dissolved peptide solution was transferred into glass vials (Waters 600000669CV; Waters Corporation, Milford, MA, USA). Peptide concentrations were determined using 5 μl of each sample using a peptide assay according to the manufacturer’s instructions (Pierce Quantitative Peptide Assay, cat #23,290; Thermo Scientific, Waltham, MA, USA). Samples were stored at − 80 °C [92].

Liquid chromatography and mass spectrometry

A bottom-up proteomics approach [90, 92] was used to identify and quantify proteins using liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS) with nanoElute2/Impact II UHR-QTOF. Two microliters of total C18 purified peptide mix (100 ng/µL) were injected for each sample and resolved on a 25 cm Pepsep C18 column (Bruker) using a linear 3–33% acetonitrile gradient. Conventional data-dependent acquisition (DDA) was used to generate peak lists [93], and MSFragger (Fragpipe v22) [94, 95] was used to annotate peptides and create an MSMS spectral library using the A. limnaeus reference proteome database downloaded from the National Center for Biotechnology Information (A. limnaeus FASTA, txid52670 downloaded on 17May20223, 35,330 protein entries, NCBI) [94]. All database searches were done against this database containing an equal number (35,330) protein decoy sequences to determine FDR at < 1%. Each sample was then acquired again with a second LC–MS/MS run in data independent acquisition (DIA) mode for peptide quantification [96].

Differential expression analysis

Protein abundance was evaluated using Python code [97]. First, raw protein abundance data were normalized using quantile normalization. The normalized data were used to perform a principal component analysis (PCA) and to calculate log₂ fold-change data. Two-tailed t tests, corrected for multiple comparisons, were used to assess statistical significance for proteins across different time points with an adjusted p-value less than 0.05 considered significant [98]. Highly differentially expressed proteins were defined as those that were statistically significant and had a log₂ fold-change of ± 1 or greater.

Co-expression analysis

Gene co-expression patterns for all A. limnaeus proteins with human orthologs were generated using Clust (Version 1.18.1) to identify clusters (i.e., groups) of proteins that were consistently co-expressed across all three experimental conditions [38]. Human orthologs (EnsemblIDs) were used to maximize the usefulness of gene ontology analyses, given the large amount of information known for Homo sapiens genes compared to almost any other model organism [31]. Co-expressed protein clusters from the Clust analysis were uploaded to the STRING database (https://string-db.org/, Version 12.0) [99]. Expression patterns were analyzed using various overrepresentation or enrichment analyses to determine significant groups of proteins that were overrepresented using Fisher’s exact test with the Benjamini–Hochberg correction applied to account for multiple hypothesis testing [99]. These enrichment analyses included the KEGG orthology (KO, Kyoto Encyclopedia of Genes and Genomes https://www.kegg.jp/) [100], gene ontologies (GO, biological process, molecular function, cellular component) [101], reactome pathways [102], and human phenotype ontology (HPO) [103]. Interaction maps with predicted functional partners were created for enrichment pathways of biological interest. The top 15 GO terms with the highest overrepresentation (i.e., signal) of genes were evaluated for each cluster, and interaction networks with predicted function partners were created for terms of biological interest. Signal was chosen to order terms over false discovery rate (FDR) or observed/expected ratios because FDR tends to emphasize larger GO terms due to their potential for achieving lower p-values while observed/expected ratios highlight smaller terms, which have a high foreground to background ratio but cannot achieve low FDR values due to their size [99]. Signal, as a weighted harmonic mean between the observed/expected ratio and -log(FDR), balances these metrics [99]. Functional protein partners (default medium-confidence score threshold of ≥ 0.4) were predicted by independently scoring experimental data, curated pathway databases, co-expression, genomic context, text mining, and homology-based inference using a Bayesian framework [99]. Each GO term of interest was further subsampled with Markov cluster algorithm (MCL) clustering (inflation parameter 1.5) to identify functional clusters of contributing proteins and their predicted functional partners. The shown interaction networks were chosen based on which database best represented the process of interest. All interaction maps with predicted functional partners and dot plots were generated using the STRING database (https://string-db.org/, Version 12.0) [99]. All code is publicly available (https://github.com/hughcj11/WS40NE_Proteomics_Analysis).

Abbreviations

ATP

Adenosine triphosphate

mitosRNAs

Mitochondrial small noncoding RNAs

WS

Wourms’ Stage

DDX41

DEAD-box helicase 41

SF3B1

Splicing factor 3B subunit 1

SRSF2

Serine/arginine-rich splicing factor 2

Cul2

Cullin-2

UPR

Unfolded protein response

MRPS

Mitochondrial ribosomal small subunit

PDH3

Pyruvate dehydrogenase E1 component subunit beta

ACO₂

Aconitate hydratase

IDH3G

Isocitrate dehydrogenase [NAD] subunit gamma

OGDH

2-Oxoglutarate dehydrogenase

HAGH

Hydroxyacylglutathione hydrolase

SDHB

Succinate dehydrogenase [ubiquinone] iron-sulfur subunit

ME2

NAD-dependent malic enzyme

TARBP2

RISC-loading complex subunit TARBP2

PRKRA

Interferon-inducible double-stranded RNA-dependent protein kinase activator A

XRCC4

X-ray repair cross-complementing 4

CETN2

Centrin 2

NER

Nucleotide excision repair

XPC

Xeroderma pigmentosum group C

ROS

Reactive oxygen species

FBS

Fetal bovine serum

TCA

Trichloroacetic acid

DTT

Dithiothreitol

IAA

Iodoacetamide

BCA

Bicinchoninic acid

LCMS

Liquid chromatography mass spectrometry

DDA

Data-dependent acquisition

DIA

Data-independent acquisition

PCA

Principal component analysis

KO

KEGG orthology

KEGG

Kyoto Encyclopedia of Genes and Genomes

GO

Gene ontologies

HPO

Human phenotype ontology

FDR

False discovery rate

MCL

Markov cluster algorithm

Authors’ contributions

JEP and CH conceived and designed the study. CH and MMC performed the experiments and acquired the data. CH analyzed the data. DK and JEP contributed reagents/materials/analysis tools. CH and JEP wrote the paper. All authors read and approved the final manuscript.

Funding

Funded by National Science Foundation Grant IOS-2209383 and IOS-2025832.

Data availability

All DDA and DIA raw data are available at Panorama Public (https://panoramaweb.org/ch02jpodlab.ur) and ProteomeXchange (PXD074475, https://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD074475) [104, 105].

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.