Global maintenance of histone post-translational modifications during the transition into anoxia in embryos of the annual killifish Austrofundulus limnaeus

Abstract

Many organisms have adapted to survive anoxic or hypoxic environments, but the epigenetic responses involved in this successful stress response are not well described in most species. Embryos of the annual killifish Austrofundulus limnaeus have the greatest tolerance to anoxia of all vertebrates, making them a powerful model to study the cellular mechanisms necessary for anoxia tolerance. However, the global histone landscape of this species has never been quantified or explored in relation to stress tolerance. Liquid chromatography–mass spectrometry and a Python bioinformatics workflow were used to identify histones and their post-translational modifications. This pipeline resulted in the detection of 252 unique biologically relevant histone post-translational modifications (hPTMs) (unimod + residue). These PTMs represent 16 types of biologically relevant hPTMs present during both anoxia and normoxia in Wourms’ stage 36 embryos. This hPTM library presents an exciting opportunity to study histone modifications across development and in response to environmental stressors. No significant changes in PTM or histone abundance were observed between anoxic and normoxic embryos, suggesting that 24 h of anoxia is not sufficient to induce epigenetic or histone isoform changes at the organismal level. This result is inconsistent with data presented for similar stresses in mammalian cells and thus stabilization of the hPTM landscape may be an adaptation that supports anoxia tolerance.

Graphical Abstract

Graphical Abstract.

Introduction

Alteration of chromatin structure through histone post-translational modifications (hPTMs) is a mechanism of critical importance to both developmental gene regulation and control of gene expression in response to stress [1–5]. Embryonic development requires the accurate and specific expression of genes at the proper time and in the proper cells. Yet, under natural conditions, the vast majority of animal embryos are faced with an ever-changing environment that can impose stressful conditions such as extremes in temperature or oxygen availability. Exposure to stressful conditions disrupts typical gene expression patterns in favour of mounting a stress response to protect cellular structures and reestablish homeostasis [6, 7]. In the context of development, there is very little information on how global changes in hPTMs are altered in vertebrate embryos responding to stressful environments [5, 8, 9]. In this study, we use stress-tolerant embryos of the annual killifish Austrofundulus limnaeus to profile global patterns of hPTMs in response to anoxia to explore how hPTMs may be altered in response to extreme stresses encountered during development.

The histone landscape of the cell is critical to modulating gene expression [10, 11]. Histones are typically classified into five groups: core nucleosome histones H2A, H2B, H3, and H4 and linker histone H1, which stabilizes the nucleosome [12]. Histones are functionally differentiated into two classes of variants: replication-coupled (RC) and replication-independent (RI) variants [12, 13]. RC variants are expressed during the S phase of the cell cycle (DNA synthesis) and are typically encoded by genes that do not contain introns, while RI variants are expressed throughout the cell cycle and their genes usually contain introns [12, 13]. An organism can have multiple isoforms of each histone, including several non-canonical isoforms with distinct biological functions [14]. Diversified histone H2A variants are of particular interest due to their location in the nucleosome and subsequent ability to control access to the DNA. H2A/H2B dimers flank the central (H3/H4)₂ tetramer and therefore occupy the entry and exit positions of DNA from the core nucleosome [15]. Therefore, several RI isoforms of H2A have significant potential to impact gene expression due to their proximal position to DNA [16]. For example, phosphorylated histone H2A.X isoforms are critical for recruiting DNA repair enzymes to DNA double-strand breaks [17, 18]. Another histone H2A isoform, H2A.Z, is a transcriptional regulator that is also involved in chromatin remodelling [14, 19]. Lastly, macroH2A isoforms can stabilize gene expression patterns by inhibiting acetylation by p300 [20, 21]. Therefore, the diverse nature of histone isoforms can significantly shape the function of the histone landscape at any given time.

The histone landscape is further diversified by a variety of hPTMs that are critical in regulating the expression, replication, repair, and organization of DNA [22–24]. hPTMs are covalent modifications of amino acid residues on a protein, typically through the addition of a modifying group [25]. Common hPTMs include methylation, acetylation, phosphorylation, SUMOylation, proline isomerization, and ubiquitylation [26]. On histones, these modifications can affect transcription in two ways: (i) by changing the overall structure of chromatin or (ii) by regulating effector molecule binding. Either of these changes can expose or block exposure of the underlying DNA sequences to polymerases and other proteins, thereby modulating transcription [27]. In response to environmental stress, changes in hPTMs can occur quickly, within seconds to hours, and result in reprogramming of the chromatin [28–31]. Yet, very few studies have reported on global patterns of hPTMs in response to stress, and there is particularly little known about the role of hPTMs in supporting the extreme stress tolerance observed in some extremophile fishes [32–35].

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 changes in oxygen concentration on multiple timescales [36]. Continuous development of A. limnaeus is interrupted by up to three stages of developmental arrest and metabolic depression termed diapause I, II, and III [37]. Both developing and diapausing embryos can survive long bouts of the complete absence of oxygen (anoxia) by using anaerobic metabolic pathways [38, 39]. Embryos in Wourms’ stage (WS) 36, which occurs after 4 days of post-diapause II development at 25°C, are particularly interesting. WS36 embryos have a differentiated brain and heart (normoxic heart rate is ∼80 beats/min), and yet these embryos can survive, successfully recover, and complete development after enduring anoxia for hundreds of days (mean LT₅₀ = 74.3 days at 25°C) by entering a state of anoxia-induced quiescence [36, 40–42]. When faced with anoxia, there is an immediate downregulation of heat dissipation that is depressed by 96% within 12–16 h [40]. Further, levels of adenosine triphosphate (ATP) decline quickly and measures of cellular energetic status based on adenylate ratios suggest a severely reduced energetic status after 14 h of anoxia [40]. Despite these rapid declines in metabolism and ATP levels, a strong response is observed in the small non-coding RNA transcriptome after only 4 h of anoxia [43]. Thus, it appears that long-term survival of anoxia is supported by a rapid response that leaves the embryo in a state of metabolic depression with severely reduced ATP levels after 24 h of anoxia despite the ability to survive for months after initial exposure [40, 44, 45].

In order for embryonic cells to enter and exit metabolic depression, there must be extensive and immediate changes in gene expression [43]. As the most anoxia-tolerant vertebrate, embryos of A. limnaeus are a unique model for understanding molecular strategies that enable survival during anoxia [42]. While distinct patterns of global ubiquitylation and SUMOlyation have been identified in these embryos, there has never been an investigation of global hPTMs or histone isoform changes associated with anoxia [41]. Due to their rapid physiological response to anoxia and their severely restricted energetic status after the first few hours, we expect cells of A. limnaeus to quickly modulate hPTMs in response to decreased oxygen availability [29, 46]. Here, we present the first global analysis of hPTMs in response to anoxia, in an anoxia-tolerant vertebrate.

Materials and methods

Collection of embryos and anoxic exposure

Embryos were collected according to protocols approved by the PSU Institutional Animal Care and Use Committee (PSU protocol #33, IACUC protocol #81) and allowed to develop until diapause II at 25°C in the dark in plastic culture dishes (100 × 15 mm; VWR Scientific) [37, 50, 51]. Embryos were induced to break diapause II by exposing them to a temperature of 30°C in constant light for 48 h [41]. Embryos were observed daily and placed back at 25°C in the dark once they had broken diapause II. When the embryos reached WS36 (typically 4 days post-diapause II), they were exposed to anoxia in a Bactron III anaerobic chamber (Sheldon Manufacturing, Cornelius, OR, USA) in an atmosphere of 5% hydrogen, 5% carbon dioxide, and 90% nitrogen gas. Once inside the chamber, the embryos were immediately placed into anoxic medium prepared by first bubbling the media with compressed nitrogen gas for 30 min and then equilibrating the media with the atmosphere of the chamber for 24 h [41]. Embryos were exposed to anoxia for 24 h at 25°C as previously described [41, 42]. WS36 embryos can survive for months of anoxia at 25°C, and thus 24 h of anoxia was chosen to allow for sufficient time to characterize the initial and likely most critical responses to anoxia, while ensuring 100% survival [41, 42]. Further, this will allow the hPTM data to be compared to transcriptomic data from previous studies collected at 24 h of anoxia [43]. Normoxic embryos were sampled from the same batch of embryos and harvested at the same developmental stage to serve as controls.

Isolation of embryonic cells

Cells from control and anoxia-treated embryos were harvested and separated from the yolk by mechanical separation as previously described for A. limnaeus [41]. This method allows for cells to be quickly separated from the chorion and yolk components. Groups of 40–45 embryos and 1 ml of ice-cold Yamamoto’s solution [52] (0.75% NaCl, 0.02% KCl, 0.02% CaCl₂, 0.002% NaHCO₃, pH = 7.3) containing protease and phosphatase inhibitors (Pierce A32959, ThermoFisher Scientific, Waltham, MA, USA) were loaded into a 15-mm NetWell screen (500 μm mesh size, catalogue number 3478, Corning Life Sciences, Lowell, MA, USA) mounted on top of a 15 ml conical vial with a collar of Tygon tubing. Cells were mechanically separated by grinding the embryos through the screen with a flat glass pestle. The screen was washed twice with 1 ml of ice-cold Yamamoto’s solution with inhibitors as earlier, and the embryonic material forced through the screen with the pestle. The cells were pelleted in the tube by spinning at 100× g for 7 min at 4°C. The supernatant containing the yolk was removed and the cell pellet resuspended in 1 ml of Yamamoto’s solution prior to transfer into a preweighed 1.5 ml low protein binding microcentrifuge tube (ThermoFisher Scientific, Waltham, MA, USA). The cells were pelleted by centrifugation at 500× g for 5 min at 4°C. The supernatant was removed, and the pellets were flash frozen in liquid nitrogen. Final tissue pellets ranged in mass from 9.9 to 12.9 mg. Six groups of pooled embryos (40–45 embryos) were processed for each anoxic and normoxic replicate (n = 6). Samples were stored at −80°C until histone acid extraction.

Histone acid extraction and protein digestion

A middle-down approach to proteomics was used for liquid chromatography–mass spectrometry (LCMS) using previously described methods [53, 54]. To enrich the samples with histones, a histone acid extraction was performed by resuspending cell pellets in 400 μl of 0.4 N H₂SO₄ (40× dilution) and incubating the samples with constant rotation at 4°C for 4 h. Samples were then centrifuged at 16 000× g for 10 min at 4°C, and the supernatant was transferred into a fresh 1.5 ml tube. This process was repeated twice to create the histone acid extract. Trichloroacetic acid was added to each sample drop by drop until a volume equal to one fourth of the sample volume had been added. Samples were inverted to mix and incubated overnight at 4°C. The precipitated histone proteins were centrifuged at 3400× g for 5 min at 4°C and the supernatant was discarded. The remaining pellet was rinsed twice: first in ice-cold acetone with 0.1% HCl and then with 100% acetone. For both rinses, the samples were centrifuged at 3400× g for 2 min at 4°C to remove the supernatant. After both rinses were complete, histone pellets were air-dried in a fume hood for 5 min. Samples were stored at −80°C until the in-solution digestion.

For digestion, each sample was brought to room temperature and the histone pellet was resuspended in 138 μl of 8 M urea containing 10 mM dithiothreitol to denature and reduce proteins. Samples were briefly vortexed and centrifuged, before incubation at 37°C for 30 min. Iodoacetamide (IAA) was added to a final concentration of 30 mM to alkylate reduced cysteine residues and prevent protein refolding. Following the addition of IAA, samples were briefly vortexed and centrifuged before a 30 min incubation in the dark at room temperature. Samples were briefly vortexed and centrifuged again after the incubation and a bicinchoninic acid protein assay (Thermo Scientific, catalogue number 23250) was performed to determine protein concentration. A total of 50 μg of protein for each sample was diluted 5:6 (v:v) in sodium phosphate buffer (pH 7.8). V8 protease (bead-conjugated; Thermo Scientific, catalogue number 20151) was added at a ratio of 1 μg protease to 50 μg total protein. V8 protease in sodium phosphate buffer cleaves proteins at the carboxyl end of both glutamate and aspartate residues [55]. After a 20 h incubation on a rotator at 35°C, samples were centrifuged at 500× g for 2 min to pellet the protease beads. The supernatant was transferred to a new tube and centrifuged at 19 000× g for 5 min. After each centrifugation, the supernatant containing peptides was transferred into a new 1.5 ml tube and concentrated using a SpeedVac (Thermo Savant, model ISS110) until all of the buffer had evaporated. Peptides were reconstituted to a final peptide concentration of 333 ng/μl in LCMS-grade water containing 0.1% formic acid. The final peptide solutions were cleaned over a C18 spin column (Thermo Scientific Pierce C18 Spin Columns, Thermo Scientific, catalogue number #89873).

Liquid chromatography and mass spectrometry

Conventional data-dependent acquisition (DDA) was used to identify proteins and generate spectral libraries [56], and PEAKS Suite X Plus and MSFragger were used to annotate histone peptides using the A. limnaeus reference proteome database [57]. Each sample was then analysed by a second LCMS run in data-independent acquisition (DIA) mode to quantify individual peptides as previously described [58].

Data and statistical analyses

A preliminary search of the A. limnaeus genome [59, 60] for the keyword ‘histone’ was used to identify several histone proteins. Each of these proteins was used in a BLAST search [61] of the A. limnaeus genome to identify proteins with similar sequences. Any new histone proteins identified by this method were then entered into an additional BLAST search until no novel histones or histone-like sequences were identified.

A list of 27 possible PTMs with known mass shifts listed in the Unimod database for processing with V8 protease was used to interrogate peptides for post-translational modifications [62] using Skyline software (Version 22.2). This list contained the following potential modifications: acetylation, ADP-ribosylation, amidation, biotinylation, carbamidomethylation, carbamylation, lactyl/carboxyethylation, carboxymethylation, sodiated (cation: Na), deamidation, dehydration, methylglyoxal-derived carboxyethyllysine (delta H(4)C(3)O(2)), dimethylation, dioxidation, Gln-pyr-Glu, Gly, 4-hydroxynonenal, hydroxamic acid, methylation, oxidation/hydroxylation, phosphorylation, propionamide, radical anion, trimethylation, and ubiquitylation (GlyGly). The absence of other PTMs does not indicate that they are not present in the data, but rather that they were not detectable using our methods.

A list of hPTMs was compiled according to the histone protein name, amino acid residue/position, and modification type using Python code. All hPTMs detected were evaluated for their biological relevance according to their Unimod accession and amino acid residue. Modifications were considered biologically relevant if the Unimod accession on the given amino acid residue was classified as ‘Post-translational, Other, or Multiple’. Seventy-four additional hPTMs were classified as ‘Unspecified’ because the amino acid on which they were detected was not given a classification on Unimod.org. These were originally included in the data to determine whether any significant patterns were observed. However, they were removed once they were deemed insignificant or found to be possible artefacts of processing. Modifications that were classified as ‘Artefact’ or ‘Chemical Derivative’ were not considered biologically relevant and were immediately removed from analysis. Modifications that could not be differentiated from one another by mass shift were listed together with a ‘/’. For example, lactylation and/or carboxyethylation are listed as lactylation/carboxyethylation because they cannot be differentiated in this dataset due to their identical mass shift (+72.021129).

Histone peptide abundance, defined as the normalized area under the curve, and raw protein abundance were obtained through Skyline (Version 22.2) (Fig. S1) [63]. Python code was developed and used for all hPTM calculations and statistics, utilizing the packages pandas and NumPy [64, 65]. First, the relative abundance of each hPTM was calculated, which is defined as the percentage of specific amino acid residue on a histone occupied with the PTM. From the relative abundance, the β-value and M-value were calculated. The β-value is the relative abundance displayed as a number between 0 and 1, with 100 added to the denominator to standardize low intensity peptides [66]. The M-value, which is a logit transformation of the β-value [67], was used to perform a principal component analysis (PCA). To evaluate significance, the relative abundance of each hPTM was used to calculate the log₂ fold change, and a two-tailed t-test was performed with the Benjamini–Hochberg correction applied to account for multiple hypothesis testing [68]. This analysis was repeated to analyse global PTM levels across all identified histones in addition to individual histone residues. Lastly, the global relative coverage of each modifiable amino acid residue was calculated to determine what percentage of each PTM type was present at each modifiable residue. For each histone protein, hPTM maps were constructed for anoxic and normoxic samples using the mean relative abundance of hPTMs. All abundance plots were made in R using the package ggplot2 [69].

Protein abundance was also evaluated using Python code. The log₂ fold change (log₂FC) was calculated from the raw protein abundance data and used to perform a PCA. A two-tailed t-test was performed with the Benjamini–Hochberg correction applied to account for multiple hypothesis testing [68], to evaluate the statistical significance of the log₂FC values.

Global hPTM abundance, protein relative abundance, and global relative residue coverage were visualized using GraphPad Prism version 10.0.0. All statistics (t-statistic, P-value, and corrected P-value) for each hPTM (Table S2) and all Python and R code are available on GitHub at https://github.com/hughcj11/Histone_PTM_Analysis.

Multiple sequence alignment

To visualize the conservation between histone isoforms of common model organisms and A. limnaeus, a search for isoforms was conducted on Histone Database 2.0 (HistoneDB 2.0) for A. limnaeus, Danio rerio, Mus musculus, and Homo sapiens [70]. The peptide sequences of all histones with NP and XP IDs were downloaded as FASTA files from HistoneDB 2.0 [70]. Jalview (Version 2.11.3.3) web services were used to conduct a multiple sequence analysis using ClustalOmega to visualize the alignment for all histone classes (H1, H2A, H2B, H3, and H4) [71, 72]. Relevant protein structures were visualized in Chimera [73].

Results

Histone protein expression patterns

Forty histone open reading frames, including 38 genes encoding for histone isoforms and 2 pseudogenes, were identified in the A. limnaeus genome assembly (Table S1). Of the 38 isoforms, 16 unique histone groups were identified in the mass spectrometry data. In total, 597 histone-specific peptides were identified based on their unique protein accession number. The 16 proteins were representative of histone H1, H2A, H2B, and H3 isoforms (Table S1). No H4 isoforms were detected. Isoforms had varying degrees of sequence coverage, ranging from 11% to 87% (Fig. 1). No significant changes in histone isoforms were detected between normoxic and 24 h anoxic embryos (P   >  0.05).

Figure 1.

Raw abundance of histone isoforms in WS36 embryos of A. limnaeus. Sixteen distinct histone protein groups were identified based on their unique protein accession number and unique peptides. (A) Histone abundance during anoxia (green) and normoxia samples (blue). Bars are means ± SEM (n = 6). Pie charts represent the sequence coverage as defined by the total detected sequence length compared to the total protein length, for each protein (purple). Due to different peptide ionization efficiencies, histone isoform abundance can only be compared across conditions and not within a sample. (B) Histone protein abundance is comparable in normoxia and after 24 h of anoxia as demonstrated by the large overlap of log₂ expression values of all histone isoforms on a PCA plot. Anoxic (green) and normoxic (blue) samples are plotted according to PC1 and PC2 with a 95% confidence ellipse surrounding each cluster (n = 6). (C) Representative images of WS36 embryos [37].

Global hPTM patterns

Out of 478 unique hPTMs detected, 252 unique hPTMs (unimod + residue) were considered biologically relevant, representing 16 types of biologically relevant PTMs (Fig. 2). These types of modifications included commonly studied modifications such as methylation, acetylation, phosphorylation, and ubiquitylation, and less studied modifications such as lactylation and 4-hydroxynonenal. Globally, hPTMs do not cover the majority of the modification sites that are theoretically available. The global relative abundances of the three most abundant hPTMs ranged from only 12.3% to 16.2% (Fig. 3). However, hPTMs that compete for residues often showed very different relative abundances. For example, acetylation, lactylation/carboxyethylation, and ubiquitylation can all occur on lysine residues, and under normoxic conditions, they have a global relative abundance of 6.8%, 4%, and 1.1%, respectively. Lysine residues were the most modified (31%) and had the greatest diversity of PTMs (Fig. 3), with 11 types of PTMs and as many as 6 different PTMs competing for a single lysine residue.

Figure 2.

Global hPTMs detected in normoxic and anoxic embryos. (A) The M-values of all hPTMs were used to construct a PCA plot illustrating the similarity in the hPTM landscape between normoxic and anoxic embryos. Anoxic (green) and normoxic samples (blue) are plotted according to PC1 and PC2 with a 95% confidence ellipse surrounding each cluster (n = 6). (B) Relative abundances of all 17 classes of hPTMs across all histones identified in this study. Relative abundance is defined as the percentage of histones where the specific amino acid residue is occupied with the PTM. Bars are means ± SD (n = 6).

Figure 3.

Global relative amino acid residue coverage by hPTM type during normoxia. Seventeen types of hPTMs were detected across 10 types of amino acid residues. Lysine residues had the highest number and diversity of modifications with ∼35% of the residues modified. All other residues had relatively low abundances of PTMs As there were no significant changes in residue coverage in response to anoxia, only normoxic global relative amino acid residue coverage is shown.

All hPTMs were present in both anoxic and normoxic samples and no significant changes in global levels of hPTMs were identified. The PTMs with the highest relative abundance across the global histone landscape were phosphorylation, dehydration, and acetylation. Seventeen unique hPTMs were highly abundant during both anoxia and normoxia, as defined by an average relative abundance >50%. Of the high abundance hPTMs, nine unique hPTMs displayed a relative abundance of exactly 100% across all samples, meaning that specific residue was always modified by that PTM. Five of these nine sites were lysine residues, each with a different PTM, including lactylation/carboxyethylation. Seven histone isoforms had at least one hPTM site with 100% relative abundance, representing H1, H2A, and H2B (Table 1).

Table 1.

hPTMs with 100% relative abundance.^a

Protein accession	Protein description	Amino acid + position	PTM description
XP_013857090.1	H2B_1/2-like	42K	Carboxyethyl/lactylation
XP_013878240.1	H1-like	31P	Oxidation/hydroxylation
XP_013854970.1	H2A.V	10S	Ubiquitylation
XP_013855658.1	H1-like	102K	Acetald + 28/dimethylation
XP_013879563.1	H1-like	40K	Methylation
XP_013885354.1	H2B.L4-like	66T	Dehydrated
XP_013857090.1	H2B_1/2-like	45K	4-Hydroxynonenal
XP_013854970.1	H2A.V	12K	Acetylation
XP_013887205.1	H2AX-like	60T	Ubiquitylation

^aNine hPTMs displayed an average relative abundance of exactly 100%, meaning that specific residue was always modified by that PTM. Seven histone isoforms had at least one hPTM site with 100% relative abundance, representing H1, H2A, and H2B.

Characterization of histone H1

Three H1 proteins were identified: H1-like (XP_013878240.1), H1-like (XP_013879563.1), and H1-like (XP_013855658.1). Twelve types of PTMs were identified across these isoforms; however, no PTM was present on all three isoforms (Fig. 4). Histone H1-like (XP_013878240.1) 31P oxidation/hydroxylation and H1-like (XP_013855658.1) 102K acetald + 28/dimethylation were detected at 100% relative abundance in all samples. Additionally, H1-like (XP_013879563.1) 46P dioxidation had high abundance in both anoxic and normoxic conditions. All other PTMs occurred in low abundance.

Figure 4.

Characterization of the histone H1 isoform post-translational modifications. (A) H1-like (XP_013878240.1), (B) H1-like (XP_013879563.1), and (C) H1-like (XP_013855658.1). Each panel represents a different amino acid position where at least one hPTM was detected. Amino acid residues are abbreviated using their one-letter code. The first methionine of the peptide sequence is included in the amino count as number 1. The x-axis displays the condition of the samples: anoxic (A) or normoxic (N).

Characterization of histone H2A

Six H2A proteins were identified: H2A-like (XP_013879560.1), H2A (XP_013874651.1), H2A-like (XP_013878193.1), H2A.V (XP_013854970.1), H2A.Z (XP_013858796.1), and H2AX-like (XP_013887205.1). Thirteen types of PTMs were identified across these isoforms and no single PTM was present on all six isoforms (Figs 5 and 6). The following three PTMs displayed an average relative abundance of exactly 100% in all samples: H2A.V (XP_013854970.1) 10S ubiquitylation and 12K acetylation and H2AX-like (XP_013887205.1) 60T ubiquitylation. Additionally, the following hPTMs had high abundance in both anoxic and normoxic conditions: H2A-like (XP_013878193.1) 51Y and 102S dehydration; H2A-like (XP_013879560.1) 58Y dioxidation; and H2AX-like (XP_013887205.1) 77T phosphorylation.

Figure 5.

Characterization of the post-translational modifications on histone H2A isoforms. Isoforms represented are (A) H2A-like (XP_013879560.1), (B) H2A (XP_013874651.1), and (C) H2A-like (XP_013878193.1). Each panel represents a different amino acid position where at least one hPTM was detected. Amino acid residues are abbreviated using their one-letter code. The first methionine of the peptide sequence is included in the amino count as number 1. The x-axis displays the condition of the samples: anoxic (A) or normoxic (N).

Figure 6.

Characterization of histone H2A isoform post-translational modifications. (A) H2A.V (XP_013854970.1), (B) H2A.Z (XP_013858796.1), and (C) H2AX-like (XP_013887205.1). Each panel represents a different amino acid position where at least one hPTM was detected. Amino acid residues are abbreviated using their one-letter code. The first methionine of the peptide sequence is included in the amino count as number 1. The x-axis displays the condition of the samples: anoxic (A) or normoxic (N).

Characterization of histone H2B

Three H2B proteins were identified: H2B_1/2-like (XP_013857090.1), H2B.L4-like (XP_013885354.1), and H2B_1/2 (XP_013879561.1). A fourth low-quality H2B protein, H2B.3-like (XP_013886452.1), was also identified by a single unique peptide sequence. As this single sequence was unmodified, H2B.3-like (XP_013886452.1) was not included in any calculations. Fifteen types of PTMs were identified across these isoforms. Phosphorylation, methylation, dimethylation, and oxidation/hydroxylation were present on all isoforms (Fig. 7). The following three PTMs displayed an average relative abundance of exactly 100% in all samples: H2B_1/2-like (XP_013857090.1) 45K 4-hydroxynonenalytion, H2B 1/2-like (XP_013857090.1) 42K carboxyethylation/lactylation, and H2B.L4-like (XP_013885354.1) 66T dehydration. H2B_1/2 (XP_013879561.1) 114T dehydration had high abundance in both anoxic and normoxic conditions.

Figure 7.

Characterization of histone H2B isoform post-translational modifications. (A) H2B_1/2-like (XP_013857090.1), (B) H2B.L4-like (XP_013885354.1), and (C) H2B_1/2 (XP_013879561.1). Each panel represents a different amino acid position where at least one hPTM was detected. Amino acid residues are abbreviated using their one-letter code. The first methionine of the peptide sequence is included in the amino count as number 1. The x-axis displays the condition of the samples: anoxic (A) or normoxic (N).

Characterization of histone H3

Three H3 proteins were identified: H3-like (XP_013879564.1), H3.3 (XP_013887037.1), and H3-like (XP_013887218.1). Fourteen types of hPTMs were identified across these isoforms. Deamidation was present on all six isoforms (Fig. 8). H3-like (XP_013887218.1) 17K acetylation had high abundance in both anoxic and normoxic conditions.

Figure 8.

Characterization of histone H3 isoform post-translational modifications. (A) H3-like (XP_013879564.1), (B) H3.3 (XP_013887037.1), and (C) H3-like (XP_013887218.1). Each panel represents a different amino acid position where at least one hPTM was detected. Amino acid residues are abbreviated using their one-letter code. The first methionine of the peptide sequence is included in the amino count as number 1. The x-axis displays the condition of the samples: anoxic (A) or normoxic (N).

Discussion

Characterization of the hPTM landscape in A. limnaeus embryos

This paper is the first global characterization of histone isoform expression and hPTMs in any annual killifish embryo, and in any vertebrate embryo in response to anoxia. While many studies identify hPTMs via antibody detection [32, 74, 75], including in A. limnaeus embryos [47], these methods often cannot distinguish between histone isoforms. Additionally, many well-described hPTMs have been identified in humans or other mammalian models, and these antibodies may not accurately reflect the hPTM location on histones from other species (Table 2; Figs S2–S8). Thus, mass spectrometry offers a way to identify hPTMs in an isoform-dependent manner, providing a rich source of data to characterize the hPTM landscape [53]. The hPTMs varied across isoforms by type and quantity. For example, H3-like (XP_013879564.1) had only 2 biologically relevant hPTMs detected while H3-like (XP_013887218.1) had 14, suggesting isoform-specific modifications. Importantly, we have identified nine previously unknown hPTMs with 100% relative abundance that may be developmental markers of WS36 embryos, critical to cell function in this species, or perhaps are associated with the extreme anoxia tolerance exhibited by these embryos (Table 1). Development-specific hPTMs are mostly unknown in embryos of A. limnaeus, and future work should investigate the expression and function of these hPTMs across development [47]. This is particularly important, as many of the hPTM classes detected in this study (acetylation, methylation, lactylation, and hydroxylation) are known to regulate gene expression in other species [74, 75, 76, 77]. If these are developmental markers, they would represent unique developmental hPTMs, as most embryonic developmental markers described in fish species to date occur on H3 and H4 and are limited to acetylation and methylation [47, 78, 79].

Table 2.

Corresponding A. limneaus amino acids by isoform of known, biologically relevant hPTMs.^a

Known hPTM	Corresponding A. limneaus amino acid by isoform			Biological relevance
	XP_013879564.1	XP_013878241.1	XP_013870443.1
H3K36me3	K36	K150	K146	Human hypoxia marker [29]
H3K27me1	K27	K141	K137	Embryonic A. limneaus developmental marker [47], upregulation of gene expression [48]
H3K27me2				Embryonic A. limneaus developmental marker [47]
H3K27me3				Marker of inactive genes [27, 49]
H3K4me1	K4	K118	K114	Decreased in post-DII A. limneaus embryos [47]
H3K4me3				Embryonic A. limneaus pre-DII marker [47], human hypoxia marker [29]

^aThe first methionine of the peptide sequence is included in the amino count as number 0 for all peptides to be consistent with the numbering of the known PTMs.

Caveats and precautions in the interpretation of these data

No histone H4 isoforms were detected, and this underrepresentation may be a methodological limitation of the V8 digestion of histones in A. limnaeus. For example, in Mozambique tilapia (Oreochromis mossambicus), histone H4 isoforms were identified in acid extracts as were prepared in this study, but only after using multiple types of enzyme digestions to produce a variety of peptides for each histone [53]. Further, the genome assembly for A. limnaeus contains three histone H4 loci that all encode for the exact same amino acid sequence, which could potentially produce five peptides of the correct size for this analysis pipeline (Table S1 and Fig. S9). Therefore, the lack of histone H4 isoforms in our analysis is very likely due to the ionization efficiency and the particular biochemical nature of the peptides produced from histone H4 isoforms in A. limnaeus. In the future, it may be prudent to use multiple digestion methods, such as the addition of V8 protease in ammonium bicarbonate, to generate different peptides and thus potentially expand the hPTM library to provide better coverage of histone H4 [53].

Importantly, WS36 embryos are composed of multiple tissues, and thus our results are an average of the responses in all cell types. It is possible, and in fact highly likely, that hPTMs are tissue- or cell type-specific, and thus our ability to observe these changes is potentially limited when looking at whole embryo global patterns [10, 80]. Even within the same tissue type, local changes on specific genes may be quantitatively too small to be captured when analysing the global hPTM landscape, although highly pervasive changes can be detected [81]. Additionally, hPTMs change significantly during embryonic development of teleost fishes, making it unclear whether the hPTMs detected in this study are representative of only WS36 embryos, or are possibly species-specific [82, 83]. Future work should be conducted on individual cell or tissue types. For example, PSU-AL-WS40NE is a neuroepithelial cell line that was isolated from an embryonic A. limnaeus tissue explant [84]. These cells can survive for 49 days without oxygen, and they show similar responses to anoxia at the molecular level when compared to embryos [84–86]. Thus, this cell line may provide insight into how cells can utilize hPTMs to successfully mitigate survival and recovery from stresses such as anoxia. With these limitations in mind, there are still many interesting findings in this study that suggest a different response to oxygen limitation in cells of A. limnaeus compared to mammals.

Austrofundulus limnaeus embryos maintain a stable hPTM landscape during initial exposure to anoxia

The lack of any global anoxia-induced changes in the hPTM landscape was somewhat surprising given the extreme nature of exposing cells to anoxia, and the large changes in gene expression that are known to accompany this initial transition [85, 87]. This is also surprising given that in other species many hPTMs respond quickly to stressors, suggesting that 24 h should have been sufficient time to induce an epigenetic response [28–31]. A delayed or absent response in histone modifications may suggest that A. limnaeus embryos are primed for survival of anoxia. Thus, it may be that the nine hPTMs with 100% relative abundance in this study are key to mediating the initial response to anoxia and must be present prior to exposure to ensure proper function. It is also possible that survival of anoxia does not depend on hPTMs, but rather on other mechanisms such as total RNA stabilization and differential abundance of small non-coding RNAs [43, 87]. Finally, 24 h of anoxia may not be sufficient time to induce global epigenetic changes to the hPTM landscape, and longer time points may yield interesting results. A delayed hPTM response to anoxia would be consistent with the salinity response of the Mozambique tilapia (O. mossambicus), which complete their organismal acclimation before hPTMs respond on a global level [81]. In addition, we cannot rule out that changes in the hPTM landscape were occurring, but we did not capture them with our methodology given that coverage of many histone isoforms is highly variable and sometimes low. Future studies of hPTMs in response to anoxia across different developmental stages may help to clarify the potential importance of histone isoforms and modifications in mediating survival of anoxia.

The stable hPTM landscape of A. limnaeus embryos exposed to anoxia is in sharp contrast to human cells, which experience global changes in histone methylation within 1 h of hypoxic exposure, including upregulation of histone modifications associated with active gene transcription [29]. In a variety of human cells, H3 and H4 histones are particularly modified in response to hypoxia, and H2A.X histones are modified differentially across cell types [46]. Thus, the stable hPTM landscape in A. limnaeus embryos may be a mechanism to avoid unwanted or dysregulated changes in gene expression. Specifically, histone-3 lysine-4 trimethylation (H3K4me3) and H3K36me3 occurs in HeLa cells in response to hypoxia [29], while H3-like (XP_013887218.1) trimethylation remains stable in anoxic A. limnaeus cells. However, while H3 histones are highly conserved across model species, canonical H3s of A. limnaeus (H3-like XP_013879564.1 and H3-like XP_013887218.1) have longer peptide sequences than human and murine canonical H3s (Figs S6 and S7). Therefore, human H3K4 and H3K36 best corresponds (by sequence similarity) to more interior residues in A. limnaeus isoforms due to a long N-terminal extension (Figs S6 and S7). Thus, while the same region of the core conserved histone is methylated in the same way in mammals and killifish embryos, it is not clear whether they serve the same function, nor is it clear whether a longer N-terminal tail contributes to a change in histone structure (Fig. 9). Unfortunately, the shorter H3 isoform in A. limnaeus (H3-like, XP_013878241.1) does not appear in this dataset, nor do the relevant residues on the H3-like (XP_013887218.1) protein, so we cannot explore the potential conservation of function with other species in these shorter histone isoforms. Future work should be done to investigate these regions on shorter isoforms. A longer N-terminal peptide sequence could be an adaptation that supports extreme anoxia tolerance of WS36 embryos [42, 45] by preventing changes in methylation that are observed in anoxia-sensitive mammals.

Figure 9.

Comparison of H3K36me3 sites in Homo sapiens and A. limnaeus. The H3 peptide sequences of H. sapiens NP_003 520 (blue) and A. limnaeus XP_013878241.1 (beige) were visualized in Chimera. Based on multiple sequence alignment, the lysine site (red) of H3K36me3 in H. sapiens occurs at K150 on this A. limnaeus isoform.

Histone ubiquitylation is a proteomic hPTM that warrants additional discussion. Twenty-five unique sites of ubiquitylation were identified in our dataset, including nine on H2B histone isoforms. Ubiquitylation is a reversible modification that attaches one (monoubiquitylation) or more (polyubiquitylation) ubiquitin proteins to a protein. Previous studies on WS36 and WS40 embryos of A. limnaeus exposed to longer-term anoxia identified a specific 23 kDa ubiquitylated protein that is absent during anoxia but present in both normoxia and aerobic recovery from anoxia [41]. Based on the molecular weight of ubiquitin, it is likely that monoubiquitylation is occurring in this situation. The ubiquitylated protein, which is likely about 14 kDa, is suspected to be a histone (H2B), suggesting that the ubiquitylation and deubiquitylation of histones may play a vital role in long-term anoxia tolerance [41]. The lack of any specific changes in histone ubiquitylation or protein expression may be due to the short time course of exposure in this study compared to previous studies. These modification sites should be investigated during long-term anoxia. This also highlights the need for hPTMs to be quantified for additional embryonic stages, including WS40 and diapausing embryos.

Several of the hPTMs detected also provide interesting avenues for further study beyond methylation. For example, lactylation may be a key histone modification during anoxia in mammalian species [88]. Lactate is a glycolytic end product often accumulated when oxygen is limiting. Anoxic embryos of all stages of A. limnaeus accumulate mM quantities of lactate [39, 45]. Anoxic cells derived from A. limnaeus WS40 embryos (WS40NE) accumulate extracellular lactate as well, but at a significantly lower rate compared to anoxia-sensitive mammalian cell lines [84]. This lactate buildup has been hypothesized as a symptom of substrate exhaustion in the mammalian cell lines, but lactate production may have utility in A. limnaeus embryos in the form of histone lactylation [89]. For example, histone lactylation stimulates gene transcription in macrophages but remains severely understudied in anoxia-tolerant animals [76]. Interestingly, lactylation may also block other PTMs from occurring. In astrocytes, NDRG2 lactylation was found to prevent ubiquitylation at Lys176, causing upregulation of NDRG2 and decreasing astrocytic inflammation during hypoxia [90]. This suggests a competing relationship on some proteins between lactylation and ubiquitylation in response to oxygen stress in mammalian cells. We identified three specific lysine histone residues where lactylation/carboxyethylation and ubiquitylation were both present in A. limnaeus: (XP_013879561.1) K107, (XP_013879561) K115, and (XP_013855658.1) K15. However, even the most modified of these three lysines, (XP_013879561) K115, was only modified ∼56% of the time, suggesting that lysine residue availability may not be the primary driver of this competition. Alternatively, this competition may be occurring in only certain populations or subpopulations of cells and these changes are masked by our global characterization approach. As lactylation and ubiquitylation are both attached enzymatically by p300 and multiple E3 ligases, respectively [76, 91–93], this suggests that enzyme activity or differential expression may also play a major role in driving this competition more than site availability. The activity balance of these enzymes is worthy of future investigation.

There is a need to better understand stress-related hPTMs in non-model organisms

While the hPTM landscape of model organisms continues to be catalogued [94, 95], the hPTMs of stress-tolerant, non-model or emerging model organisms are often unexplored despite evidence that epigenetic changes occur in response to stressors that may support stress tolerance [96]. Many stress-related hPTMs are described in model organisms as biomarkers of stressors or disease, including cancer-specific modifications [97–100]. Stress-related PTMs and hPTMs such as oxidations and deamidations are associated with aging in human tissues [101–104]. Proline hydroxylation throughout the proteome is critical in oxygen sensing via the Hypoxia-inducible factor (HIF) hydroxylase system [105, 106]. In addition, in mammalian histones, the rapid loss of H3P16 hydroxylation during hypoxia appears to act as an HIF-independent oxygen sensing pathway that regulates gene expression [75]. Lastly, 4-hydroxynonenal, a reactive end product of lipid peroxidation associated with oxidative stress, is an hPTM believed to increase exposure of DNA to oxidative damage in Alzheimer’s disease [107, 108]. Oxidation/hydroxylation, deamidation, and 4-hydroxynonenal are all hPTMs found in A. limnaeus embryos during normoxic conditions, as well as the tissues of other fish [53]. However, it would be expected that these hPTMs would increase during anoxic stress due to increasing reactive oxygen species, yet we observe no significant global changes in the levels of these hPTMs

The stability of hPTMs in A. limnaeus embryos, particularly in response to stress, may be due to their life history and the highly variable nature of their environment [109]. Development is essentially an epigenetic process, with a variety of epigenetic mechanisms ensuring the proper expression of genes in space and time [110]. Thus, certain aspects of epigenetic gene regulation must be protected and conserved over organismal and evolutionary timescales [111, 112]. While transgenerational plasticity of hPTMs is a beneficial evolutionary strategy in some conditions, individual plasticity of hPTMs may be a more appropriate form of adaptive plasticity for species in variable environments [113]. Because A. limnaeus embryos experience anoxic stress as a normal part of their development, they may have biological mechanisms that ensure epigenetic stability and therefore developmental stability, by preventing the changes observed in anoxia-sensitive species, which may indeed be maladaptive [114]. As A. limnaeus embryonic stages have different levels of anoxia tolerance, it is also possible that this stability does not persist as anoxia tolerance decreases [41, 42]. Interestingly, stabilizing histone phosphorylation may be a component of apoptosis avoidance in A. limnaeus embryos, as induction of apoptosis in mammalian cells is associated with dephosphorylation of H1 histone isoforms and phosphorylation of H2B isoforms [115, 116]. Future work should investigate whether hPTM stability correlates with anoxic survival in later embryonic stages, particularly in regards to prevention of apoptosis [44].

Conclusions

This is the first study to explore the global landscape of hPTMs in an anoxia-tolerant vertebrate. We report a global stabilization of the hPTM landscape in anoxic compared with normoxic embryos, a result that is in stark contrast to studies on mammalian cells. We also see no changes in global levels of histone oxidation or lactylation, both of which were expected outcomes based on data from mammalian models. The mechanisms that maintain a stable global histone landscape in the face of a major stress and the role of these patterns in supporting anoxia tolerance are topics that merit further investigation.

Author contributions

Chelsea Hughes (Data curation [lead], Formal Analysis [lead], Software [lead], Validation [lead], Visualization [lead], Writing—original draft [lead], Writing—review & editing [supporting]), Elizabeth A. Mojica (Investigation [equal], Methodology [supporting], Software [supporting], Visualization [supporting], Writing—review & editing [supporting]), Dietmar Kültz (Data curation [supporting], Funding acquisition [equal], Methodology [lead], Project administration [equal], Resources [equal], Supervision [equal], Writing—review & editing [supporting]), Jason E. Podrabsky (Conceptualization [lead], Data curation [supporting], Formal Analysis [supporting], Funding acquisition [equal], Investigation [equal], Project administration [equal], Resources [equal], Supervision [equal], Visualization [supporting], Writing—review & editing [lead])

Conflict of interest

None declared.

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/JPLab01.url) and ProteomeXchange (PXD058622: https://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD058622, DOI: https://doi.org/10.6069/4bpv-yq94).