Comparative proteomic changes in rabbit vocal folds undergoing systemic dehydration and systemic rehydration

Abstract

Background:

A considerable body of clinical evidence suggests that systemic dehydration can negatively affect voice production, leading to the common recommendation to rehydrate. Evidence for the corrective benefits of rehydration, however, is limited with mixed conclusions, and biological data on the underlying tissue changes with rehydration is lacking. In this study, we used a rabbit model (n = 24) of acute (5 days) water restriction-induced systemic dehydration with subsequent rehydration (3 days) to explore the protein-level changes underlying the molecular transition from euhydration to dehydration and following rehydration using LC-MS/MS protein quantification in the vocal folds. We show that 5-day water restriction led to an average 4.3% decrease in body weight with relative increases in anion gap, Cl-, creatinine, Na+, and relative decreases in BUN, iCa2+, K+, and tCO2 compared to control (euhydrated) animals. A total of 309 differentially regulated (p < 0.05) proteins were identified between the Control and Dehydration groups. We observed a noteworthy similarity between the Dehydration and Rehydration groups, both well differentiated from the Control group, highlighting the distinct timelines of resolution of the clinical symptoms of systemic dehydration and the underlying molecular changes.

Significance:

Voice disorders are a ubiquitous problem with considerable economic and psychological impact. Maintenance of proper hydration is commonly prescribed as a general vocal hygiene practice. There is evidence that dehydration negatively impacts phonation, but our understanding of the state of vocal folds in the context of systemic dehydration are limited, particular from a molecular perspective. Further, ours is a novel molecular study of the short-term impact of rehydration on the tissue. Given the relatively minimal difference in vocal fold proteomic profiles between the Dehydration and Rehydration groups, our data demonstrate a complex physiological response to acute systemic dehydration, and highlight the importance of considering persistent underlying molecular pathology despite the rapid resolution of clinical measures. This study sets a foundation for future research to confirm the nature of potential beneficial outcomes of clinical recommendations related to hydration.

Introduction

Voice disorders are a common health concern, affecting millions of people annually, worldwide. Dysphonia negatively impacts quality of life [1] and can translate into substantial economic burden [2], especially among occupational voice users. Thus, minimally invasive therapeutic and prophylactic measures such as maintaining proper vocal hygiene are valuable. Proper hydration is commonly recommended to facilitate vocal hygiene. However, while considerable clinical evidence supports an etiological link between systemic dehydration and dysphonia, evidence for the utility of rehydration in an acute context is scarce. Further, human studies of phonation are limited to objective measures of voice (acoustic and aerodynamic measurements or visual assessment) or subjective measures of phonatory effort or discomfort. While insightful, these measures exhibit limited sensitivity and fail to describe the biological underpinnings for observed dysphonic changes associated with dehydration. The present study sought to address this gap in our understanding, focusing on the impact of systemic dehydration and subsequent rehydration on the vocal fold proteome.

Dehydration and rehydration as they relate to voice may manifest through distinct mechanisms: systemically as water from or into tissue via systemic circulation; or superficially from drying or hydration of the laryngeal lumen. Recent work has examined transcriptional and protein changes in systemically-dehydrated animals. Systemic dehydration was induced by water-restriction [3], furosemide-induced diuresis [4, 5], and complete water deprivation [6]. Pertinent changes across studies suggest potential inflammatory processes, disruptions to cellular junctional integrity, and changes to vocal fold extracellular matrix, all of which could contribute to dysphonia. To the best of our knowledge, there are no molecular investigations of the effects of rehydration following vocal fold dehydration. Interestingly, a recent study of lung function shows that acute systemic rehydration can improve dehydration-induced dysfunction, in contrast to nebulized rehydration [7], supporting the hypothesis that systemic rehydration may diminish dehydration-induced changes. The challenge in mechanistic research of systemic dehydration and systemic rehydration is determining the magnitude of realistic dehydration and rehydration levels as it pertains to human physiology. But exercise medicine provides some insights. Cellular hydration is integral to the maintenance of muscle function [8], and exercise-induced dehydration of as little as 2–3% loss in body weight has been linked to diminished endurance [9, 10] with observed differences in response between sexes [11]. Perceived exertion of exercise is also increased at body weight loss of at least 3% [12]. Dehydration has also been linked to impaired cognitive performance [13], but this is not a universal finding[14]. In general, a review of literature show that dehydration manifested as within 5% body weight loss is observed in the exercise literature. In this study, we achieved an approximately 4% reduction in body weight.

The present study used a New Zealand White rabbit model of water restriction-induced systemic dehydration followed by systemic rehydration to examine protein-level changes associated with each hydration state. Rabbits are a well-described surrogate model of vocal fold histological and molecular physiology [15, 16] and were chosen given that molecular studies in humans are ethically precluded or are restricted to cadaver or surgically resected tissue with concerns for pathological bias. Most importantly, the in vivo model includes the homeostatic mechanisms associated with dehydration stress tolerance which we hypothesized to be fundamental to the dehydration and rehydration responses.

Materials and methods

Rabbit care and tissue collection

Experiments were conducted in accordance with the guidelines and after approval of the Purdue Animal Care and Use Committee (Protocol # 1606001428) and following ARRIVE guidelines. New Zealand White male rabbits were obtained from Envigo Global (Indianapolis, IN). Twenty-four rabbits were included in the study. Rabbits were between 20.4–22.6 weeks of age on arrival, and measurements were taken between 22.4–26.6 weeks of age. Rabbits were allowed at least one week for acclimatization and were treated prophylactically for intestinal coccidiosis with a 5-day course of Amprolium in their drinking water.

Rabbits were randomly assigned to one of three experimental groups: Control (euhydrated), Dehydration, and Rehydration (n = 8/group). No rabbits were excluded from the analysis, but blood chemistry data (2^nd blood collection) is missing for one rabbit in the Rehydration group due to instrumental malfunction. Sample sizes were selected based on previous experiments from our group and related studies [4, 17, 18]. Food was provided ad libitum for the entire experiment. Rabbits were left in their cages except for body weight measurements and blood collection.

Blood was collected via venipuncture of the lateral ear vein with a 23 gauge needle. Packed cell volume (PCV) was measured by visual inspection following centrifugation of the hematocrit tubes. Blood chemistry was analyzed with the i-STAT Chem8+ cartridge that includes creatinine, blood urea nitrogen (BUN), glucose, sodium, chloride, potassium, total CO2, ionized calcium, anion gap, hematocrit (Hct), and hemoglobin using the i-STAT Alinity blood analyzer (Abaxis by Zoetis Inc., Parsippany-Troy Hills, NJ, USA). Euthanasia was completed with a single IV dose (1 mL; 100 mg/Kg) of Phenytoin/Pentobarbital (Beuthanasia-D Special, Schering Plough Animal Health Corp., Union, NJ, USA) through a catheter in the lateral ear vein.

Following euthanasia, the larynx was immediately removed, bisected along the dorsal sagittal midline, pinned to wax, and the bilateral full-thickness vocal folds removed by microscopy-assisted microdissection. Vocal fold tissue was collected into a cryovial, flash-frozen in liquid nitrogen, and stored at −80 °C until processing.

Water restriction protocol

Systemic dehydration was induced with a water restriction protocol. A full daily supply of water was provided at 500 mL. Water volume was measured with a 500 mL graduate cylinder to an accuracy limit of 2.5 mL. The Control group was provided full water supply for the entire experiment. The Dehydration and Rehydration groups were water restricted for five days at 50% of their average individual water intake as measured over baseline period (five days prior to starting the experiment). The Dehydration group was euthanized immediately following the dehydration period. The Rehydration group was provided full water supply for three days following the dehydration period. Control (euhydrated) and rehydrated rabbits were euthanized immediately following the rehydration period. Rabbit body weight was measured throughout the protocol.

Proteomic sample processing and data acquisition

Tissue sample preparation and mass spectrometry analysis were performed as previously described. Briefly, tissues samples were processed and protein extracted and quantified. Ultracentrifugation used to obtain separate soluble and insoluble fractions subject to separate LC-MS/MS analysis [17]. Modifications to the previous protocol are noted. Both fractions were digested with Pierce^™ Trypsin Protease, MS Grade (Thermo Fisher Scientific, Waltham, MA, USA). Peptides were separated with an Aurora UHPLC C18 packed emitter column (25 cm long × 75 μm ID) packed with 1.6 μm 120 Å (Ionopticks, Victoria, Australia). The column temperature was maintained at 50°C. A flow rate of 150 nL/min was used. As in the previous study, soluble and insoluble fractions were run separately during LC-MS acquisition, and data were merged when searching the database. These fractions are not interpreted by their cellular localization; rather, the Purdue Proteomics Core has observed that this improves overall protein identification. For a detailed description, see supplementary Method S1. All of the raw LC-MS/MS data are deposited in MassIVE data repository (ftp://massive.ucsd.edu/MSV000089151/) under ID MSV000089151.

Analysis of protein expression

The resulting complete set of proteins was filtered for half non-zero LFQ in at least one of the groups (at least n = 4 samples). LFQs were log2 transformed, median centered, and imputation performed sample-wise from a downshifted normal distribution. Group similarities were further assessed with principal component analysis and agglomerative hierarchical clustering with Spearman correlation distance and average linkage. Two differential protein expression comparisons were considered for further analysis and are discussed. The first comparison was between the Control and Dehydration groups alone (“DEHY”), and the second was between the Control and average of the Dehydration and Rehydration groups (“CDR”). The data were averaged for the dehydration and rehydration groups after observing that the proteomic profiles for these groups exhibited minimal difference. Data from two additional comparisons between the Control and Rehydration groups alone (“REHY”) and between the Dehydration and Rehydration groups alone (“DR”) are included illustratively but are not considered for biological interpretation given the confounding influence of the dehydration state on the rehydration state.

Enrichment and protein-protein interaction analysis

The UniProt Retrieve/ID mapping tool (https://www.uniprot.org/uploadlists/) was used to convert UniProtKB AC/ID from significant protein subsets to available “Gene name,” which were used as the input for gene enrichment analysis with Metascape [19] (https://www.metascape.org). Gene enrichment databases included GO Biological Processes [20, 21], KEGG Pathway [22, 23], Reactome Gene Sets [24], Transcription factor targets [25], and WikiPathways [26]. Enrichment parameters were set to a minimum overlap length of 10, p-value cut-off of 0.001, and minimum enrichment factor of 5; the GPEC option was selected. Protein-protein interaction enrichment databases included the Metascape Physical Core: STRING [27] (physical interactions), BioGrid [28] (physical interactions), OmniPath [29], and InWeb_IM [30]. Enrichment parameters were set with a network size of between 10 and 500. Results were filtered for minimum overlap length of 6, p-value cut-off of 0.001, and minimum enrichment factor of 5. Full details of the underlying analysis are available from Metascape. Briefly, an enrichment score and a hypergeometric distribution-based p-value are determined for each specific enrichment term based on represented genes. Enrichment terms from GO Biological Processes, KEGG Pathway, and Reactome Gene Sets are further clustered together based on shared member gene identity. These clusters often, but may not strictly, represent a specific functional annotation. The total number of gene sets was further reduced by collapsing them into the largest unique sets of genes to minimize redundancy. Protein-protein interaction in-network clusters are identified with the MCODE algorithm [31]. Associated enrichment terms are mapped onto these clusters independently of the gene-based enrichment. Data presented are modified from the original Metascape output.

Statistical analysis

All data analyses and visualization were completed using R (v4.1.2) with RStudio^™ version 1.4.1717 (RStudio, PBC, Boston, MA, http://www.rstudio.com) except when otherwise specified. Statistical consultancy was provided by colleagues specializing in high throughput data, as described in the acknowledgments. General linear models were used for all quantitative comparisons between groups with pertinent contrasts specified with the emmeans package (v1.7.2). Significance was defined at α = 0.05. Pairwise comparisons were corrected by Tukey’s adjustment. The Benjamini-Hochberg False Discovery Rate is provided where appropriate with multiple comparisons, although it was not used to filter statistically significant results. P-values and Q-values are reported on the log scale, where appropriate, for ease of interpretation.

Results

Water intake

The baseline daily water intake range within rabbits was between 30 mL -170 mL. In general, the baseline water intake across individual rabbits was markedly variable with mean of 291 mL (± 65 mL; Fig S1 A). There was no difference in mean water intake between groups during the baseline period (p > 0.24). The dehydration period allowed for an intake of only 50% of the average baseline intake for the Dehydration and Rehydration groups (Fig S1 B). The apparent increase in the mean water intake of the Rehydration group from the Control group during the first day of rehydration is significant (p < 0.0001), but the groups show no difference by the second day of rehydration (Fig S1 B).

Body weight

Body weight across all rabbits on the last day of the baseline period ranged from 2.61 Kg to 3.55 Kg with mean of 2.99 Kg and standard deviation of 0.26 Kg (Fig 1A). Given the rapidity with which young rabbits gain body weight and the range of rabbit ages, age in weeks at the time of measurement was included as a covariate. No significant difference between mean body weights between groups was seen at baseline (all p > 0.27). Following the first day of water restriction (Deh 1), the Dehydration and Rehydration groups exhibited a significant deviation in body weight from the Control group, which was maintained through the water restriction period (Fig 1B). The Dehydration and Rehydration groups exhibited no difference at any time point. There were no significant differences between the Control and Rehydration groups during the rehydration period.

Fig 1. Body weight of individual rabbits.

(A) at the last day of baseline period and (B) over the course of the experiment, standardized to the weight at baseline. Following the first day of water restriction (Deh 1), the Dehydration and Rehydration groups differed significantly from the Control group (body weight loss (BWL) = −3.1%, p = 0.0005 and −2.5%, p = 0.0035, respectively), which was maintained throughout the water restriction period (Deh 5: BWL = −4.6%, p < 0.0001 and −4.0%, p < 0.0001, respectively) while the Control group exhibited a 2.9% increase. The Dehydration and Rehydration groups exhibited no difference from each other at any time point (all p > 0.85). Differences between the Rehydration and Control groups are non-significant from the first day following rehydration (Reh 1, 2, and 3; p = 0.06, 0.08, 0.14, respectively).

Packed cell volume and blood chemistry

Packed Cell Volume (PCV) at baseline ranged from 37% to 48%. No significant effect of any group was seen at baseline (F(2, 21) = 0.88, p = 0.43) or following either the dehydration or rehydration periods (F(7, 56) = 0.9614, p = 0.47). BUN was the only analyte with a significant mutual difference between all three groups following dehydration (all p < 0.025) (Fig S2 A), while more analytes showed differences only within the Dehydration group. No additional analytes showed differences between the Control and Rehydration groups. Following this observation and considering the identical treatment state of both Dehydration and Rehydration groups at this time point (Deh 5), a comparison between the Control group and the combined mean of the Dehydration and Rehydration groups was performed, representing Control versus “water restricted” status. PCV remained non-significant (F(3,44) = 0.97, p = 0.41). In contrast, blood chemistry demonstrated significant changes that are shown in Table 1 and Fig S2 B–I. No significant differences were found between the Control and Rehydration groups during the rehydration period.

Table 1.

Summary statistics of significant findings of blood chemistry from the Control compared to pooled Dehydration and Rehydration groups.

	Group	Mean	Contrast Estimate	P-value
AnGap	Control	0.889	0.125	0.031
	Dehydration	1.053
	Rehydration	0.976
BUN	Control	1.027	−0.167	<0.0001
	Dehydration	0.800
	Rehydration	0.919
Cl	Control	1.014	0.032	0.014
	Dehydration	1.059
	Rehydration	1.034
Crea	Control	0.981	0.128	0.004
	Dehydration	1.087
	Rehydration	1.131
iCa	Control	1.041	−0.037	0.022
	Dehydration	1.003
	Rehydration	1.006
K	Control	0.961	−0.095	0.009
	Dehydration	0.828
	Rehydration	0.904
Na	Control	1.010	0.023	<0.0001
	Dehydration	1.038
	Rehydration	1.028
TCO2	Control	1.109	−0.141	0.005
	Dehydration	0.913
	Rehydration	1.023

Mean is the model estimated mean for the respective group. Contrast estimate is the difference between pooled mean of Dehydration and Rehydration and the mean of Control with the associated p-value.

Protein expression

Effects of dehydration

A total of 2990 unique FASTA identifiers were determined, leaving 1827 unique proteins when filtered for non-zero value LFQ criteria. Comparison between Control and Dehydration groups identified 309 significant differentially regulated proteins (p < 0.05), with 240 upregulated and 69 downregulated relative to the Control group (Fig 2A). Among these, 92 proteins exhibited at least a 2-fold change (78 up and 14 downregulated; mean difference of 1 on the log2 scale) and 9 exhibited at least a 4-fold change (5 up and 4 downregulated; mean difference of 2 on the log2 scale). Summary details for the top 15 most significant proteins by p-value and by group mean difference on the log2 scale are provided in Table 2, and a complete list is provided in Table S1. Principal component analysis indicates an apparent difference between the two groups with the entire protein set (Fig 2B). A more pronounced difference is seen with the significant subset (Fig 2C), with the first principal component explaining 43.4.5% of the overall variation. An intermediate similarity between samples D6 and D10 with C2, C3, and C10 is appreciated. Hierarchical clustering, including the full set of proteins, identifies a similar trend of differentiation between the groups (Fig 2 D).

Fig 2. Comparisons for DEHY subset.

(A) Volcano plot of protein expression comparisons between Control and Dehydration groups with mean difference (log2 scale) between groups on the x-axis and statistical significance (-log(p)) on the y-axis. There are 309 proteins with p < 0.05 (blue line). A p-value of 0.1 is indicated by the red horizontal line. There are 171 proteins with a log2-fold change of at least ±1.5, indicated by the gray vertical lines. (B) Principal component analysis using the full protein set (n=1827) and (C) the subset of 309 statistically significant proteins. Variance explained by the respective principal component is provided in parentheses. (D) Hierarchical clustering from the full protein set ignoring the Rehydration group.

Table 2.

Top 15 Differentially Regulated Proteins by P-value or Group Mean Difference in the DEHY Comparison.

	Uniprot ID	NAME	LogP	LogQ	D	C
P-value	A0A5F9D693	NEDD8-activating enzyme E1 catalytic subunit	−5.96	−2.70	1.53	0.70
	A0A5F9DCQ8	Clathrin heavy chain	−3.92	−0.98	0.84	0.79
	G1SM52	Leucine rich repeat containing 59	−3.60	−0.98	1.30	0.56
	G1SD44	Metaxin 1	−3.50	−0.98	1.69	0.64
	G1U754	Histidine-rich glycoprotein	−3.40	−0.98	0.67	0.79
	A0A5F9C6C7	Coatomer subunit beta	−3.36	−0.98	1.52	0.83
	G1TZA1	C1q domain-containing protein	−3.35	−0.98	1.19	0.74
	Q9N0Z6	Sodium/potassium-transporting ATPase subunit alpha-1	−3.34	−0.98	0.64	0.83
	G1SGL0	Sarcoglycan delta	−3.27	−0.96	0.78	0.77
	G1TDJ3	Extended synaptotagmin 1	−3.22	−0.96	1.69	0.69
	G1TM88	Serpin family A member 3	−3.12	−0.95	0.96	0.88
	P07293	Voltage-dependent L-type calcium channel subunit alpha-1S	−3.10	−0.95	0.58	0.71
	G1TIZ1	Ubiquitin carboxyl-terminal hydrolase	−3.09	−0.95	−0.60	−0.68
	A0A5F9DIY4	Myosin IC	−3.07	−0.95	0.70	0.90
	G1SES8	Mitochondrial ribosomal protein S22	−2.99	−0.92	−1.34	−0.41
Mean Difference	A0A5F9DT67	SERPIN domain-containing protein	−1.53	−0.62	4.01	0.43
	A0A5F9DDP4	Beta-microseminoprotein	−1.35	−0.55	2.58	0.43
	G1SEN8	Sacchrp_dh_NADP domain-containing protein	−2.24	−0.84	2.44	0.74
	G1U442	Mitochondrial pyruvate carrier	−1.38	−0.56	2.29	0.52
	A0A5F9C4W7	SERPIN domain-containing protein	−2.16	−0.79	2.18	0.66
	G1SZP0	Target of myb1 membrane trafficking protein	−2.96	−0.92	1.85	0.75
	G1SMI2	Acyl-CoA thioesterase 9	−2.30	−0.84	1.83	0.54
	G1T5A5	Reticulon 4 interacting protein 1	−2.45	−0.87	1.76	0.70
	G1U8F0	AP-2 complex subunit alpha	−2.56	−0.87	1.74	0.89
	G1T338	Peptidyl-prolyl cis-trans isomerase	−1.73	−0.68	1.70	0.52
	G1TDJ3	Extended synaptotagmin 1	−3.22	−0.96	1.69	0.69
	G1SD44	Metaxin 1	−3.50	−0.98	1.69	0.64
	G1SWS5	Phospholysine phosphohistidine inorganic pyrophosphate phosphatase	−1.95	−0.76	1.67	0.58
	G1TDC2	Solute carrier family 37 member 4	−2.08	−0.78	1.64	0.72
	G1SKE6	Gamma-sarcoglycan	−1.69	−0.68	1.64	0.56

LogP: log10(p) for the particular protein. LogQ: log10(FDR). D: mean difference on the log2 scale. C: Correlation to principal component 1.

Persistent effects of dehydration following rehydration

A total of 418 proteins were found with a significant difference between the Control group and the joint mean of Dehydration and Rehydration groups (CDR comparison), 331 upregulated and 87 downregulated relative to the Control group. Among these, 237 proteins exhibited at least a 2-fold change (207 up and 30 downregulated; contrast estimate of 2 on the log2 scale), and 102 exhibited at least a 4-fold change (84 up and 18 downregulated; contrast estimate of 4 on the log2 scale). Summary details for the top 15 most significant proteins by p-value and by contrast estimates on the log2 scale are provided in Table 3, and a full list is provided in Table S1. This contrast was coded contrast estimates are twice the magnitude of the true mean differences used to calculate the fold change. Principal component analysis provides evidence of differences between the Control and Rehydration groups with or without the inclusion of the Dehydration group (Fig 3A–B). The similarity among the Dehydration and Rehydration groups is apparent by principal component analysis of the entire protein set with no clear divergence from each other when restricted to either the significant CDR or REHY (Control vs. Rehydration group alone) subsets (Fig 3C–D). The Control and Rehydration groups cluster moderately well alone (Fig 4A) or when considering all three groups from the complete protein set (Fig 4B) and the significant CDR subset (Fig 4C).

Table 3.

Top 15 Differentially Regulated Proteins by P-value or Group Mean Difference in the CDR Comparison.

	Uniprot ID	NAME	LogP	LogQ	D	C
P-value	A0A5F9D693	NEDD8-activating enzyme E1 catalytic subunit	−6.51	−3.25	2.88	0.58
	A0A5F9DCQ8	Clathrin heavy chain	−4.64	−1.68	1.68	0.78
	Q9N0Z6	Sodium/potassium-transporting ATPase subunit alpha-1	−4.33	−1.55	1.36	0.82
	G1U754	Histidine-rich glycoprotein	−4.05	−1.54	1.34	0.76
	A0A5F9DPU2	Cullin 3	−4.01	−1.54	2.16	0.82
	G1T6E9	CDGSH iron sulfur domain 2	−3.89	−1.54	3.90	0.64
	G1SM52	Leucine rich repeat containing 59	−3.83	−1.54	2.36	0.59
	G1TIZ1	Ubiquitin carboxyl-terminal hydrolase	−3.80	−1.54	−1.23	−0.71
	G1U1V6	CSD domain-containing protein	−3.79	−1.54	−5.70	−0.58
	A0A5F9CSX3	Amine oxidase	−3.72	−1.54	2.56	0.77
	P07293	Voltage-dependent L-type calcium channel subunit alpha-1S	−3.69	−1.54	1.15	0.66
	G1T2Z8	S-methyl-5-thioadenosine phosphorylase	−3.68	−1.54	3.48	0.52
	G1TZA1	C1q domain-containing protein	−3.67	−1.54	2.21	0.69
	G1SK52	X-prolyl aminopeptidase 1	−3.65	−1.54	−0.79	−0.59
	G1SGL0	Sarcoglycan delta	−3.51	−1.43	1.43	0.73
Mean Difference	A0A5F9DT67	SERPIN domain-containing protein	−2.50	−1.21	9.92	0.50
	A0A5F9DDP4	Beta-microseminoprotein	−1.62	−0.83	5.07	0.41
	G1U442	Mitochondrial pyruvate carrier	−1.79	−0.90	4.76	0.52
	G1SEN8	Sacchrp_dh_NADP domain-containing protein	−2.59	−1.21	4.69	0.66
	A0A5F9DV00	Platelet activating factor acetylhydrolase 1b catalytic subunit 2	−2.76	−1.24	4.25	0.45
	G1T6E9	CDGSH iron sulfur domain 2	−3.89	−1.54	3.90	0.64
	A0A5F9C4W7	SERPIN domain-containing protein	−2.23	−1.10	3.85	0.64
	G1T7Q5	Carboxylic ester hydrolase	−1.42	−0.73	3.52	0.45
	G1T2Z8	S-methyl-5-thioadenosine phosphorylase	−3.68	−1.54	3.48	0.52
	G1T338	Peptidyl-prolyl cis-trans isomerase	−2.13	−1.05	3.41	0.54
	G1TDC2	Solute carrier family 37 member 4	−2.62	−1.21	3.37	0.71
	G1TBL1	Solute carrier family 25 member 20	−2.85	−1.30	3.30	0.68
	A0A5F9CVH3	Chloride intracellular channel protein	−1.94	−0.96	3.26	0.48
	G1SMI2	Acyl-CoA thioesterase 9	−2.36	−1.14	3.23	0.35
	G1SLF8	Ecm29 proteasome adaptor and scaffold	−3.05	−1.37	3.19	0.78

LogP: log10(p) for the particular protein. LogQ: log10(FDR). D: contrast estimate representing twice the mean difference on the log2 scale. C: Correlation to principal component 1.

Fig 3. Principal component analysis.

showing (A) the Control and Rehydration groups and (B) all three groups from the complete protein set (n= 1827). The same patterns are seen in (C), the significant CDR subset (n= 418), and (D), the significant REHY subset (n= 332). Variance explained by the respective principal component is provided in parentheses.

Fig 4. Hierarchical clustering.

(A) Clustering using full protein set (n=1824), ignoring dehydration group, (B) the full protein set (n=1827) considering all three groups, and (C) the significant CDR subset (n= 418) considering all three groups.

The Dehydration and Rehydration groups were further compared, ignoring the Control group. The Dehydration and Rehydration groups exhibit exceedingly few differentially regulated proteins from each other (n = 34, uncorrected p < 0.05; n = 15, Tukey’s adjusted p < 0.05; a complete list is provided in Table S1). Principal component analysis with the full protein set suggests a high level of similarity between the two groups, with only 30.4% of the overall variance explained (Fig 5A) with a marginal improvement to 34.8% when considering only the significant CDR and REHY subsets (Fig 5B–C). Among the significant proteins in the REHY subset, the majority of proteins exhibit very similar magnitudes of mean difference from the Control group and statistical significance in both Dehydration and Rehydration groups (Fig 5D).

Fig 5. Comparison of DEHY and REHY groups.

Principal component analysis of Dehydration and Rehydration groups from (A) the full protein set (n=1827), (B) the significant CDR subset (n= 418), and (C) the significant REHY subset (n= 332). Variance explained by the respective principal component is provided in parentheses. (D) Illustrative plot of differential expression between the Dehydration and Rehydration groups relative to Control. Dehydration-Control difference (log2 scale) is shown along the x-axis and Rehydration-Control difference (log2 scale) along the y-axis. Point color represents the magnitude of the ratio of -log10P for the respective Dehydration to Rehydration comparison, with red indicating higher significance. There are 34 proteins with p < 0.05 and 62 with p < 0.1. The dashed lines indicate a band corresponding to ±1.5 fold change expression, with the dotted representing equality.

Enrichment analysis and protein-protein interaction

A total of 116 enrichment terms were identified for the DEHY subset and 159 for the CDR subset. There is considerable overlap between the two sets. The top enrichment terms following gene set reduction, up to 3, from the first 5 Metascape-defined clusters from each comparison are provided in Table 4, and a complete list is provided in Table S2. Protein-protein interaction networks are shown in Fig 6, and a complete list is provided in Table S3. A total of 8 MCODE clusters were identified for the DEHY subset and 13 for the CDR subset; after filtering criteria similar to that used for the enrichment analysis were applied, 3 and 5 MCODE clusters were represented, respectively. The top enrichment terms, up to 3, for the filtered MCODE clusters are shown in Table 5. The protein interactions described within each cluster are shown in Fig 6. The Transcription Factor Targets analysis provided two results for the DEHY comparison (M40825, NR1H4; M14141, NRF2) and a single, shared result for the CDR comparison (M40825, NR1H4).

Table 4.

Gene Enrichment from Differentially Regulated Proteins.

Metascape Cluster	Enrichment ID	Description	LogP	LogQ	E	Prots
DEHY1	GO:1990542	mitochondrial transmembrane transport	−19.4	−15.3	103.9	11
	GO:0006839	mitochondrial transport	−6.2	−4.1	7.9	10
DEHY2	R-HSA-109582	Hemostasis	−18.3	−14.9	6.8	35
	hsa04610	Complement and coagulation cascades	−18.0	−14.6	111.1	10
	GO:0050878	regulation of body fluid levels	−14.8	−11.8	31.0	12
DEHY3	R-HSA-174824	Plasma lipoprotein assembly, remodeling, and clearance	−18.9	−15.3	134.9	10
DEHY4	R-HSA-9711123	Cellular response to chemical stress	−10.3	−7.6	10.9	14
	R-HSA-162906	HIV Infection	−7.2	−5.0	6.9	13
	R-HSA-195721	Signaling by WNT	−7.0	−4.9	5.6	15
DEHY5	R-HSA-5653656	Vesicle-mediated transport	−13.5	−10.6	5.6	30
	GO:0048193	Golgi vesicle transport	−6.3	−4.3	5.8	13
	R-HSA-446203	Asparagine N-linked glycosylation	−5.9	−3.9	5.3	13
CDR1	R-HSA-109582	Hemostasis	−23.8	−19.5	16.0	26
	hsa04610	Complement and coagulation cascades	−19.2	−15.8	58.5	13
CDR1	R-HSA-76002	Platelet activation, signaling and aggregation	−17.9	−14.8	9.5	27
CDR2	GO:0003013	circulatory system process	−22.0	−18.2	17.9	23
	GO:1903522	regulation of blood circulation	−18.8	−15.5	25.4	17
	GO:0090257	regulation of muscle system process	−15.9	−13.2	22.9	15
CDR3	R-HSA-6798695	Neutrophil degranulation	−19.7	−16.1	7.2	36
CDR4	R-HSA-9711123	Cellular response to chemical stress	−18.8	−15.5	35.9	15
	WP183	Proteasome degradation	−16.9	−14.0	65.7	11
	hsa05020	Prion disease	−15.3	−12.7	21.0	15
CDR5	hsa04961	Endocrine and other factor-regulated calcium reabsorption	−17.9	−14.8	79.4	11
	R-HSA-174824	Plasma lipoprotein assembly, remodeling, and clearance	−16.4	−13.6	60.1	11
	R-HSA-9679506	SARS-CoV Infections	−7.9	−6.1	7.9	13

LogP: log10(p) for the enrichment term. LogQ: log10(FDR). E: Enrichment factor of the enrichment term. Prots: The number of proteins represented by the enrichment term from the collapsed gene set.

Fig 6. PPI clusters identified by MCODE.

Nodes represent proteins, and edges represent the interaction between those proteins. Full clusters are shown beyond the number of the specific enrichment terms in Table 5. The 3 clusters on the left represent the DEHY subset MCODE clusters, and the five on the right, the CDR subset.

Table 5.

Protein-protein interaction enrichment terms.

MCODE Cluster	Enrichment ID	Description	LogP	LogQ	E	Prots
DEHY 1	R-HSA-1280218	Adaptive Immune System	−6.4	−4.6	14	7
DEHY 2	R-HSA-1280218	Adaptive Immune System	−11	−8.3	21	10
	R-HSA-9006934	Signaling by Receptor Tyrosine Kinases	−6.2	−4.5	18	6
	R-HSA-199991	Membrane Trafficking	−5.7	−4.1	15	6
DEHY 4	R-HSA-114608	Platelet degranulation	−25	−21	180	12
	R-HSA-76005	Response to elevated platelet cytosolic Ca2+	−25	−21	170	12
	R-HSA-8957275	Post-translational protein phosphorylation	−24	−20	190	11
CDR 1	GO:0043604	amide biosynthetic process	−5.7	−4.3	15	6
	R-HSA-382551	Transport of small molecules	−4.6	−3.5	9.9	6
	GO:0043603	cellular amide metabolic process	−4.6	−3.4	9.7	6
CDR 2	R-HSA-446203	Asparagine N-linked glycosylation	−6.9	−5.4	25	6
	R-HSA-199991	Membrane Trafficking	−5.1	−3.8	12	6
	R-HSA-5653656	Vesicle-mediated transport	−4.9	−3.7	11	6
CDR 3	R-HSA-8957275	Post-translational protein phosphorylation	−28	−23	190	13
	R-HSA-381426	Regulation of Insulin-like Growth Factor (IGF)*	−27	−23	170	13
	R-HSA-114608	Platelet degranulation	−27	−23	160	13
CDR 4	R-HSA-162909	Host Interactions of HIV factors	−15	−12	110	8
	R-HSA-4086400	PCP/CE pathway	−13	−11	140	7
	R-HSA-162906	HIV Infection	−13	−9.9	61	8
CDR 5	R-HSA-9711123	Cellular response to chemical stress	−9.6	−7.5	67	6
	hsa05022	Pathways of neurodegeneration - multiple diseases	−6.8	−5.3	22	6
	R-HSA-2262752	Cellular responses to stress	−5.6	−4.3	14	6

LogP: log10(p) for the enrichment term. LogQ: log10(FDR). E: Enrichment factor the enrichment term. Prots: The number of proteins represented by the enrichment term from the collapsed gene set.

Discussion

Water intake, restriction, and dehydration

Dehydration is a common physiological state that may be induced through many different mechanisms. As it relates to our current study, water restriction-induced dehydration is a physiologically relevant and realistic condition. Restricting volume to half the average intake conceptually mimics the circumstance of a human individual with full access to water who consumes a sustainable but suboptimal amount. We accounted for “optimal” water intake of the rabbits at the individual level by measuring baseline water intake of each rabbit. Interestingly, we observed a dramatic range of daily water intake among rabbits. While other restrictive means such as total water deprivation may result in a more rapid and pronounced state of dehydration, this is less applicable to the assumed lived experience of everyday voice users. The 4% and 4.6% magnitudes of bodyweight loss observed here are consistent with clinically evaluated mild-moderate pathological dehydration in children and adults as reviewed [32, 33] and realistic by intentional effort as reported in a group of male wrestlers [34]. Our choice of five days of water restriction and three days of rehydration is based on previous work in our lab in which rats returned to baseline body weight following moderate dehydration [35]. We conclude confidently that the changes we observed are due to dehydration induced by water restriction and avoid the confounds identified in other dehydration studies. While the definition of dehydration is straightforward in describing suboptimal water content, means by which dehydration is measured clinically are more nuanced and present additional challenges.

PCV, blood chemistry, and body weight

We make the fundamental assumption that rabbits are euhydrated at baseline; however, the ability to test this empirically is limited. Common clinical measures of dehydration include packed cell volume or hematocrit, blood analytes including BUN and creatinine, and physical measurements such as body weight and skin turgor. However, individual measurements may have limited diagnostic power, evidenced by the rabbits’ baseline PCV, which ranged from 37–48%, with a comparable spread at the other observed time points. Despite our repeated within-rabbit measures, PCV failed to indicate a shift from euhydrated to dehydrated state. Conversely, a number of blood analytes measured within the i-STAT Chem8+ panel were significantly different following water restriction. As expected, we observed increases in Na⁺, Cl⁻, and creatinine. Interestingly, BUN decreased following dehydration and was the only measure to exhibit significant mutual difference among all three groups following water restriction. The Rehydration group fell below the Control and above the Dehydration group. Nevertheless, creatinine is a more reliable marker of prerenal azotemia than BUN in rabbits. Besides hydration and renal function, BUN concentration can be influenced by nutrition status, circadian rhythm, intestinal absorption, and urease activity in the intestinal flora [36]. The Dehydration group may have diminished protein intake, influencing BUN levels. Body weight is our strongest indicator of water restriction-induced dehydration, with a clear decrease following restriction and a rapid increase following the rehydration protocol. Taken together, our 5-day water restriction and 3-day rehydration protocol induced clinically observable changes consistent with mild-moderate dehydration and resolution, respectively.

Group comparisons

Our study was motivated by the documented observation that systemic dehydration negatively impacts voice outcomes [37, 38]. Recent work from our group has begun to describe associated transcriptional changes in the vocal folds [4, 17, 39], but our novel approach here was a shift to focus on the functional level of the proteome. Our results were consistent with the hypothesis that systemic dehydration would induce changes in protein regulation. The primary focus was to identify the effects of rehydration following the systemic dehydration induced by water restriction. We hypothesized partial reversal of changes induced by water restriction back towards a Control state. The remarkable similarity between the Dehydration and Rehydration groups was unexpected and novel. The Rehydration status has no context without the animal having first been dehydrated, so while we consider this group distinct from a protein regulation perspective, we are cautious to differentiate Rehydration as its own condition. Thus, the decision was made to pool Dehydration and Rehydration groups in a second analysis following the observed proteomic similarities between the Dehydration and Rehydration groups. We interpret these results as changes from dehydration that persist despite the rehydration protocol. While we emphasize the proteomic similarity between these two groups, we recognize that these are disparate physiological hydration states.

At the level of individual proteins, both the Dehydration versus Control (DEHY) and pooled Dehydration-Rehydration versus Control (CDR) comparisons demonstrated upregulation of a substantial number of pertinent structural proteins, including collagens (1A2, 4A1, 4A2, and 6A3), fibrillin 2, fibulin 3, multiple serpins (A1, A3, F2, G1, and H1), and versican. These findings suggest alterations of structural maintenance of the vocal fold lamina propria with considerable biomechanical implications. Collagen is one of the fundamental structural components of the vocal folds [40] that, along with elastin and other extracellular matrix components like hyaluronic acid, contribute to the viscoelastic nature of the vocal folds that underlie phonatory capacity. The serpins (serine protease inhibitors) are a large family of protease inhibitors with pleiotropic effects. The three identified here could play a role in vocal pathology induced by dehydration: A1 is an inhibitor of a variety of proteases, including elastase, which may degrade elastin or collagen fibers [41] and stimulate fibroblasts [42], A3 may indirectly influence collagen stability by protection of the fibular accessory protein decorin [43], and H1 is a collagen-specific chaperone important for the successful maturation of collagen fibers [44]. Interestingly, serpin H1 has recently been shown as a potential therapeutic target against fibrosis in the vocal folds [45]. Versican and fibulin 3 influence collagen superstructure among many other extracellular matrix modifying activities [46–48], while fibrillin 2 is associated with elastic fibers. Although we have not tested it empirically, there is a considerable capacity for changes in these proteins to negatively impact voice by influencing the underlying vocal fold biomechanics.

The annotated enrichment through Metascape further suggests a diverse set of physiological changes associated with dehydration. Of particular interest are changes related to hemostasis. A direct link between dehydration and hemostasis is not well established, but it is reasonable to consider the physiological implications of dehydration on hemodynamics within the vocal fold vasculature. Virchow’s triad is a well-described relationship between blood flow, coagulability, and endothelial injury that provides contexts for how the enrichment of hemostatic factors may be explained by dehydration-mediated increase in blood viscosity eliciting a response from the vascular endothelium [49, 50]. Interestingly, dehydration has been suggested as a potential modifying factor for the risk of developing deep venous thromboembolism [51], and while thromboembolism is not a typical vocal fold pathology, diminished blood flow could directly impact the mechanics of voice production negatively.

Additional dehydration-responsive roles for endothelial cells are substantiated by a recent study demonstrating diversity among and transcriptional changes of renal endothelial cells in response to water deprivation-induced dehydration [52]. As dehydration is expected to increase both serum and interstitial osmolarity, it is reasonable to anticipate an endothelial contribution to a homeostatic response as well. Taken together, the vocal fold endothelium and implications to the submucosal interstitium are attractive targets for further analysis.

Oxidative changes secondary to dehydration are an anticipated stress and are evidenced by enrichments for mitochondrial function, cellular stress response, and targets of the transcription factor NRF2. NRF2 is a transcription factor that regulates response to oxidative and other stresses [53]. Oxidative stress has long been associated with dehydration. Our study used young adult male rabbits; previous studies showed that rapid bodyweight loss by water restriction increased markers of oxidative stress in young-adult male wrestlers [34], and chronic water restriction in a lizard species manifested oxidative changes that varied between sexes [54]. Evaluation of the in vivo effects of oxidative stress in the vocal folds is limited, but reactive oxygen species production is associated with early wound healing [55] and oxidative stress is associated with age-related changes [56]. More generally, chronic oxidative stress could damage the vocal fold microenvironment leading to dysphonia.

The most striking observation of this study was the overwhelming similarity of the proteomic profiles of the Dehydration and Rehydration groups. Dehydration is commonly identified in the clinical setting, and fluid restoration is a ubiquitous intervention often intended to rapidly restore a euhydrated state. We demonstrate herein that our clinical markers of dehydration fully resolve within 1 to 2 days following an oral rehydration protocol. However, the persistence of dehydration-induced changes has profound implications on how we should conceptualize the clinical intervention of systemic hydration within the vocal folds. Our results suggest that euhydration is necessary but not enough to present a baseline molecular state and that an unremarkable superficial clinical evaluation may be uncoupled from the underlying molecular pathology. We assume the observed changes from dehydration are transient, but we conclude that the 3-day rehydration period is insufficient time to restore the tissue biology to a pre-dehydrated state.

Limitations and future directions

Our proteomic study is limited to a between-group design because of the inherent attrition involved in taking vocal fold samples from the same animal repeatedly. The variation in baseline water intake among rabbits could be construed as a limitation but was minimized by defining water restriction relative to each animal. We chose proteomic analysis to focus on describing functional changes in the tissues following dehydration and rehydration. While the presence of differentially regulated proteins associated with enrichment terms implies upregulation of the associated genes at some point before sample collection, we recognize the limitation of proteomics to infer real-time transcriptional processes. We acknowledge that this study does not use a secondary validation of proteomic results. The potential for immunoblotting in this study is limited by the availability of validated and specific anti-rabbit probes. Given the exploratory nature of this study, we have relied instead on stringent filtering criteria of the LC-MS/MS and protein-level data to ensure the fidelity of our analysis. Future studies will benefit from resources dedicated to testing specific hypotheses informed by the present study. Future longitudinal studies that assess a more granular time interval approach and combine transcriptomic and proteomic analyses will also be informative. Lastly, we used full-thickness tissue samples in this study which confound the distinct contributions of individual tissue layers. Our data suggest that parallel interrogation of epithelial, mesenchymal, and immune cells would improve our molecular understanding and should be considered in future studies.

Supplementary Material

Supplemental Figure S1

Fig. S1. Water intake. (A) Baseline water intake over five days for individual rabbits. (B) Water intake for the duration of the experiment standardized to a 5-day baseline average. Base: baseline water intake period (5 days). Deh: dehydration period (5 days). Reh: rehydration period (3 days).

Supplemental Figure S2

Fig S2. Significant findings of blood chemistry (A) between Control, Dehydration and Rehydration groups and (B-I) from the Control compared to pooled Dehydration and Rehydration groups. Y-axis represents %-change from baseline measure. Dashed line lies along the x-axis: values above represent an increase, and values below represent a decrease from baseline measure. Green bars: (A-I) Control group. Blue bar: (A) Rehydration group. Red bars: (A) Dehydration group and (B-I) pooled Dehydration and Rehydration groups; “water restricted” status.

Supplemental Table 2

Table S2. List of enrichment terms from DEHY and CDR comparisons.

Supplemental Table 3

Table S3. List of PPI terms from DEHY and CDR comparisons.

Supplemental Table 1

Table S1. List of proteins from all group comparisons.

Supplementary Method 1

Method S1. Fully detailed list of LC-MS/MS proteomic data acquisition and processing.

List of Abbreviations

CDR

comparison between mean of Control and pooled means of Dehydration and Rehydration groups

DEHY

comparison between Control and Dehydration groups

DR

comparison between Dehydration and Rehydration groups

LC-MS/MS

liquid chromatography mass spectrometry/mass spectrometry

LFQ

label free quantification

PCV

packed cell volume

REHY

comparison between Control and Rehydration groups