Furosemide-induced systemic dehydration alters the proteome of rabbit vocal folds

Abstract

Whole-body dehydration (i.e., systemic dehydration) leads to vocal fold tissue dehydration. Furosemide, a common diuretic prescribed to treat hypertension and edema-associated conditions, induces systemic dehydration. Furosemide also causes voice changes in human speakers, making this method of systemic dehydration particularly interesting for vocal fold dehydration studies. Our objective was to obtain a comprehensive proteome of vocal folds following furosemide-induced systemic dehydration. New Zealand White rabbits were used as the animal model and randomly assigned to euhydrated (control) or furosemide-dehydrated groups. Systemic dehydration, induced by injectable furosemide, was verified by an average body weight loss of −5.5% and significant percentage changes in blood analytes in the dehydrated rabbits compared to controls. Vocal fold specimens, including mucosa and muscle, were processed for proteomic analysis using label-free quantitation LC-MS/MS. Over 1,600 proteins were successfully identified across all vocal fold samples; and associated with a variety of cellular components and ubiquitous cell functions. Protein levels were compared between groups showing 32 proteins differentially regulated (p ≤ 0.05) in the dehydrated vocal folds. These are mainly involved with mitochondrial translation and metabolism. The downregulation of proteins involved in mitochondrial metabolism in the vocal folds suggests a mechanism to prevent oxidative stress associated with systemic dehydration.

Graphical Abstract

Introduction

Voice disorders affect humans of different demographics globally [1–3]; in the United States, the prevalence in the general population is up to 7.6% [4] but voice disorders can affect 30% of the population over a lifetime [5]. The concept that a well-hydrated body (systemic hydration) contributes to healthy vocal folds and proper voice function is shared among voice specialists. It stems from studies demonstrating the detrimental effects of systemic dehydration on phonation and vocal fold biomechanics [6–8]. Systemic dehydration results from reduced fluid within the body due to decreased water intake or fluid loss (e.g., sweat, diarrhea) [9–11]. It can also be induced by dialysis, water withholding or restriction, and diuretics (e.g., furosemide) [6,10,12–14]. In vivo animal studies showed that systemic dehydration by water withholding leads to vocal fold dehydration [13,15]. Different modalities to induce systemic dehydration affect body physiology in distinct ways and cause molecular changes in vocal folds. Systemic dehydration by water withholding in rats affects the amount of hyaluronan in the vocal folds, a major component of the extracellular matrix involved in hydration homeostasis and vocal fold viscoelastic properties maintenance [16]. Water restriction in rats alters vocal fold gene and protein expression reducing pro-inflammatory cytokines, and interfering with plasma membrane integrity and hyaluronan network [14]. Furosemide-induced systemic dehydration altered the transcriptome of rabbit vocal folds by downregulating genes related to pro-inflammatory cytokines and epithelial cell barrier, suggesting a disruption in epithelial barrier integrity independent of inflammation [17]. Despite these significant molecular changes, the proteomic profile of systemically dehydrated vocal folds remains to be elucidated.

One critical consideration when studying the larynx is the choice of an animal model. Different animal species are used as models for laryngeal studies [18–21]. The rabbit is a desirable vocal fold tissue model given similarities to human vocal fold anatomy and biology [22,23]. Many studies have used rabbits to explore histological and molecular changes of vocal folds in response to vocal fold injury [24,25], phonotrauma [26–29], exposure to low humidity [30], and systemic dehydration by diuretic [17]. We have previously established a rabbit model of acute systemic dehydration with the loop diuretic furosemide, which revealed transcriptome changes compared to euhydrated tissue [17]. Systemic dehydration induced by furosemide also caused adverse changes in voice function in human subjects [6]. Hence furosemide was selected to induce systemic dehydration to inquire about vocal fold protein changes in the current investigation.

The advance and refinement of the technology for protein characterization and quantification over the years, together with user-friendly platforms for data analysis, make mass spectrometry (MS)-based proteomics an attractive alternative to traditional assays such as Western blot analysis and enzyme-linked immunosorbent assay (ELISA) [31,32]. Additionally, high-throughput proteomics permits large-scale protein characterization and quantification at a relatively low cost and high accuracy, sensitivity, specificity, and without the need for specific antibodies. The use of MS-based proteomics has been successfully applied in studies of laryngeal cancer and tissue engineering for vocal fold restoration [33–46]. However, the number of proteomic studies on the larynx is limited, with most of these studies focused on vocal fold fibroblasts, thyroarytenoid muscle, or laryngeal tumor tissue [33,35,38–41,45,47–52], and one study analyzing the proteomic profile of healthy vocal fold mucosa using a rat model [53]. In the study herein, we opted to use LC-MS/MS to obtain a complete proteome of rabbit vocal fold tissue.

We hypothesized that furosemide-induced systemic dehydration would lead to changes in vocal fold protein levels. The euhydrated group (control) was injected with isotonic saline. Systemic dehydration following furosemide administration was measured by multiple markers of body weight loss and blood analytes, including packed cell volume (PCV), total plasma proteins (TPP), creatinine, blood urea nitrogen (BUN), among others. Our criterion for systemic dehydration was ~5% body weight loss reflecting a moderate body dehydration level [9,54]. Once this level of dehydration was achieved, vocal fold tissues were collected, and the proteomic profiles were obtained using label-free LC-MS/MS. The proteome analysis revealed a subset of proteins differentially regulated in the systemically dehydrated vocal fold tissue, associated with downregulation of mitochondrial translation and metabolism. These findings suggest a mechanism to prevent oxidative damage associated with furosemide-induced systemic dehydration in the vocal folds.

Materials and methods

Rabbits care and systemic dehydration protocol

The study was approved by Purdue Animal Care and Use Committee under protocol number 1606001428, and it was conducted in compliance with the ARRIVE guidelines and following the National Institutes of Health Guide for the Care and Use of Laboratory Animals [55]. Eight 6-month-old New Zealand White male rabbits were acquired from Envigo (Envigo, Indianapolis, IN, USA) for this study. Male rabbits were chosen for this study to prevent interference of female sex hormones with body fluid-regulatory hormones reported in hormonally-cycling humans and rodents [56–58]. Rabbits were housed in an animal facility at Purdue University with controlled temperature (15–20°C) and humidity (~40%), and a 12h light/12h dark schedule. A socialization protocol involving toys and food enrichment, and human contact was part of the study to reduce animal stress and facilitate animal handling.

After a 7-day acclimation period, the rabbits were randomly divided into euhydrated (control, N= 4) and furosemide-dehydrated (N= 4) groups. Furosemide is a loop diuretic used to treat hypertension and edema associated with cardiac, renal, and hepatic failure [59]; it causes loss of water and ions by blocking the reabsorption of sodium and chloride in the ascending loop of Henle [60]. Prior to the furosemide-dehydration experiment, all rabbits received 0.1 M sucrose solution for 48 hours as a pre-hydration protocol to minimize variability in the baseline hydration levels. The average body weight of all rabbits was 3.3 Kg. Rabbits in the dehydrated group received one or two doses of 5.0 mg/Kg furosemide intraperitoneal (IP) injection (average of 0.3 mL/rabbit), and euhydrated rabbits received the same volume of saline IP injection. Although the volume of injected saline is too small to have any physiological effects, the IP injection procedure, which requires the animal to be held upside down, is a stress factor that needs to be ruled out as a confounding factor; therefore, all rabbits were handled the same way. The criterion to achieve systemic dehydration was body weight loss (BWL) of ~5% to reflect a moderate dehydration level, which can be observed after intense physical activity or due to illnesses [9,54]. Body weights of dehydrated rabbits were registered before furosemide injection and monitored every hour for approximately 6 hours, the duration of the experiment, following injection until they reached ~5% BWL. Body weights of euhydrated rabbits were recorded prior to saline injection and pre-euthanasia to assess body weight change. Rabbits in the dehydrated group did not have access to food or water during the experiment (~6 hours), while the euhydrated group had unrestricted access to food and water throughout the study. The experiment was conducted from 7 AM to ~1 PM. Laboratory rabbits consume most food during the night (after 5 PM) and have a low food intake during the morning (6 AM to 12 PM) [61]. Therefore, the food restriction during the morning time had a minimal impact in the furosemide-dehydrated rabbits. Fig. 1 shows the workflow of the study, including the systemic dehydration model, sample preparation for LC-MS/MS, and the proteomic analysis described below. The figure was created with BioRender.com.

Fig. 1.

Systemic dehydration experiment and label free quantitative proteomic analysis workflow. (A) Systemic dehydration protocol: rabbits in the dehydrated group (N= 4) were intraperitoneally (IP) injected with furosemide while rabbits in the euhydrated-control group (N= 4) were IP injected with saline. After ~6 hours, when dehydrated rabbits reached ~5% dehydration based on body weight loss, all animals were euthanized. (B) Sample preparation for LC-MS/MS analysis: larynges were collected and vocal fold tissue (VFT) was microdissected immediately following euthanasia and kept at −80 °C. Proteins were extracted as soluble and insoluble fractions and in-solution digestion with trypsin/Lys-C was performed for each sample (N= 8/group). Samples were analyzed by LC-MS/MS. (C) Proteomic data analysis: soluble and insoluble fractions data from each biological replicate (rabbit VFT) were combined for MaxQuant protein searches. Proteomic data were filtered and analyzed by Perseus software to identify differentially regulated proteins in the dehydrated group. Functional annotation, enrichment analysis and predicted protein-protein interaction network were performed using Metascape online software.

In addition, blood samples were collected from the ear vein into heparinized mini tubes at the beginning (pre-furosemide or saline injection) and the end (~6h post-injection; pre-euthanasia) of the experiment from all rabbits to evaluate packed cell volume (PCV) and total plasma proteins (TPP) as markers of dehydration. PCV and TPP were assessed using heparinized microhematocrit capillary tubes centrifuged at 15,000 × g for 2 minutes and manually verified using a microhematocrit reader card and a Reichert’s VET 360 refractometer (Ametek Reichert Technologies, Depew, NY, USA), respectively. Other blood analytes were evaluated using the iSTAT Chem8+ cartridge that includes creatinine, blood urea nitrogen (BUN), glucose, sodium, chloride, potassium, total CO₂, ionized calcium, anion gap, PCV (%), and hemoglobin (Abaxis by Zoetis Inc., Parsippany-Troy Hills, NJ, USA). These analytes were measured using the iSTAT Alinity blood analyzer (Abaxis by Zoetis Inc.). The percentage change in body weight and blood analytes was calculated for each rabbit as [% change = (final value – initial value)/initial value*100] and compared between dehydrated and euhydrated groups. The initial values were obtained pre-furosemide or saline injections, and final values were obtained ~6 hours after injections (pre-euthanasia).

Vocal fold samples collection and preparation for proteomic analysis

Larynges were collected following euthanasia, completed using 1.0 mL IV Beuthanasia-D Special (Schering Plough Animal Health Corp., Union, NJ, USA). Vocal fold tissue was microdissected from larynges under a dissection microscope as previously described [17] and immediately flash-frozen in liquid nitrogen and kept at −80 °C until protein extraction procedure. The vocal fold tissue was collected bilaterally (right and left sides) from each rabbit and consisted of soft tissue comprising mucosa and thyroarytenoid muscle identified at the level of the arytenoid cartilages at the transverse level of the glottis where the vocal folds reside in the larynx. Each bilateral section of the vocal fold had approximately 3 mm length × 2 mm depth.

Sample preparation for proteomic analysis was performed at the Purdue Proteomics Facility. Briefly, vocal fold samples (2 bilateral sections per rabbit) were thawed on ice and washed twice with 25 mM ammonium bicarbonate (ABC) before centrifugation at 6,500 rpm for 90 sec. Lysis was achieved by transferring the tissue pellets to 2 mL vials containing 350 μL of 100 mM ABC and 1.4 mm ceramic beads, followed by homogenization using a Precellys tissue homogenizer (Bertin Technologies, Rockville, MD, USA). Homogenates were transferred to new tubes and centrifuged at 13,500 rpm for 5 min at 4 °C. Supernatants were recovered, and the protein concentration was measured using the bicinchoninic acid assay (BCA) [62].

Protein fractionation was performed to separate each sample’s soluble and insoluble protein fractions; 50 μg of protein (equivalent volume) from each sample was used, and ultrapure water was added to achieve the same final volume across all samples. Next, samples were centrifuged at 55K rpm for 40 min in an Optima MAX-XP ultracentrifuge (Beckman Coulter, Indianapolis, IN, USA). The supernatant containing the soluble protein fraction was mixed with four volumes of cold acetone, mixed, and incubated at −20 °C overnight for protein precipitation. After overnight precipitation, soluble fractions were centrifuged at 13,500 rpm for 15 min at 4 °C to collect pellets, the supernatant was removed, and pellets were then dried in a vacuum centrifuge. Pellets from both soluble (after overnight precipitation) and insoluble fractions were treated with 10 μL of 10 mM dithiothreitol (DTT), 8 M urea in 25 mM ABC for 1 h at 37 °C for reduction, followed by incubation with 10 μL of alkylating reagent [2% iodoethanol, 0.5% triethylphosphine in acetonitrile (ACN)] for 1 h at 37 °C for alkylation.

After reduction and alkylation, samples were dried in a vacuum centrifuge. Digestion was achieved by adding mass spectrometry grade trypsin and Lys-C mix (Promega, Madison, WI, USA) to each sample at a 1:25 enzyme to substrate ratio and performed at high pressure using a NEP2320 barocycler (Pressure Biosciences, South Easton, MA, USA) set at 50 °C; 60 cycles of 50 seconds at 20 kpsi, and 10 seconds at atmospheric pressure. Digested peptides were desalted using MicroSpin C18 silica columns (The Nest Group, Inc., Southborough, MA, USA) and eluted with 80% ACN and 0.1% formic acid (FA). Purified peptides were dried in a vacuum centrifuge and stored at −80 °C until LC-MS/MS. Both soluble and insoluble peptides were resuspended in 3% ACN, 0.1% FA in water at a final concentration of 1 μg/μL, and 1 μL was loaded to the HPLC system. The objective of the soluble and insoluble fractionation was to enhance proteome coverage and not to define the subcellular localization of proteins. Each fraction was run independently during mass spectrometry acquisition and then merged for each correspondent sample before database searches.

LC-MS/MS for peptide sequencing and data analysis

Proteomic analysis was performed at the Purdue Proteomics Facility. Peptides were analyzed in a Dionex UltiMate™ 3000 RSLCnano System (ThermoFisher Scientific, Waltham, MA) coupled online to an ETD-enabled Orbitrap Fusion Lumos Mass Spectrometer (ThermoFisher Scientific) as previously described [63]. Briefly, reverse phase peptide separation was accomplished using a trap column (300 μm ID × 5 mm) packed with 5 μm 100 Å PepMap C18 medium coupled to a 50-cm long × 75 μm inner diameter analytical column packed with 2 μm 100 Å PepMap C18 silica (ThermoFisher Scientific). The column temperature was maintained at 50 °C. Samples were loaded to the trap column in a loading buffer (2% ACN, 0.1% FA) at a flow rate of 5 μL/min for 5 min and eluted from the analytical column at a flow rate of 200 nL/min using a 120-min LC gradient as follows: linear gradient of 6.5 to 27% of solvent B in 82 min, 27–40% of B in next 8 min, 40–100% of B in 7 min at which point the gradient was held at 100% of B for 7 min before reverting back to 2% of B, and hold at 2% of B for next 15 min for column equilibration. The column was further washed and equilibrated by using three 30-min LC gradients before injecting the following sample. All data were acquired in the Orbitrap Fusion Lumos Mass Spectrometer (ThermoFisher Scientific). Orbitrap mass analyzer and data were collected using a higher-energy collisional dissociation (HCD) fragmentation scheme. For MS scans, the scan range was from 350 to 1500 m/z at a resolution of 120,000, the automatic gain control (AGC) target was set at 4 × 10⁵, maximum injection time 50 ms, dynamic exclusion 60s, and intensity threshold 5.0 ×10³. MS data were acquired in data-dependent mode with a cycle time of 5s/scan. MS/MS data were collected at a resolution of 7500. The mass spectrometry data underlying this study were deposited in the Center for Computational Mass Spectrometry database under MassIVE ID: MSV000086612 (https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp).

LC-MS/MS data were analyzed using MaxQuant software (version 1.6.3.3) [64] against rabbit (Oryctolagus cuniculus) database downloaded from the UniProt (www.uniprot.org) containing 44,508 protein sequences. The following parameters were edited for the searches: precursor mass tolerance of 10 ppm; enzyme specificity of trypsin/Lys-C enzyme allowing up to 2 missed cleavages; oxidation of methionine (M) as a variable modification and iodoethanol (C) as a fixed modification. False discovery rate (FDR) of peptide spectral match (PSM) and protein identification was set to 0.01. The unique plus razor peptides (non-redundant, non-unique peptides assigned to the protein group with most other peptides) were used for peptide quantitation. LFQ (label-free quantitation) intensity values were used for relative protein abundance measurement. Only proteins detected with at least one unique peptide and MS/MS ≥ 2 (spectral counts) were considered as valid identification and used for statistical analysis using Perseus software [65].

Principal component analysis

Principal component analysis (PCA) was performed for the list of proteins identified across all eight samples. The list of proteins was organized by ascending p-value and PCA was performed on subsets of proteins to determine the larger subset of proteins that provided discrimination between control (euhydrated) and dehydrated group.

Proteomics functional enrichment analysis and data visualization

Functional pathway and process enrichment analysis was performed with Metascape online software [66] for two datasets: 1) all proteins identified in both control and furosemide-dehydrated rabbit vocal folds (N=8) regardless of expression level, and 2) differentially regulated proteins identified in the furosemide-dehydrated vocal folds (N=4) compared to control (N=4). For dataset #1 comprising all vocal fold proteins, gene names were retrieved using the entire list of proteins identified across all eight rabbits’ samples. The available gene identifiers were uploaded to Metascape and converted to Entrez Gene IDs of Homo sapiens as the species available in the database with the highest number of orthologues for rabbits’ genes. The unique IDs were then annotated and used in the functional enrichment analysis applying the software default significant criteria settings and filtering for the following ontology categories: GO Molecular Functions, GO Biological Processes, GO Cellular Components, and KEGG Pathway. Subsequently, a network of the top 20 most enriched clusters organized by Metascape based on membership similarities established by the software, with each cluster represented by the most statistically significant term within, was created to provide a global picture of the main functions represented in the proteome of rabbit’s vocal folds. A Venn diagram was generated with the complete set of proteins identified in both the control and furosemide-dehydrated groups to illustrate the overlap between the two groups. A heatmap based on Z-transformed LFQ values of each protein identified across all eight biological replicates was created with Morpheus software (https://software.broadinstitute.org/morpheus).

For dataset #2 of the differentially regulated proteins in the furosemide-dehydrated vocal folds, gene names were retrieved and annotated as described above, and the enrichment analysis was filtered for GO Biological Processes. A detailed description of the enrichment analysis and functional clusters network is available on the Metascape website: https://metascape.org/gp/index.html#/menu/manual#Enrichment. In addition, for dataset #2, a protein-protein interaction (PPI) network based on predicted physical connections, applying the Molecular Complex Detection (MCODE) algorithm that identifies densely connected network neighborhoods was also performed. A heatmap representation of the furosemide-differentially regulated proteins (p < 0.05) based on Z-scored LFQ values of each protein across the biological replicates (N=4/group) of each treatment group was created with the Morpheus software.

Statistical analyses

The percentage change in body weight (% body weight loss) and blood analytes was compared between groups by applying the Mann-Whitney test using GraphPad Prism (version 6.0e for Mac). Differences between groups were considered statistically significant for a p-value ≤ 0.05.

The statistical analysis to identify proteins with different expression levels between dehydrated and control-euhydrated groups was performed using Perseus software platform (v1.6.12.0) [65] similarly as described before [67]. MaxQuant data was uploaded to Perseus and analyzed following the described steps: “contaminants”, “reverse”, and “only identified by site” proteins were removed; the LFQ values were Log2 transformed; the proteins were filtered to have a minimum of two valid LFQ values representing two replicates (50% of samples) in at least one group (dehydrated or euhydrated); missing values were imputed based on normal distribution; and two-sample two-tailed Student’s t-test was performed to compare the means between groups. Proteins with a p-value ≤ 0.05 and absolute Log2(LFQ) ≥|0.38| were considered significantly regulated.

Results

Rabbit furosemide-systemic dehydration model

We assessed systemic dehydration by evaluating the % BWL and the % change in the level of hematological analytes in the rabbits. The average BWL of dehydrated rabbits was −5.5% (range= 4.5 to 6.7%), after approximately 6 hours following furosemide injection, compared to −1.2% in the euhydrated group (p= 0.0143; Fig. 2A). The percentage changes in PCV, hemoglobin and TPP were significantly greater, as were the blood creatinine and BUN levels in the dehydrated rabbits (p < 0.05) (Fig. 2B and C), confirming that systemic dehydration was achieved. Finally, we observed a significant decrease in the percentage change of potassium and ionized calcium blood levels and a significant increase in the blood glucose level in the dehydrated rabbits (Fig. 2D); findings related to the dehydration method with furosemide. Table 1 summarizes the statistical analysis results for the % changes of all blood analytes compared between the euhydrated and dehydrated groups.

Fig. 2.

Systemic dehydration markers. (A) Percentage body weight loss and (B-D) % change in blood levels of tested analytes between euhydrated and dehydrated rabbits. PCV: packed cell volume; TPP: total plasma proteins; BUN: blood urea nitrogen. Bars represent the mean and standard error of the mean (± SEM) of four rabbits in a respective group. *significant statistical difference between group means with p-value ≤ 0.05.

Table 1.

Summary of statistical analysis of percentage change in blood levels of analytes between control-euhydrated and furosemide-systemically dehydrated rabbits.

	Control-euhydrated (N=4)	Dehydrated (N=4)
Blood analyte	(mean ± SEM)		p-value*
PCV	2.4 ± 0.97	11.96 ± 1.29	0.0143
TPP	−3.63 ± 2.21	15.33 ± 2.34	0.0143
iSTAT Chem8+
Creatinine	10.27 ± 6.77	61.66 ± 3.58	0.0143
BUN	21.07 ± 4.32	62.71 ± 8.44	0.0143
Glucose	−3.65 ± 8.48	67.12 ± 34.92	0.0286
Sodium	0.002 ± 0.41	0.73 ± 0.6	0.2143; ns
Chloride	−1.10 ± 2.87	−3.67 ± 0.47	0.1571; ns
Potassium	−5.60 ± 3.09	−22.42 ± 4.06	0.0286
Total CO2	48.87 ± 28.19	5.81 ± 13.95	0.6571; ns
iCa	−2.33 ± 0.47	−6.02 ± 1.36	0.0286
Anion gap	−20.42 ± 6.92	16.59 ± 14.89	0.0857; ns
Hematocrit	−3.21 ± 1.24	14.91 ± 2.09	0.0143
Hemoglobin	−3.22 ± 1.19	14.67 ± 2.13	0.0143

PCV: packed cell volume; TPP: total plasma protein; BUN: blood urea nitrogen; iCa: ionized calcium.

^*Mann-Whitney exact p-value: statistically significant differences between euhydrated and dehydrated group when p ≤ 0.05. ns: non-significant.

Proteomic profile of rabbit vocal folds

The MaxQuant software identified 2,414 proteins, combining soluble and insoluble protein fractions, across the eight vocal fold tissue samples against the rabbit database. These data were uploaded to the Perseus platform and filtered as described in the Method section. After filtering the data, a list comprising 1,684 proteins (Table S1) was analyzed for differential regulation, i.e., decreased and increased levels, in the dehydrated vocal folds compared to the euhydrated tissue. Before exploring the differentially regulated proteins, we analyzed the complete set of proteins identified across all eight vocal fold samples to obtain a global picture of the main functions and pathways represented in this tissue. A total of 1,496 proteins had correspondent gene names and were uploaded to Metascape for functional enrichment analysis; the remaining 189 Uniprot IDs did not retrieve gene names and were not included in this analysis. The input gene identifiers were converted into 1,314 unique Entrez Gene IDs of Homo sapiens, the target species selected in the Metascape database, since rabbit is not available, providing the largest number of orthologues. The enrichment analysis resulted in thousands of gene ontology (GO) enriched terms representing various biological processes, subcellular localization, and pathways. Metascape grouped significantly enriched terms into clusters based on function similarities, and a network of these enriched functions was created. The top 20 most enriched clusters and the functional enrichment network are shown in Figs. 3A and 3B, respectively, and the complete list of enriched terms in each cluster is provided in Table S2. These enriched clusters suggest pathways associated with vocal fold mucosa (epithelial layer and lamina propria: e.g., cell-substrate junction, cell adhesion molecule binding, collagen-containing extracellular matrix) and muscle (e.g., contractile fiber, actin binding, muscle system process). The most enriched terms include biological functions related to mitochondrial metabolism and structure and represent eight clusters forming the largest network, followed by muscle function and structure represented by four clusters, and extracellular matrix components, also represented by four clusters (Fig. 3B).

Fig. 3.

Pathway and Process Enrichment Analysis of rabbit vocal folds proteomic profile. (A) Top 20 most enriched ontology terms (clusters) identified for the complete vocal folds proteome. These enriched clusters suggest a successful representation of proteins found across the entire vocal fold tissue: mucosa (epithelial layer and lamina propria: e.g., cell-substrate junction, cell adhesion molecule binding, collagen-containing extracellular matrix), and muscle (e.g., contractile fiber, actin binding, muscle system process). (B) Network of enriched terms colored by cluster IDs (shown in 3A). Each cluster ID represents the most enriched term within that cluster. Nodes of the same color belong to the same cluster and represent up to 10 enriched terms (selected based on the lowest p-values) within a given cluster. The size of the nodes is proportional to the number of input genes (correspondent to proteins) that fall into that term. The largest network contains eight clusters related to mitochondrial structure and functions denoting a significant representation of mitochondrial proteins in the rabbit vocal folds. The next largest network, with four clusters and 32 enriched terms, represent functions predominantly related to muscle. Complete enrichment analysis is provided in Table S2. Figures were created with Metascape.

Most of the proteins (1,541) were identified in both control and furosemide-dehydrated groups (Fig. 4A). From the proteins identified exclusively in the control (68 proteins) or furosemide-dehydrated (75 proteins) group, 12 showed significant differences (p ≤ 0.05) after the imputation step for statistical analysis (Fig 4A; Table S3). These proteins are shown in Table 2 and discussed according to their relevance to this study. The heatmap of the 1,684 proteins identified in rabbit vocal folds shows the distribution of protein expression patterns across individual samples (Fig. 4B). The variability between replicates of the same group is expected when using animals as biological replicates as opposed to genetically and physiologically homogenous samples such as cell culture.

Fig. 4.

Rabbit vocal fold proteins distribution and expression pattern across biological replicates. (A) Venn diagram of all proteins identified across the eight vocal fold samples. Majority of proteins were identified in both control and furosemide-dehydrated groups. (B) Heatmap representation of all proteins identified across eight biological replicates. Color gradient represent the Z-scored LFQ values of each protein across the samples of each group; row colors are normalized based on each row’s minimum and maximum value. Blue and red colors represent the minimum and maximum expression pattern, respectively.

Table 2.

Proteins differentially regulated in vocal folds in response to acute exposure to furosemide-induced systemic dehydration.

Regulation	Molecular Pathway§	Protein IDs*	Gene symbol*	Protein Name	p-value**	Log2 FC
Downregulated	mitochondrial translation	G1U3G6	PTRH2c	Peptidyl-tRNA hydrolase 2	0.0001	−2.702
		G1SI29	TUFM	Elongation factor Tu, mitochondrial	0.0103	−0.104
		G1T752	MCTS1; LOC100352615	Malignant T-cell-amplified sequence	0.0247	−0.423
		A0A5F9CFH0	FARSB	Phenylalanyl-tRNA synthetase beta subunit	0.0289	−0.203
		G1TB33	MRPL13	Mitochondrial ribosomal protein L13	0.0308	−1.326
		G1SE37	MRPL58; ICT1	Mitochondrial ribosomal protein L58	0.0345	−2.035
		G1SZN2	MRPL38	Mitochondrial ribosomal protein L38	0.0465	−0.386
	oxidative stress	G1SND1	FMO2	Flavin Containing Dimethylaniline Monooxygenase 2	0.0035	−0.591
		G1TEA3	CAVIN2; SDPR	Caveolae associated protein 2	0.0282	−0.457
		G1SLY5	HSPB8c	Heat shock protein beta-8	0.0404	−0.700
	endocytosis	A0A5F9DCU4	DTNAc	Dystrobrevin alpha	0.0039	−1.826
		G1TKS5	OCIAD2	OCIA domain containing 2	0.0309	−0.601
		B7NZM8	YWHAH	Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein eta	0.0455	−0.482
	dephosphorylation	G1SS49	HDHD2c	Haloacid dehalogenase like hydrolase domain containing 2	0.0031	−3.248
	proteasomal protein catabolic process	A0A5F9C5T7	SKP1; LOC100348953	S-phase kinase-associated protein 1	0.0057	−0.776
	immune response	B5L650	RLA-A1c	MHC class I antigen	0.0134	−2.094
	RNA processing	G1SVQ4	PIN4; LOC100351985	Peptidyl-prolyl cis-trans isomerase	0.0482	−0.472
	synaptic vesicle transport	G1SKY8	TMEM230c	Transmembrane protein 230	0.0491	−1.219
Upregulated	stress response	A0A5F9CEF1	PCCA	Propionyl-CoA carboxylase subunit alpha	0.0113	0.149
		A0A0A0MQQ2	S100A12; LOC100008703	Protein S100	0.0399	4.273
	macroautophagy	G1T2T9	RAB5Af	RAB5A, member RAS oncogene family	0.0030	1.341
		G1U3Q0	ATP6V0A1	V-type proton ATPase subunit a	0.0204	0.404
	muscle tissue development	G1U3C5	IGFBP5	Insulin like growth factor binding protein 5	0.0361	0.426
		A0A5F9CCB2	KLHL40	Kelch like family member 40	0.0381	0.336
		A0A5F9DDG7	ALDH1A2f	Aldehyde dehydrogenase 1 family member A2	0.0417	1.103
		G1TKN4	RAC1	Rac family small GTPase 1	0.0434	0.253
	protein folding and sorting	A0A5F9DCR4	CALUf	Calumenin	0.0010	3.223
	N/A	A0A5F9DEL1f	N/A	Ig-like domain-containing protein	0.0115	3.368
	translation	G1SUA2	TMA7	Translation machinery associated 7 homolog	0.0226	0.308
	response to salt stress	G1TR45	EFHD2	EF-hand domain family member D2	0.0321	1.257
	acute inflammatory response	G1T127	KLKB1f	Kallikrein B1	0.0366	1.661
	energy metabolism	A0A5F9CQS8	D2HGDHf	D-2-hydroxyglutarate dehydrogenase	0.0495	1.244

^§Molecular pathways based on the most enriched GO biological process (GO-BP) or individual GO-BP annotation (in the absence of enriched term) for each protein, provided by Metascape analysis. N/A: no BP annotation provided.

^*Protein IDs and gene symbols were retrieved from Uniprot database (https://www.uniprot.org/)

^cProteins identified in the control-euhydrated group only.

^fProteins identified in the furosemide-dehydrated group only.

^**P-values from Student’s t-test. The bolded Log2 Fold Change (≥ |0.38|) proteins are considered of significant magnitude.

Principal component analysis (PCA) was performed with the complete proteome dataset (1,684 proteins) and for smaller subsets of proteins arranged by ascending p-value to determine the largest number of proteins that provided discrimination between euhydrated control and furosemide-dehydrated groups. As expected, the full dataset did not provide clear separation as the majority of proteins had similar expression between groups (Fig. S1). PCA with the top 500 proteins provided clear linear separation between groups with 36.2% and 15.7% of the overall variance explained by PC1 and PC2, respectively (Fig. 5A). PCA with top 250, 100 and the 32 proteins differentially regulated in the furosemide-dehydrated group (described below) discriminated the groups with 57.1%, 63%, and 74.7% overall variance between PC1 and PC2, respectively (Fig. 5B–D).

Fig. 5.

Principal component analysis. (A) PCA for the top 500 proteins, (B) top 250 proteins, (C) top 100 proteins, (D) and 32 proteins differentially regulated in the furosemide-dehydrated group arranged by ascending p-value show clear separation between control and furosemide-dehydrated groups. Complete list of proteins is provided in Table S1.

Proteomic profile of rabbit vocal folds systemically dehydrated with furosemide

Thirty-two proteins were identified with differential regulation in the vocal folds of furosemide-systemically dehydrated rabbits (p ≤ 0.05). Eighteen proteins were found downregulated, while fourteen were upregulated (Table 2). Among the downregulated group, proteins involved with mitochondrial translation such as Mitochondrial Ribosomal Proteins (MRPL13, MRPL38, MRPL58), Tu Translation Elongation Factor, Mitochondrial (TUFM), Peptidyl-tRNA hydrolase 2 (PTRH2), and Phenylalanyl-tRNA synthetase beta subunit (FARSB) were identified. Proteins associated with oxidative stress, including Flavin Containing Dimethylaniline Monooxygenase 2 (FMO2), Cavin-2 (SDPR), and the mitochondrial small heat shock protein HSPB8 were also downregulated. While in the upregulated group, we identified Propionyl-CoA carboxylase subunit alpha (PCCA) and S100 Calcium Binding Protein A12 (S100A12), both associated with a stress response. Heatmap representation of these 32 proteins across biological replicates (rabbits) in control and dehydrated groups is shown in Fig. 6A. Enrichment analysis resulted in four main biological processes, including translation, with the gene list associated with mitochondrial translation and muscle development (Fig. 6B; Table S4). Protein-protein interaction was predicted for 19 proteins, with MCODE components associated with translation (Fig. 6C).

Fig. 6.

Proteins differentially regulated in furosemide-dehydrated rabbit vocal folds. (A) Heatmap representation of 32 proteins with p ≤ 0.05. Color gradient and dot size represent the Z-scored LFQ values of each protein across the biological replicates of each treatment group. Blue color and smaller dot sizes represent proteins with Z-score < 0, and red color and larger dot sizes represent proteins with Z-score > 0. Analysis show that majority of these proteins are downregulated. (B) Enriched Gene Ontology (GO) biological processes for proteins differentially regulated (p ≤ 0.05) with the most enriched pathway related to mitochondrial translation, followed by macroautophagy, muscle tissue development and endocytosis. (C) Protein-protein interaction (PPI) network of proteins differentially regulated (p ≤ 0.05). PPI show predicted interaction between mitochondrial proteins suggesting and orchestrated response to furosemide-dehydration in the rabbit vocal folds. Connection lines represent putative protein-protein interaction, using the “Physical all” database in Metascape. Proteins from an MCODE network are highlighted in red. Enrichment analysis results are provided in Table S4.

Discussion

This study aimed to investigate the changes in protein expression levels in furosemide-induced, systemically dehydrated vocal folds. Systemic dehydration was successfully verified by different dehydration markers, including % body weight loss and increased % change of blood analytes such as PCV, TPP, creatinine, and BUN [68]. Clinically, PCV and TPP increase when total blood volume decreases, most commonly due to dehydration [68]. A parallel increase in creatinine and BUN concentrations is collectively designated as azotemia and reflects decreased renal glomerular filtration rate (GFR). The transient decreased GFR results from reduced blood volume in dehydration, classified as pre-renal azotemia [69]. All these significant changes observed in the dehydrated rabbits were expected due to the furosemide diuretic effect. Also, as a pharmacological consequence of the furosemide mechanism of action, the blood levels of potassium and ionized calcium were significantly decreased while the glucose level was significantly increased in the dehydrated rabbits compared to euhydrated controls [70,71].

The rabbit vocal folds reveal a diverse array of proteins

The proteome of vocal fold tissue from euhydrated and furosemide-dehydrated rabbits comprises 1,684 proteins representing the mucosal and thyroarytenoid muscle layers. Our study identified a substantially higher number of proteins compared to the proteome of rat vocal fold mucosa (1,684 versus 179) [53]. However, this difference can be explained by the presence of the muscle layer within the vocal fold specimens in our study, while the rat study targeted only the mucosa. Besides, our protein extraction method applied in-solution protein digestion with fractionation of soluble and insoluble proteins, increasing the number of proteins recovered instead of in-gel digestion that limits the recovery of proteins. Extracellular matrix (ECM) rich-tissues like vocal folds are challenging for proteomic analysis due to the digestion-resistant nature of components [72]. In our study, the identification of a group of proteins related to ECM such as fibronectin, collagens, decorin, fibrillin, fibulins, laminins, elastin, proteoglycan, among others, indicates that the fractionation method used for protein extraction was successful in solubilizing and recovering a number of these proteins (highlighted proteins in Table S1). The functional enrichment analysis of the proteome of vocal folds from both euhydrated and furosemide-dehydrated rabbits reveals a diverse group of cellular components and biological processes that work together to maintain the tissue structure and function. These data revealed several newly described proteins in this tissue, which can be explored in future studies related to vocal folds and associated vocal pathologies.

Systemic dehydration and oxidative stress

Systemic dehydration at a level as low as 3% BWL is associated with fluid loss due to physical exercise in humans and can lead to increased production of reactive oxygen species (ROS) in the mitochondria [73]. Additionally, a moderate level of systemic dehydration, caused by either water deprivation or heat stress, induced oxidative stress in the brain of laboratory and wild mice, respectively [74,75]. In our study, the proteins differentially regulated in the vocal folds of furosemide-induced systemically dehydrated rabbits suggest a significant involvement of the mitochondria, likely in response to increased intracellular reactive oxygen species (ROS). ROS are generated continuously in all aerobic cells as byproducts of metabolism [76] and cell signaling [77]. Cells have developed enzymatic and non-enzymatic antioxidant defense mechanisms to prevent ROS formation [76,78]; however, when ROS generation exceeds the antioxidant capacity of the cell, oxidative stress occurs. Oxidative stress is a deleterious process that can result in the oxidation of biological molecules such as nucleic acids, proteins, carbohydrates, and lipids, especially in aerobic environments [76–78]. Dehydration can potentially cause damage (denaturation and conformational changes) to the cell membranes and other proteins [79,80]. There is evidence that dehydration is linked to increased cell susceptibility to ROS. However, the origin of ROS formation in dehydration is unknown [81]. The capacity of ROS formation by vocal fold tissue has been demonstrated after acute vocal fold injury (early phase of wound healing) in a rat model [82]. The same authors also showed that preventive treatment with antioxidants significantly decreased ROS formation after acute injury [83] and weakened age-associated changes in the vocal folds [84]. In addition, our group recently observed evidence of oxidative stress response in the larynx of rabbits exposed to low humidity [30].

Mitochondrial translation-related proteins

Mitochondria are especially vulnerable to oxidative damage and are a significant source for intracellular ROS formation as a byproduct of the electron-transport chain in cellular respiration [77,85]. Mitochondrial ROS have been implicated in regulating cell signaling, immune responses, autophagy, and maintenance of cell homeostasis [86]. There is increasing evidence that the mitochondria also acts to minimize intracellular oxidative stress through a complex ROS defense system [85]. Mitochondrial DNA (mtDNA) is particularly susceptible to ROS-induced damage. mtDNA is also responsible for the transcription of all mitochondrial machinery, crucial for their primary mitochondrial function of providing ATP as the energy source to sustain cell viability [85]. Our data showed that several core mitochondrial proteins responsible for protein synthesis (TUFM, MRPL13, MRPL38, MRPL58, PTRH2, FARSB) and RNA processing (PIN4) are downregulated. Interestingly, one study of the proteome of S. cerevisiae and cultured human cells subjected to dehydration by desiccation also noted the downregulation of several core mitochondrial proteins [80]. Other studies have shown the downregulation of such proteins in response to different oxidants, suggesting a general defense mechanism against oxidative stress [87–89]. The downregulation of mitochondrial metabolism favors its ROS defense system and simultaneously prevents ROS accumulation by the mitochondria [85,88]. Only one mitochondrial protein was upregulated in our study, the Propionyl-CoA Carboxylase (PCCA), which has a protective role in response to intracellular ROS [90]. Although the fold change of PCCA was below the significant cut-off, its biological impact in the dehydrated vocal folds needs to be explored. Our data suggest that acute systemic dehydration induced by furosemide interferes with mitochondrial metabolism, likely in response to increased ROS formation. The increase of ROS in vocal fold tissue subjected to systemic dehydration warrants further investigation.

Oxidative stress-related proteins

Our study has also identified differential regulation of proteins with a documented role in oxidative stress response, such as the downregulation of the flavin-containing monooxygenase 2 (FMO2), Cavin-2 (SDPR), and the mitochondrial small heat shock protein HSPB8, as well as the upregulation of the protein S100A12. FMO2 has a significant role in the oxidative metabolism of various xenobiotics compounds [91]; however, studies have shown that FMO2 may also serve as a source of ROS in the form of hydrogen peroxide [92]. Cavin-2 is a membrane-bound protein responsible for the organization of the caveolae (small invaginations in the plasma membrane of several cell types). This protein also regulates the production of nitric oxide (NO) by endothelial cells [93]. HSPB8, also known as heat shock protein 22, has several cellular functions, including protection against oxidative stress in a NO-dependent mechanism in cardiac mitochondria [94,95]. The downregulation of HSPB8 in the vocal folds of dehydrated rabbits may indicate reduced tolerance to oxidative stress. The last protein from this group, S100A12, is a protein that regulates inflammatory and immune responses acting as an “alarmin”, released in response to cellular stresses. In addition to serving as an activator of the innate immune response, its protective functions include oxidant scavenging, and its presence may indicate tissue oxidative stress [96]. Notably, the gene for S100A12 was found downregulated in our previous transcriptome study [17]. The differences between the transcriptome and proteome may reflect the complexity of posttranscriptional and posttranslational changes that can impact the relationship between mRNA and protein abundances [97,98]. Nevertheless, our data demonstrate that oxidative stress-related proteins are differentially expressed in the vocal fold tissue of furosemide-systemically dehydrated rabbits.

Study limitations and future investigations

Posttranscriptional and posttranslational events affect RNA and protein stability and abundance [31,97,98] having a major impact when comparing transcriptome and proteome analysis. Therefore, the differences between the set of differentially regulated vocal fold proteins in this study and the set of vocal fold genes differentially expressed in our previous study with furosemide-induced systemically dehydrated rabbits [17] are not surprising. Furthermore, it is worth mentioning that the proteomic data in this study was obtained from a different cohort of rabbits, and batch effects may also impact the results observed between the transcriptome and the present proteome study [99]. In addition, a large number of proteins associated with muscle were identified across all vocal fold samples; this is not surprising considering the higher representation of muscle tissue compared to the mucosal layer in the analyzed specimens. A study targeting the different layers of the vocal folds (epithelium, lamina propria, thyroarytenoid muscle) is warranted to investigate the tissue-specific contributions in response to systemic dehydration. Another limitation of this study is the use of male rabbits only, which was a choice to avoid female sex hormonal fluctuations that knowingly interact with body fluid-regulatory hormones [56–58]. Our next studies will include female animals to expand the findings of vocal fold biology to both sexes. In sum, the findings from this study increase our understanding of the molecular effects of systemic dehydration on rabbit vocal folds, especially at the level of the proteome. Specific to this study, the induction of systemic dehydration by furosemide, a known diuretic which adversely impacts the voice, serves to increase our understanding of the molecular underpinnings of laryngeal tissue change.

Conclusion

In the present study, we successfully applied an established systemic dehydration protocol with rabbits to investigate the proteome of vocal fold tissue. The protein fractionation method and the label-free quantitation LC-MS/MS approach identified more than 1,600 proteins in the euhydrated and dehydrated vocal folds, representing muscle and mucosa. This proteome will provide a database for future studies investigating the rabbit larynx/vocal folds and associated disorders. Finally, the proteome of acute systemically dehydrated rabbit vocal folds reveals novel insights on the response to a moderate dehydration level (~5%) that involves the downregulation of mitochondrial metabolism likely to prevent oxidative damage.

Supplementary Material

1

Fig. S1 (TIFF). Principal component analysis with all proteins identified across all eight rabbit samples. Complete list of proteins is provided in Table S1.

2

Table S1 (Excel). Proteins identified in the vocal folds of euhydrated (control) and systemically dehydrated rabbits, with at least one unique peptide and at least two MS/MS counts. Data was filtered as described in the Materials and methods section and LFQ (label-free quantitation) values were Log2 transformed before comparison between groups. Proteins with p < 0.05 are considered significantly differentially regulated in the dehydrated group. Proteins highlighted in yellow represent extracellular matrix components.

3

Table S2 (Excel). Pathway and Process Enrichment Analysis of all proteins identified in the vocal folds of euhydrated and systemically dehydrated rabbits. Analysis was performed with Metascape.

4

Table S3 (Excel). Venn diagram results of proteins exclusively identified in either control-euhydrated or furosemide-dehydrated group.

5

Table S4 (Excel). Enriched Gene Ontology (GO) biological processes for proteins differentially regulated (p ≤ 0.05) in the vocal folds of dehydrated rabbits. Enrichment analysis was performed with Metascape.

Significance:

Voice disorders affect different population demographics worldwide with one in 13 adults in the United States reporting voice problems annually. Vocal fold systemic hydration is clinically recognized for preventing and treating voice problems and depends on optimal body hydration primarily achieved by water intake. Herein, we use the rabbit as a translatable animal model, and furosemide as a translatable method of systemic dehydration, to reveal a comprehensive proteomic profile of vocal fold mucosa and muscle in response to systemic dehydration. The significant subset of proteins differentially regulated due to furosemide-induced dehydration offer novel insights into the molecular mechanisms of systemic dehydration in the vocal folds. These findings also deepen our understanding of changes to tissue biology after diuretic administration.

• The diuretic furosemide causes systemic dehydration affecting vocal folds

• Vocal fold proteome of euhydrated and furosemide-dehydrated rabbits were analyzed

• More than 1,600 proteins were identified by LC-MS/MS in the rabbit vocal folds

• Furosemide-induced systemic dehydration regulates vocal fold mitochondrial proteins