Identifying Phenotypically Distinct Fibroblast Subsets in Type 2 Diabetes–Associated Iatrogenic Laryngotracheal Stenosis

Abstract

Objective.

Iatrogenic laryngotracheal stenosis (iLTS) is the pathologic narrowing of the glottis, subglottis, and/or trachea secondary to intubation or tracheostomy related injury. Patients with type 2 diabetes mellitus (T2DM) are more likely to develop iLTS. To date, the metabolomics and phenotypic expression of cell markers in fibroblasts derived from patients with T2DM and iLTS are largely unknown.

Study Design.

Controlled in vitro cohort study.

Setting.

Tertiary referral center (2017-2020).

Methods.

This in vitro study assessed samples from 6 patients with iLTS who underwent surgery at a single institution. Fibroblasts were isolated from biopsy specimens of laryngotracheal scar and normal-appearing trachea and compared with controls obtained from the trachea of rapid autopsy specimens. Patients with iLTS were subcategorized into those with and without T2DM. Metabolic substrates were identified by mass spectrometry, and cell protein expression was measured by flow cytometry.

Results.

T2DM iLTS-scar fibroblasts had a metabolically distinct profile and clustered tightly on a Pearson correlation heat map as compared with non-T2DM iLTS-scar fibroblasts. Levels of itaconate were elevated in T2DM iLTS-scar fibroblasts. Flow cytometry demonstrated that T2DM iLTS-scar fibroblasts were associated with higher CD90 expression (Thy-1; mean, 95%) when compared with non-T2DM iLTS-scar (mean, 83.6%; P = .0109) or normal tracheal fibroblasts (mean, 81.1%; P = .0042).

Conclusions.

Scar-derived fibroblasts from patients with T2DM and iLTS have a metabolically distinct profile. These fibroblasts are characterized by an increase in itaconate, a metabolite related to immune-induced scar remodeling, and can be identified by elevated expression of CD90 (Thy-1) in vitro.

Laryngotracheal stenosis results in cicatricial airway scar that affects voice and breathing, frequently progressing to respiratory failure.¹ Iatrogenic laryngotracheal stenosis (iLTS), secondary to postintubation or posttracheostomy injury, is the most prevalent etiology.² Patients with iLTS require more frequent surgical intervention, including temporary and permanent tracheostomy.^2,3 The propensity for severe disease in iLTS is in part due to the biology of pathologic fibroblasts and increased comorbidities, which adversely affect wound healing.⁴ The most prevalent comorbidity in iLTS is type 2 diabetes mellitus (T2DM), which is present in up to 53% of all patients with iLTS.^2,5,6 Clinically, patients with T2DM are more likely to develop acute laryngeal injury⁷ as well as interarytenoid scar resulting in posterior glottic stenosis.⁶ Biologically, T2DM iLTS-scar fibroblasts are more contractile and metabolically distinct than non-T2DM iLTS-scar fibroblasts.⁸ With a rising incidence of T2DM nationally,⁹ there is a growing need to understand how T2DM affects iLTS-scar fibroblasts and wound healing in laryngotracheal injury.

The metabolic pathways involved in T2DM iLTS-scar pathogenesis remain largely unknown. Metabolomics, or the study of metabolite substrates, can isolate unique metabolic pathways that are critical in fibrotic diseases such as iLTS.^10,11 In T2DM iLTS-scar fibroblasts, there is an increased mitochondria energy demand as compared with non-T2DM iLTS, suggesting a role for metabolic alteration in disease.⁴ In other diseases, such as idiopathic pulmonary fibrosis, lung biopsy metabolomics analyzed through mass spectrometry show alterations in multiple metabolic pathways when compared with nondiseased controls.¹² Moreover, unique metabolic profiles are present in asthma-related airway fibrosis and phenotypically correlate with distinct cellular markers allowing for quantification of fibroblasts subsets.¹³ Elucidating the metabolic profile of T2DM iLTS-scar fibroblasts may provide insight into iLTS pathogenesis.

Identifying activated fibroblast subsets helps to distinguish the pathologic effector cell responsible for fibrotic disease.^14-16 Many studies have defined fibroblast subsets based on expression of phenotypical intra- and extracellular markers. For instance, fibroblast activation protein (FAP), a surface glycoprotein, is one marker expressed on fibroblasts that is linked to pathologic phenotypes in other fibrotic diseases.^17,18 Murine models of rheumatoid arthritis also contain pathologic synovial fibroblast subsets that express FAP⁺CD90⁺ via flow cytometry and are associated with inflammation.¹⁷ CD90 (Thy-1) expression is further elevated in fibroblasts derived from thyroid-associated ophthalmopathy (formerly known as Graves’-associated ophthalmopathy), which are more contractile than non-CD90⁺ fibroblasts.¹⁹ CD34, an extracellular marker prevalent on undifferentiated progenitor cells, is also seen on fibroblasts, and may be another marker for activated fibroblasts in wound healing.²⁰ Identifying cell surface markers in T2DM iLTS-scar fibroblasts will allow us to identify pathologic subsets driving disease, which can be a target for intervention.

We performed a discovery study to investigate the metabolic profiles of T2DM and non-T2DM iLTS-scar fibroblasts. We hypothesized that T2DM iLTS-scar fibroblasts have a unique and metabolic profile distinct from non-T2DM iLTS-scar fibroblasts and nonscar fibroblasts. We further hypothesized that T2DM iLTS-scar fibroblasts have specific cell surface markers that can identify activated fibroblast subsets responsible for a more severe phenotype in T2DM iLTS.

Methods

Fibroblast Isolation and Cell Culture

Endoscopic biopsy samples of airway scar or normal tissue were taken from 6 patients following informed consent in accordance with the Johns Hopkins University Institutional Review Board (NA_00078310). Patients with iLTS were divided into those with and without a preoperative diagnosis of T2DM. All primary fibroblasts were isolated, cultured, and expanded as previously described.^21,22 In brief, biopsy specimens were cut into Petri dishes with a scalpel blade. Cells were grown in standard Dulbecco’s modified Eagle media + 10% fetal bovine serum + 1% antibiotic-antimycotic solution (Thermo Scientific).²² Passages 0 to 2 were used for the duration of this study. Normal controls were derived from either nonscar tracheal mucosa in patients with non-T2DM iLTS or autopsy specimens processed within 24 hours of death.²³

Mass Spectrometry Metabolomics

Fibroblasts (passages 1 and 2) were passaged onto 10-cm Petri dishes under standard culture conditions. Fibroblasts were grown to confluence, washed twice with phosphate buffered saline, and then collected by mass spectrometer–grade methanol (80%) and a cell scraper. Samples were stored in a −80 °C freezer until all cell lines were collected to minimize the freeze-thaw cycle and perform analysis in bulk. Once thawed, samples were sonicated (Hielscher Ultrasonic Technology) in three 1-second pulses to release cellular contents and placed on a rotator (Thermo Scientific) at low speed overnight. Samples were then centrifuged at 4 °C and the supernatant collected. Cell pellets were isolated and refrozen at −80 °C for total protein quantification. The supernatant (metabolite containing) was dehydrated with nitrogen gas at low flow until there was a 50% reduction in volume. Samples were again centrifuged to remove additional cellular debris, and the supernatant was reconstituted in mass spectrometer–grade acetonitrile. Samples were then analyzed with a Sciex 5500 quadrupole LC/MS mass spectrometer.²⁴ Sufficient data quality was available for 71 metabolites (Supplemental Table S1, available online) based on peak expression. Fibroblasts were grouped by etiology: non-T2DM iLTS-scar (n = 3), T2DM iLTS-scar (n = 3), and normal tracheal biopsies from patients with iLTS (n = 2). Each metabolite peak intensity (mz/rt) was normalized through total protein (μg/mL) and analyzed with MetaboAnalyst 4.0 software.^25,26

For total protein concentration, sample pellets were thawed and lysed with an AllPrep RNA/Protein kit (Qiagen). Total protein was then measured with the Qubit Protein Assay Kit (Thermo Scientific).²⁷ Metabolite peaks were subsequently normalized relative to total protein concentration.

Flow Cytometry

Cultured fibroblasts (passages 0-2) were grown in culture to 80% to 90% confluence and trypsinized, as previously described.⁴ Single-cell suspensions were aliquoted to a 96-well plate. Cells were stained with the following antibody panel: LIVE/DEAD Fixable Aqua, CD31 (PECAM) Brilliant Violet 421 (303124; BioLegend), CD34 PerCP-CY5.5 (343612; BioLegend), FAPa PE (FAB3715P-025; R&D Systems), CD90 Alexa Fluor 647 (328120; BioLegend), and PDGFR (platelet-derived growth factor receptor) PE/Cy7 (323508; BioLegend). After staining, cells were fixed and permeabilized. Single-stained and florescence minus one–stained samples were used as controls for gate selection. Viability Aqua–negative (live) cells were subsequently evaluated by percentage population of fibroblasts (FAP⁺). Fibroblast cells line were grouped according to etiology (n = 3), with rapidly processed autopsy tracheal mucosa–derived nonscar fibroblasts used as controls. All analyses were performed in FlowJo Flow Cytometry Analysis Software (Treestar, Version 10.7.1 for Mac).

Statistical Analysis

Metabolomic data were analyzed through standard unpaired t tests for direct comparison between T2DM and non-T2DM iLTS-scar fibroblasts. A volcano plot—log₂(fold change) versus statistical significance (log₁₀[P], P < .05)—was performed to determine metabolites that were elevated in T2DM versus non-T2DM iLTS fibroblasts. A Pearson R coefficient was calculated for sample and metabolite heat map correlations. Partial least squares discriminant analysis was performed, where R² is an estimate of fit and Q² is an estimate of predictive ability of the model and is calculated via cross-validation.²⁸ Data analysis was performed with the MetaboAnalyst 4.0 metabolomic web server (metaboanalyst.ca).²⁶

Flow cytometry results were presented as percentage expressed (mean ± SEM). Multiple groups were compared through analysis of variance with Tukey correction for multiple comparisons. Significance criterion for all analyses was set at P < .05. Data analysis was performed with Prism software (GraphPad, Version 8).

Results

Experimental Cohort

Patient demographics were similar between groups (Table 1). No significant differences were noted in age, sex, or BMI. Patients with T2DM and iLTS had higher Charlson Comorbidity Index than those with non-T2DM iLTS and normal tracheal biopsies.

Table 1.

Patient Demographics.^a

	T2DM (n = 3)	Non-T2DM (n = 3)	P value
Age, y	54 (38-72)	48.3 (29-64)	.84
Sex: female	3 (100)	2 (66)	.16
Body mass index	35.6 (26-49)	32.3 (29-39)	.13
Hemoglobin A1C	6.3 (6.1-6.7)
Tobacco use
Current	0	0
Former	1	1
Never	2	2
Cotton-Meyer grade			.2733
1	0	0
2	0	0
3	2	3
4	1	0
History of tracheostomy	1	3
Medication
Metformin	2
Insulin	2
Oxygen therapy	2	0
Inhaled steroids	0	0
PPI	1	0
Comorbidity
Asthma	0	0
COPD	0	0
Depression	1	1
GERD	2	0
Hemiplegia	1	0
Hypertension	2	2
Obstructive sleep apnea	0	0
Solid tumor	0	0
Inflammatory tissue disease	0	0
Charlson Comorbidity Index	5 (4-8)	1.3 (0-5)	.006
Level of stenosis
Posterior glottis	2	0
Subglottic	2	3
Tracheal	0	1
Multilevel	1	1

Abbreviations: COPD, chronic obstructive pulmonary disease; GERD, gastroesophageal reflux disease; PPI, proton pump inhibitor; T2DM, type 2 diabetes mellitus.

^aBlank cells indicate not applicable. Values are presented as mean (range) or No. (%).

T2DM iLTS-Scar Fibroblasts Exhibit a Unique Metabolic Fingerprint vs Non-T2DM iLTS-Scar and Control Fibroblasts

A Pearson coefficient heat map of samples relative to metabolites illustrates distinct metabolite fingerprints in T2DM iLTS-scar and nonscar fibroblast samples (Figure 1). Non-T2DM iLTS fibroblasts were less clustered and more metabolically heterogeneous, most notably in sample S5. Sample-to-sample comparison via a Pearson R coefficient heat map showed that T2DM iLTS-scar fibroblast cell lines were metabolically correlated and that non-T2DM iLTS fibroblasts were more heterogenous (Figure 2a). A 2-component partial least squares discriminant analysis (Q² = −0.74, R² = 0.86) demonstrated 2 distinct groups metabolically as a function of etiology (T2DM vs non-T2DM; Figure 2b).

Figure 1.

Heat map of samples vs metabolites. Pearson R coefficient heat map of fibroblast samples vs metabolites detected by mass spectrometry. Normal, red; non-T2DM iLTS-scar fibroblasts, green; T2DM iLTS-scar fibroblasts, blue. Color scale range: −3 (blue) to +3 (red). iLTS, iatrogenic laryngotracheal stenosis; T2DM, type 2 diabetes mellitus.

Figure 2.

Clustering of iLTS-scar fibroblasts. (A) Intersample Pearson R coefficient heat map with strong correlations between T2DM iLTS-scar samples (box with white border). (B) Partial least squares discriminant analysis scoring chart shows nonrandom clustering representing distinct metabolic profiles. Shaded areas = 95% confidence regions. iLTS, iatrogenic laryngotracheal stenosis; T2DM, type 2 diabetes mellitus.

Mitochondria-Related Metabolites Are Elevated in T2DM iLTS-Scar Fibroblast

Data were collected on a total of 71 metabolites, with 8 peaks (1.9%) excluded from subsequent analysis due to an insufficient signal-to-noise ratio. Metabolic discovery (Figure 3) revealed that 1 metabolite, itaconate, was elevated in T2DM iLTS-scar fibroblasts (2.0×, P = .045). Two additional metabolites, riboflavin and pyridoxine, trended toward an increase (9.5×, P = .125; 1.6×, P = .056, respectively) in T2DM iLTS-scar. Acetylcarnitine (0.061, P = .007) and 2-aminooctanic acid (0.22×, P = .008) were significantly decreased in T2DM iLTS-scar as compared with non-T2DM iLTS-scar. Metabolites in the glutamine pathway (glutamate and hydroxyproline) were elevated (21× and 17×) in non-T2DM iLTS fibroblasts, although these did not reach statistical significance.

Figure 3.

Metabolic discovery. Volcano plot between non-T2DM (green/left) and T2DM (right/blue) iLTS-scar fibroblasts. Values represents an individual metabolite and expression in T2DM vs non-T2DM iLTS-scar fibroblast samples. Itaconate (2.0×; P = .045) was elevated in T2DM iLTS-scar fibroblasts. iLTS, iatrogenic laryngotracheal stenosis; T2DM, type 2 diabetes mellitus.

FAP⁺ CD90⁺ Subsets Are Predominant in T2DM iLTS-Scar Fibroblasts

Figure 4a shows the gating strategy. The fibroblast marker FAP (>98% expression) was used for subsequent gating of cells (Figure 4b). T2DM iLTS-scar fibroblasts (n = 3) had a significantly higher expression of FAP⁺CD90⁺ cells (mean expression, 95%; P = .004) as compared with non-T2DM iLTS-scar fibroblasts (n = 3, mean = 83.6%, P = .0109) or controls (n = 3, mean = 81.1%, P = .0042). There was no difference between non-T2DM iLTS-scar and normal tracheal fibroblasts. FAP⁺PDGFR⁺ fibroblasts were less common (<5% expression) in all groups except for a T2DM iLTS-scar sample, which was slightly elevated at 11%. Less than 20% of fibroblasts expressed FAP⁺CD34⁺ and were not significantly different among the control, T2DM, and non-T2DM iLTS-scar groups. Less than 1% of living cells expressed CD31 (PECAM).

Figure 4.

High-CD90 expression in T2DM iLTS-scar fibroblasts. (A) Fibroblast flow gating schema. A, area; FSC, forward scatter; H, height; SSC, side scatter. (B) FAP⁺CD90⁺ fibroblasts were more prevalent in T2DM than in non-T2DM iLTS-scar fibroblasts (P = .011) and control (P = .0042). *P <.05. **P <.01. CD90, Thy-1; FAP, fibroblast activation protein; iLTS, iatrogenic laryngotracheal stenosis; PDGFR, platelet-derived growth factor receptor; T2DM, type 2 diabetes mellitus.

Discussion

This is the first study to investigate the metabolomics of T2DM iLTS-scar fibroblasts. T2DM iLTS-scar fibroblasts represent a metabolically homogenous population and are distinguishable from non-T2DM iLTS-scar fibroblasts in a number of metabolic pathways, including elevated itaconate. T2DM iLTS-scar fibroblasts could be identified by increased CD90 (Thy-1) expression as compared with non-T2DM iLTS-scar and nonscar fibroblasts. Together, these data demonstrate that cultured T2DM iLTS-scar fibroblasts express an etiology-specific metabolic profile that can be identified by extracellular marker phenotyping.

Although in its infancy, metabolomics has become a powerful tool for identifying metabolite substrate changes on a cellular level.²⁹ In our study, we found clinically distinct patients with T2DM and iLTS to have scar fibroblasts that are characterized by a unique metabolic profile. Similar metabolic profiles were recently noted in dermal scar and thought to affect progenitor fibroblasts resulting in excessive scaring.¹⁰ However, correlating these metabolites to biologic pathways is difficult as few have reported on metabolite function in fibrosis. For instance, acetylcarnitine and 2-aminoctonic acid are involved with fatty acid synthesis, but neither has been studied in fibroblasts.³⁰ Despite these challenges, we found that T2DM iLTS-scar fibroblasts have higher expression of itaconate when compared with non-T2DM iLTS. As a metabolic by-product of the citric acid cycle, itaconate is highly activated after inflammation and reduces collagen production in fibroblasts derived from patients with systemic sclerosis in vitro.^31,32 As a biomarker, itaconate is correlated with disease severity in models of rheumatoid arthritis and is reduced with immunomodulatory treatment.³³ Riboflavin (vitamin B2) and pyridoxine (vitamin B6), 2 metabolites heavily involved with mitochondrial metabolism, also trended toward increased expression in T2DM iLTS-scar fibroblasts.^34,35 However, non-T2DM iLTS fibroblasts had an increased trend in levels of glutamate and hydroxyproline, 2 key components of glutamine metabolism that have been leveraged for inhibiting non-T2DM iLTS fibroblasts, which are more dependent on Warburg-like physiology (aerobic glycolysis).^4,21 The metabolic profile of T2DM iLTS-scar fibroblasts in this study identifies specific substrates that are consistently elevated in the cell lines studied here.

To characterize the fibroblasts within the metabolically homogenous T2DM iLTS-scar, we evaluated for fibroblast-associated cell markers in the T2DM iLTS fibroblast cell lines. First, the purity of the cells lines was confirmed by finding that 98% of cells expressed FAP, with minimal CD31 (PECAM) staining for endothelial cells. While all samples expressed CD90, T2DM iLTS-scar had the highest expression versus non-T2DM or control iLTS-scar fibroblasts. In a study of 117 patients with keloid or hypertrophic scaring, Ho et al found similar pathologic CD90⁺ fibroblasts, which were derived from a CD34-to-CD90 fibroblast transition and histologically correlated with worsening scar progression.³⁶ In other studies, fibroblasts expressing CD90 that were derived from arthritic joints in a murine model were proinflammatory and destructive when adoptively transferred into naïve joints, characteristic of an “activated” fibroblast subset.¹⁷ The activated CD90⁺ fibroblast subset in our study is predominant with T2DM iLTS-scar and may propagate the fibrotic phenotype.

While our study successfully characterized a metabolically distinct fibroblast subset in T2DM iLTS-scar, there are several limitations. The use of growth media, for instance, likely affects metabolic processes by providing a surplus of amino acids and growth factors for replication. Although cell starving is sometimes proposed to reduce the effects of growth media, cells that are dependent on aerobic (mitochondrial) metabolism are more sensitive to starvation than other subsets and therefore this method was not utilized as it could skew our results.³⁷ Moreover, the non-T2DM iLTS-scar fibroblasts in our study were more heterogenous (Figure 1), which may be secondary to the presence of additional fibroblast subsets within the non-T2DM iLTS-scar group. Cultured fibroblasts are also known to undergo morphological changes and partial differentiation into myofibroblast-like phenotypes under normal culture conditions after plating.^38,39 To limit the impact of dedifferentiation on identifying key fibroblast subsets in vitro, we used cells of low passage (≤2), which maintain greater similarity to primary cells. This subsequently reduced our sample availability and resulted in different control groups for the 2 experiments: tracheal biopsies from patients with iLTS in mass spectrometry and normal trachea from rapid autopsy specimens in flow cytometry experiments. Although we found a significant difference between CD90⁺ expression in T2DM and non-T2DM iLTS-scar fibroblasts, the clinical impact of CD90⁺ fibroblast subsets is still unknown. Future studies utilizing primary cells derived from dissociated tissue biopsy samples (prior to culture-induced changes) may better reflect laryngotracheal stenosis disease in vivo.

Conclusions

Clinically, patients with T2DM are at increased risk of developing iLTS. Distinct metabolic profiles exist within T2DM iLTS and are likely linked to clinical phenotype. In this study, we discovered a metabolically distinct phenotype in T2DM iLTS-scar fibroblasts that is correlated with elevated levels of itaconate. These T2DM iLTS-scar fibroblasts can be identified by increased expression of the extracellular marker CD90 (Thy-1) and represent an activated subset that may perpetuate fibrosis.