Alveolar Vascular Remodeling in Nonspecific Interstitial Pneumonia: Replacement of Normal Lung Capillaries with COL15A1-Positive Endothelial Cells

To the Editor:

Nonspecific interstitial pneumonia (NSIP) is an interstitial lung injury pattern, which can arise idiopathically or secondary to other causes (e.g., connective tissue diseases or drug toxicity). Histologically, fibrotic NSIP (fNSIP) is characterized by uniform widening of the alveolar septa due to homogeneous mild fibrosis with or without inflammation, while the lung architecture remains preserved (1, 2). Recently, we have reported an increased vessel diameter of the capillary network and increased intervascular distances in NSIP compared with control subjects (3), suggesting a vascular contribution to the pathogenesis of fNSIP.

Recent advances in lung vascular biology by single-cell RNA sequencing enabled the discrimination of two capillary cell types in the murine and human lung, termed aerocytes and general capillaries, reflecting previous characterizations of microvasculature zones (4–7). Aerocytes are hyperspecialized cells with a large surface area, which, together with the juxtaposed alveolar type 1 cell, form the air–blood barrier and thus enable gas exchange (4, 5). General capillaries, on the other hand, exhibit more regulatory functions and serve as the progenitor population to aerocytes (4, 5). In lung tissue from patients with idiopathic pulmonary fibrosis, we have observed by single-cell RNA sequencing a dramatic increase in a COL15A1⁺–expressing endothelial population that was transcriptionally indistinguishable from venous endothelial cells around airways and in pleura of healthy control subjects (8). Histologic validation localized COL15A1⁺ vessels to areas of dense fibrosis and to areas surrounding fibroblastic foci (6, 8).

Here, we further investigated the endothelial diversity in fNSIP, because, in contrast to idiopathic pulmonary fibrosis, the lung architecture is preserved, allowing for dissociation of the effect of marked remodeling from the effect of increased fibrotic matrix alone.

Methods

Six diagnostic surgical lung biopsy samples of patients with confirmed fNSIP from Yale University and six human lung explants of patients with end-stage fNSIP obtained during lung transplantation at Hannover Medical School (MHH) were included in this study (for basic patient characteristics, see Table 1). The nine nonfibrotic controls were obtained from either surgical size adjustment during lung transplantation or histologically tumor-free specimens from surgical cancer resections. Approval of the relevant ethics committees was obtained (MHH number 2702-2015, Yale 2000031225). Endothelial composition was evaluated by immunofluorescent microscopy using an established staining protocol and established staining panels as described elsewhere (6). Four targets were stained: aerocyte-specific HPGD, pancapillary PRX, COL15A1 for fibrotic endothelial cells, and panendothelial CD31. Regions of interest were analyzed using the ImageJ color threshold function of ImageJ, with the chosen color corresponding to the targeted protein of interest. We focused our analysis on the lung parenchyma by excluding large airways and large vessels. Randomly selected separate regions of interest were used for analyses within each sample, and nine replicates of intimal diameter and wall thickness measurements were made for each sample, with the mean for the sample reported. Values were compared using the Mann-Whitney U test. P values < 0.05 were considered significant.

Table 1.

Basic Patient Characteristics

	Control (n = 9)	NSIP Yale (n = 6)	NSIP MHH (n = 6)
Age, years	55.2 ± 16.4	51.0 ± 11.5	43.7 ± 13.2
Sex, female/male	3/6	5/1	4/2
Race, Black/White	n.a.	2/4	0/6
Diagnosis	n.a.	Myositis ILD (n = 2)	Myositis ILD (n = 1)
		CTD-ILD (n = 2)	SSC-ILD (n = 2)
		NSIP (n = 1)	CVID (n = 1)
		Unclassifiable (n = 1)	Unclassifiable (n = 2)
Serology	n.a.	Negative (n = 2)	Negative (n = 1)
		Jo1 (n = 1)	Missing (n = 3)
		Ro (n = 3)	SS-A/Ro (n = 1)
		La (n = 1)	Scl70 (n = 1)
		PL-7 (n = 1)
Immunomodulator therapy, % (n/N)	n.a.	100 (6/6)	67 (4/6)
FEV1% predicted	n.a.	76.5 ± 13.6	57.6 ± 30.9
FVC% predicted	n.a.	70.0 ± 13.6	42.6 ± 17.0
DlCO% predicted	n.a.	49.7 ± 7.5	24.5 ± 5.3*
SaO2 room air, %	n.a.	97.0 ± 1.0	87.0 ± 2.5*

Definition of abbreviations: CTD = connective tissue disease; CVID = common variable immunodeficiency; FEV₁ = forced expiratory volume in 1 second; ILD = interstitial lung disease; MHH = Hannover Medical School; n.a. = not available; NSIP = nonspecific interstitial pneumonia; Sa_O2 = arterial pulse oximetry measurement; SSC = scleroderma.

^*Only two data points were available for this measurement.

Results

In two independent cohorts and two clinical settings—diagnostic biopsies and lung explants—we observed a dramatic replacement of normal PRX+ lung capillary populations (including both aerocytes and general capillaries) with COL15A1⁺ endothelial cells (reduction of sample area fraction with normal capillaries from 95.6% ± 0.8% in control subjects vs. 21.6% ± 3.6% in NSIP, P < 0.01, Figures 1A and 1B; with significant increase in COL15A1⁺ staining of the sampled parenchymal tissue from 0.7% ± 0.3% to 13.9% ± 2.6%, P < 0.01, Figure 1C). The mean COL15A1⁺ area fraction of MHH explant samples was significantly higher than the biopsy samples from Yale (20.3% ± 2.6% compared with 7.6% ± 1.5%; P = 0.009). Within NSIP samples, fibrotic parenchymal regions also revealed greater endothelial density (7.2% ± 1.8%) compared with the normal (nonfibrotic) regions (2.4% ± 0.9%) within the same NSIP samples (data not shown; P < 0.001).

Figure 1.

(A) Representative immunofluorescent image staining for pan-endothelial CD31 and pancapillary PRX (first column), as well as aerocyte-specific HPGD and bronchial and fibrosis-specific COL15A1 (second column) in control and nonspecific interstitial pneumonia (NSIP) lungs. In control subjects, PRX stains bona fide lung capillaries, whereas COL15A1 stains bronchial vessels (white arrows) but not the alveolar microvasculature. In NSIP, PRX⁺ and HPGD⁺ microvasculature is mainly lost; inversely COL15A1⁺ vessels are dramatically expanded. However, small, circumscribed areas were found in NSIP with preserved bona fide lung capillaries (white dashed circles). (B) Comparison of sample area fraction containing PRX⁺ lung capillaries (i.e., aerocytes and general capillaries) in control and NSIP samples reveals a significant decrease in the area of PRX⁺ lung capillaries in the parenchyma of NSIP samples. (C) COL15A1 staining is significantly higher in NSIP samples. (D) Average microvascular diameter of COL15A1⁺ NSIP vessels was significantly greater than normal regions (HPGD⁺ COL15A1⁻) of NSIP and controls. MHH = Hannover Medical School.

In NSIP, small circumscribed areas were found with preserved lung capillaries with visually striking thin vascular walls (Figure 1A). Quantitative analysis revealed little difference between the thickness of the alveolar wall in control subjects (2.4 ± 0.1 μm) and in NSIP in areas with preserved HPGD⁺ COL15A1⁻ lung capillaries (2.3 ± 0.1 μm; P = 0.64; data not shown), but the vascular walls were significantly thicker in regions with HPGD⁻ COL15A1⁺ vessels in NSIP (3.7 ± 0.1 μm; P < 0.01). Furthermore, the diameter of COL15A1⁺ vessels (10.0 ± 0.5 μm; Figure 1D) was much larger than in lung capillaries of control subjects (6.6 ± 0.2 μm; P < 0.01) and regions with preserved lung capillaries of NSIP samples (6.4 ± 0.3 μm; P < 0.01; Figure 1D), matching our previous observations (3). Neither the intimal diameter nor the vascular wall thickness differed significantly comparing HPGD⁻COL15A1⁺ vessels of MHH samples with Yale samples (P = 0.10 and P = 0.5, respectively).

Discussion

Replacement of normal lung capillaries with COL15A1⁺ endothelial cells reflects alveolar vascular remodeling in NSIP, in which the parenchymal architecture of the lung is preserved but fibrotic thickening of the alveolar wall is occurring. When we compared early-stage disease diagnostic samples of the Yale cohort with the end-stage disease samples of the MHH cohort, we observed an association between the extent of the alveolar vascular remodeling and disease severity, in which a significant increase of COL15A1⁺ vessels associated with decreases in FVC and Dl_CO. Whether COL15A1⁺ cells are the result of a change in phenotype of native lung capillary cells, metaplastic or aberrant differentiation from local endothelial precursors, encroachment from venules or arterioles, or new cells migrating from the bronchial vasculature is not known. However, the consequences of the shift from differentiated lung capillaries to COL15A1⁺ blood vessels in NSIP are manifold. First, the obvious consequence of the loss of specialized capillaries is an impaired gas exchange in patients with NSIP. Second, lung capillaries are continuous, nonfenestrated vessels, whereas COL15A1⁺ vessels are fenestrated, as illustrated by the expression of the marker gene PLVAP (6), which may explain the increased lung permeability in fibrotic lung diseases (9). Third, aerocytes are the major source of HPGD (hydroxyprostaglandin dehydrogenase). HPGD is responsible for the first-pass degradation of most prostaglandins by the lung (6, 10), suggestive of disruption of prostaglandin homeostasis in patients with NSIP.

In conclusion, NSIP is characterized by the replacement of alveolar capillaries with COL15A1⁺ blood vessels. The role of COL15A1 in lung fibrosis is unclear, highlighting the need for further research regarding the endothelial contribution to the pathogenesis of NSIP and to fibrotic lung diseases in general.