Profibrotic COVID-19 Subphenotype Exhibits Enhanced Localized ER-Dependent HSP47⁺ Expression in Cardiac Myofibroblasts in Situ

Abstract

We recently described a subgroup of autopsied COVID-19 subjects (~40%), termed ‘profibrotic phenotype,’ who exhibited clusters of myofibroblasts (Mfbs), which were positive for the collagen-specific chaperone heat shock protein 47 (HSP47⁺) in situ. This report identifies increased, localized (hot spot restricted) expression of αSMA, COLα1, POSTN and FAP supporting the identity of HSP47⁺ cells as myofibroblasts and characterizing a profibrotic extracellular matrix (ECM) phenotype. Coupled with increased GRP78 in COVID-19 subjects, these data could reflect induction of the unfolded protein response for mitigation of proteostasis (i.e., protein homeostasis) dysfunction in discrete clusters of cells. ECM shifts in selected COVID-19 subjects occur without significant increases in either global trichrome positive staining or myocardial injury based quantitively on standard H&E scoring. Our findings also suggest distinct mechanism(s) for ECM remodeling in the setting of SARS-CoV-2 infection. The ratio of CD163⁺/CD68⁺ cells is increased in hot spots of profibrotic hearts compared with either controls or outside of hot spots in COVID-19 subjects. In sum, matrix remodeling of human COVID-19 hearts in situ is characterized by site-restricted profibrotic mediated (e.g., HSP47⁺ Mfbs, CD163⁺ Mφs) modifications in ECM (i.e., COLα1, POSTN, FAP), with a strong correlation between COLα1 and HSP47⁺cells within hot spots. Given the established associations of viral infection (e.g., human immunodeficiency virus; HIV), myocardial fibrosis and sudden cardiac death, early screening tools (e.g., plasma biomarkers, noninvasive cardiac magnetic resonance imaging) for diagnosis, monitoring and treatment of fibrotic ECM remodeling are warranted for COVID-19 high-risk populations.

Introduction:

Life expectancy in the US has either declined or flattened, an event which coincides with the recent pandemic whose impact continues to reverberate among disproportionately affected communities¹. Fibrosis, an adaptive response to tissue injury in the heart and other tissues, causes organ failure accounting for one-third of deaths worldwide². Both lung and extrapulmonary fibrotic manifestations contribute to the mortality of COVID-19 but the long-term morbidity from this response is not yet fully understood^3–5. We have reported increased levels of the endoplasmic reticulum ER molecular chaperone GRP78, a co-receptor of the spike protein of SARS-CoV-2, during activation of the unfolded protein response (UPR) by proteotoxic stresses within both pneumocytes and macrophages of archival lung samples from autopsied COVID-19 subjects⁶. In the current work, we focused on the UPR in hearts.

Along with immune and hypercoagulability factors, the post-acute sequalae of SARS-CoV-2 infection (PASC) at 12 months are associated with increased risk of atrial fibrillation, pulmonary embolism, heart failure, and stroke^7,8, despite limited evidence of direct cardiac infection with the virus⁹. In an era of precision medicine, human autopsies are unmatched for diagnostic accuracy, training and continuous medical education, quality assurance and especially for the clinicopathological adjudication of unexpected sudden cardiac death (SCD)¹⁰. The landmark report of Human Immunodeficiency Virus Post-Mortem Systematic Surveillance of Sudden Cardiac Death (HIV POST SCD), prospectively adjudicated the causes of sudden death at autopsy in San Francisco County. This study clearly established that the >2-fold increased risk for arrhythmias-induced SCD for HIV positive compared with residents without HIV, was causally linked to interstitial myocardial fibrosis¹¹. We recently have performed histopathological studies to identify a subgroup of autopsied COVID-19 subjects, termed ‘profibrotic phenotype,’ who exhibit strikingly increased molecular chaperone heat shock protein 47 (HSP47⁺) in cardiac myofibroblasts, as well as increased CD163⁺-monocyte derived macrophages and collagen α1(l)(COLα1)¹². Herein, we investigated activation of the UPR-dependent cardiac extracellular matrix (ECM) remodeling in situ^13,14 in COVID-19 infections. Among 40% of afflicted COVID-19 individuals, we report the profibrotic subgroup exhibits localized enrichment in myofibroblasts of ER-resident HSP47⁺, POSTN⁺, FAP⁺ and client proteins (e.g., COLα1 and fibronectin) in situ¹⁵. Myofibroblasts and client proteins collectively play essential roles for synthesis, secretion, assembly, and maturation of the fibrotic ECM^15–17. Furthermore, collagen-specific HSP47⁺, POSTN⁺, FAP⁺, αSMA⁺ myofibroblasts are co-localized and regionally distributed with increased CD163⁺/CD68⁺ macrophages in the restricted clusters of cells in the heart. These pathognomonic changes of profibrotic events were not observed in controls, COVID-19 infected subjects without “hot spots”, or in tissues outside of the hot spots in profibrotic COVID-19 subjects. We discuss potential implications of our findings for diagnosis, risk stratification and future therapeutic inventions based on UPR dependent, profibrotic signaling networks in viral-infected patients.

Methods:

Archival autopsy tissues for the investigations described in this work were made available from the Department of Pathology and the Medical College of Wisconsin Tissue Bank under IRB-approved protocols. De-identified demographic and clinical data were made available from electronic health records under IRB-approved protocols.

Subject selection:

Archival autopsy tissues were obtained from patients admitted through the emergency department and hospitalized at the Froedtert & the Medical College of Wisconsin regional hospital (Milwaukee, WI). For COVID-19 patients, we identified those in which the primary cause of death was SARS-CoV-2 pneumonia, excluding those with evidence of either myocardial infarction (MI) or acute ischemic injury, cardiac arrest, endocarditis, clinical pericarditis, myocarditis, cardiac tamponade, tumor involving the heart, pulmonary embolism, bacterial pneumonia, or bacterial sepsis. APACHE-II scores were calculated for all COVID-19 subjects at the time of hospital admission because this instrument has the greatest area under the curve compared to qSOFA, SOFA, or SAPS-II scores in individuals with COVID-19¹⁸ for predicting mortality. Data for calculation of the APACHE-II scores were extracted from the electronic medical record.

Diagnosis of COVID-19 was based on positive SARS-CoV-2 PCR tests and consistent CT chest, lab tests, and clinical exam. For control subjects (n=18), we identified candidates in which the primary cause of death was believed to have limited impact on the heart or lungs (advanced cancers or pancreatitis or cirrhosis). Control patients were selected for demographics (race, age, gender, etc.) that matched those of the COVID-19 cohort. Finally, we specified that hearts were within normal limits on gross exam and microscopically. For all subjects, demographic and clinical data were collected from electronic medical and autopsy records in a deidentified manner using an honest broker (see Table 1a and 1b).

Table 1a.

Patient characteristics (represented as n (%) or median (range)

	COVID-19 Patients			Controls
	Overall (n=12)	Profibrotic Hotspot (n=5)	Non-fibrotic (n=7)	Overall (n=18)	P-value*	P-value**
Age, years	65 (28–89)	69 (28–89)	63 (35–71)	65 (31–83)	0.98	0.57
Age group, years					0.97	0.90
< 40	2 (17)	1 (20)	1 (14)	2 (11)
40–49	0 (0)	0 (0)	0 (0)	2 (11)
50–59	2 (17)	1 (20)	1 (14)	3 (17)
60–69	5 (41)	1 (20)	4 (58)	6 (33)
70–79	2 (17)	1 (20)	1 (14)	3 (17)
≥ 80	1 (8)	1 (20)	0 (0)	2 (11)
Sex, female	7 (58)	2 (29)	5 (71)	9 (50)	0.72	0.56
Race/Ethnicity					0.06	0.36
Black	8 (66)	3 (60)	5 (71)	9 (50)
Hispanic	2 (17)	0 (0)	2 (29)	0 (0)
White	2 (17)	2 (40)	0 (0)	9 (50)
Vaccine status a					0.12	0.15
1st dose	2 (17)	2 (40)	0 (0)	0 (0)
Fully vaccinated	1 (8)	0 (0)	1 (15)	0 (0)
Partially vaccinated	0 (0)	0 (0)	0 (0)	2 (14)
Not vaccinated	9 (75)	3 (60)	6 (85)	12 (86)
Symptom to onset, days	21 (8–52)	21 (15–43)	17 (8–52)	--	--	0.63
Pre-Existing Condition
Dyslipidemia	3 (25)	1 (20)	2 (29)	7 (39)	0.69	>0.99
Hypertension	8 (67)	2 (40)	6 (86)	11 (61)	>0.99	0.22
Type 2 diabetes	4 (33)	1 (20)	3 (43)	4 (22)	0.68	0.58
Heart failure	4 (33)	1 (20)	3 (43)	6 (33)	>0.99	0.58
Coronary artery disease/myocardial infarction	0 (0)	0 (0)	0 (0)	1 (6)	>0.99	--
End stage renal disease	4 (33)	1 (20)	3 (43)	3 (17)	0.39	0.58

^*Overall COVID-19 vs controls

^**profibrotic Hotspot vs non-fibrotic COVID patients

^a4 controls did not have vaccine status available

Table 1b.

	Non-fibrotic (n=7) n (%)	Profibrotic (n=5) n (%)	P-value
ICU stay	7 (100)	4 (80)	0.42
Intubated	6 (86)	3 (60)	0.52
ECMO	0 (0)	2 (40)	0.15

Microscopic studies:

High resolution images from H&E sections of our cases were used for scoring of injuries common in hearts of subjects with COVID-19^7–9. Table 2 shows the scoring system, which was applied in a blinded manner to all subjects by a trained pathologist (PN). Pathological descriptions developed for autopsy cases were separately employed to assess for the presence of hypertrophy, old MI, endotheliitis, pericarditis, or hypertensive changes.

Table 2.

H & E Scoring System

	Assigned Score
Inflammation
No significant interstitial/perivascular inflammation	0
Focal inflammation and edema	1
Widespread inflammation with or without necrosis	2
Microthrombi
None	0
Focal	1
Widespread	2
Pericarditis
None	0
Focal	1
Widespread	2
Total H & E Inflammatory Score (Sum of Inflammation, Microthrombi and Pericarditis)	Total
H & E Fibrosis Score (Non-infarction)
None	0
Focal	1
Moderate	2
Severe	3
Is infarction present?	+/−

Immunohistochemistry (IHC) was utilized to quantify either distinct cell types or collagen. By using adjacent serial sections, we were able to quantify CD163⁺, FAP⁺, and other markers within areas enriched in HSP47⁺ cells. Primary antibodies used for both immunohistochemical and immunofluorescence staining included mouse anti-human HSP47 (monoclonal, Santa Cruz Biotechnology, Dallas, TX; Cat# sc-5293), rabbit anti-human Periostin (polyclonal, Sigma Aldrich, St. Louis, MO; Cat# HPA012306), rabbit anti-human FAP (monoclonal, Abcam, Cambridge, MA; Cat# ab207178), mouse anti-human Cardiac Troponin T(cTnT) (monoclonal, Abcam, Cat# ab8295), rabbit anti-human Collagen alpha 1(i) (monoclonal, Abcam, Cat# ab138492), DAPI (Akoya Biosciences, # FP1490), mouse CD163 (monoclonal, Leica Biosystems, Deer Park, IL; Cat# NCL-L-CD163) and CD68 (monoclonal Antibody KP1, Thermo Fisher Scientific, Grand Island, NY; Cat# 14-0688-82). A monoclonal rabbit antibody to GRP78 (BiP (C50B12) Rabbit mAb; Cell Signaling, Danvers, MA; Cat#3177; 1:200) and Envision Plus polymer (Agilent/DAKO) were used for GRP78 quantification.

Immunohistochemistry was performed using an Omnis autostainer (Agilent/DAKO, Santa Clara, CA). Deparaffinized sections of hearts from either COVID-19 decedents or from patients with non-COVID-19 conditions were incubated with primary antibodies to HSP47, POSTN, FAP, collagen alpha 1(I) (COLα1), cTNT, or CD163. High resolution Joint Photographic Experts Group (JPEG) images were used for the studies detailed below.

Multiplex immunofluorescent (mIF) staining was performed using an Opal 7-Color Automation IHC Kit (Akoya Biosciences, Marlborough, MA) to detect colocalization of cell types and markers. An optimized multiplex assay for HSP47, POSTN, FAP, COLα1, cTNT, CD163 and DAPI was conducted on a Leica Bond Rx fully automated autostainer. Biomarkers that could co-localize in the same cellular compartment were paired with a spectrally separated Opal fluorophore to avoid potential spectral interference, as recommended by the manufacturer. All mIF slides were scanned on an Akoya PhenoImager (Akoya Biosciences, Marlborough, MA) at 20 X using MOTiF^™ protocol, which generates a single unmixed whole slide scan of up to 7 colors. This single image facilities a rapid application of DIA across the entire slide in a streamlined workflow, without the requirement of stitching many spectrally unmixed image tiles.

Quantification of IHC slides:

In 5 of 12 subjects, groups of cells enriched for HSP47⁺ were observed at low power and marked digitally in high resolution JPEGs. After examination of several clusters of HSP47⁺ cells, ten images at 20X were captured inside a representative hot spot by one investigator (ERJ) with ten additional images captured outside a hot spot. In control and non-fibrotic hearts, ten images (20X) were acquired in representative areas. Values for the 10 images were averaged to provide a single value for each subject. To quantify HSP47, CD163, CD68, COLα1, POSTN and FAP images, colors from the IHC images in high resolution jpegs were deconvoluted¹⁹ from blinded images by a second investigator (GR). The respective blue channel image was thresholded and watershedded to represent the brown immune-positive stain appropriately and quantify the corresponding percent area of the slide, using image J software. In the cases of HSP47, POSTN, FAP, CD68, CD163 and COLα1, quantification of IHC signal inside a hot spot was compared to that outside of a hot spot in the same subjects. These methods were used as an alternative to visual counting of immune positive cells within a defined area to enhance scientific rigor¹². Selected studies on the spatial distribution of either HSP47⁺ or CD163⁺ cells in left ventricle (LV), right ventricle (RV) or septal regions of hearts that utilized hand counting of positive cells in a separate cluster from those evaluated by ImageJ are identified in the text. To further identify localization of HSP47⁺ cells, clusters in profibrotic subjects or individual cells in control and non-fibrotic subjects were classified as present or absent in the myocardium, endocardium, or pericardium (Table 4). This determination (positive or negative) was made by one of the pathologists who served as an author on the paper after setting a visual threshold and reviewing all sections of the high resolution JPEGs from control and COVID-19 subjects.

Table 4.

Localization of HSP47+ cells.

	N sections (subjects)	Myocardium	Endocardium	Pericardium
Control	36 (14)	7	7	14
Non-fibrotic	17 (7)	0	2	8
Profibrotic	13 (5)	10	4	10

Masson’s Trichrome:

Cardiac tissue sections were stained with Masson’s trichrome to determine the secreted fibrillar collagens as a marker of global ECM fibrosis²⁰. The staining procedure employed a plasma stain followed by a phosphotungstic phosphomolybdic acid solution followed by a collagen-fiber stain. The sections were pretreated with hot Bouin’s solution to intensify the staining colors. The red shades of muscle and cytoplasm were favored by phosphotungstic phosphomolybdic acid. The collagen fibers take up the tungstate ion and the aniline blue is subsequently bound to this complex, coloring the collagen blue. Fibrosis was observed as blue color staining. The percentage of blue area was quantified using ImageJ (NIH) software with a custom developed macro for automated image analysis for the magnitude of fibrosis²¹. Because fibrillar collagen was heterogeneously distributed throughout the tissue samples and easily quantified by our macro at low power (as opposed to antibody based IHC), images at 2.5X were captured and used for trichrome quantification. Based on the size of tissue samples, we obtained 3–5 non-overlapping images per sample which covered ≥80% of the area of the slide. Values of individual trichrome images were averaged for a single percent area per subject for statistical comparisons.

Statistical analysis.

Data were reported as n (%) or median and range. Chi-square test or Fisher’s exact test was performed to compare categorical variables. Non-parametric Mann-Whitney-Wilcoxon test was used to compare the differences in biomarkers between COVID-19 patients and controls while Kruskal-Wallis test was used to compare that among profibrotic, non-fibrotic, and controls, respectively. Wilcoxon signed rank test was performed to compare the biomarkers inside versus outside the hotspots. Box plot with jittered points was used to display and illustrate the distribution of the data. Linear mixed models or generalized linear mixed models were used to assess the various markers inside versus outside the hotspots and/or among the different groups within LV, RV, or septal region of the heart. Spearman or Pearson correlation coefficient was calculated to examine the relationships between variables. P<0.05 was considered statistically significant. The analysis was performed in SAS 9.4 (SAS Institute, Cary, NC, USA) and SPSS 28.0 (IBM Corp., Armonk, NY).

Results:

Demographics of selected patients who died from SARS-CoV-2 pneumonia (n=12) between June 2020 and December 2021, as well as controls (n=18) are shown in Table 1a. COVID-19 subjects are subdivided into those with HSP47⁺ hot spots, and those without (see below for explanation). Subjects with COVID-19 died in an acute care setting of respiratory causes; multiorgan system failure was present in five of twelve. The median (range) duration of symptom onset to death (SOTD) was 21 (8–52) days. Based upon the dominant COVID-19 variants in North America at the time of their illness, these subjects presumably presented with the D614G (n=5), alpha (n=2), or delta (n=5) strains of the virus²². Blacks (66%) outnumbered Hispanics (17%) and Whites (17%) among the COVID-19 decedents. There were equal numbers of Whites (50%) and Blacks (50%) among non-COVID-19 control subjects (n=18) who died of nonhematopoietic solid tumors without heart involvement during the same timeframe (Table 1a). Neither sex nor vaccine status was different among the groups. Pre-existing comorbidities such as coronary artery disease/myocardial infarction (CAD/MI), hypertension (HT), heart failure, end stage renal disease (ESRD), dyslipidemia and type 2 diabetes mellitus (T2DM) are listed in Table 1a. These comorbidities were similar between control and COVID-19 subjects. To further assess clinical features that might distinguish COVID-19 subjects with and without clusters of HSP47⁺ cells, we compared ICU admission, intubation, or treatment with ECMO in non-fibrotic and profibrotic COVID-19 subjects (Table 1b). Finally, we assessed APACHE II severity of illness scores on admission in these two COVID-19 groups. These data appear in Figure 1. None of the variables were different in the two groups, which might be due to the small numbers with limited statistical power.

Figure 1.

A box plot shows APACH- II scores on hospital admission for COVID-19 subjects stratified by non-fibrotic and profibrotic status. Scores were compared using Mann-Whitney-Wilcoxin; there were no differences between the groups in severity of illness on presentation.

Next heart tissue was studied by light microscopic examination of H&E stained sections. Representative images appear in Figure 2, injury scoring criteria in Table 2, and the scored injury analyses summarized in Table 3. Our clinicopathological data show LV hypertrophy and hypertensive changes in all groups, with p values for hypertensive and all variables in hearts of COVID-19 subjects versus control subjects ≥0.47(Table 3), similar to data previously reported by Bois⁹. Table 3 also shows that H&E scores for COVID-19 associated global, classical injury patterns were overall low and similar between samples from control and COVID-19 patients, data consistent with reports from other investigations of cardiac pathology from patients infected with SARS-CoV-2²³.

Figure 2a-c.

Representative H&E images of hearts from control (n=7) and COVID-19 patients (n=7 non-fibrotic and 5 profibrotic). COVID-19 hearts exhibited scattered focal myocarditis, and pericarditis, but overall, the histological appearance of the tissue was relatively preserved in all COVID-19 subjects relative to control.

Table 3.

Scoring H & E Injury

	Controls (n=5)	Non-fibrotic COVID (n=7)	Profibrotic (n=5)	P-value*	P-value**	P-value***
Histological Changes
Hypertrophy	+2/5 mild	5/7	5/5	0.65	0.62	>0.99
Box car nuclei	0/5	1/7	2/5	>0.99	0.47	0.57
Hypertensive changes noted	+2/5 mild	3/7	3/5	>0.99	>0.99	>0.99
Old MI	+1/5 milda	2/7	0/5	>0.99	>0.99	0.51
Injury Score	Controls (n=7)	Non-fibrotic COVID (n=7)	Profibrotic (n=5)	P-value*	P-value**	P-value***
Inflammation				>0.99	0.22	0.22
None	6	6	2
Focal	1	1	3
Microthrombi				>0.99	>0.99	>0.99
None	7	7	5
Focal	0	0	0
Pericarditis				0.46	0.99	0.47
None	7	5	5
Focal	0	2	0
Total Injury Score, median (IQR)b	0 (0–0)	0 (0–1)	1 (0–1)	0.71	0.20	0.43
H & E Score, median (IQR)b	1 (0–2)	1 (1–2)	1 (1–1.5)	0.37	0.79	0.53

^amild subendocardial fibrosis

^bhighest value

^*Control vs non-fibrotic COVID

^**Control vs Profibrotic COVID

^***Non-fibrotic vs Profibrotic

Because cell-specific (i.e., fibroblast, cardiomyocyte, endothelial) targeted deletion studies in mice showed that myofibroblast-specific HSP47 was essential for deposition of Type I and V collagen, post-injury scar formation and survival²⁴, we next examined sections of hearts immunostained for HSP47 at low power. HSP47⁺ cells were strikingly increased in irregular clusters between 1 and 4 mm in diameter (Figure 3c), typically with many clusters (e.g., 2–10) per myocardial tissue sample in 5/12 (42%) in COVID-19 samples. These clusters were termed “hot spots.” COVD-19 subjects with hot spots were termed “profibrotic” based on the known role of HSP47 as an ER-resident and collagen-specific chaperone^25,26 as well as precedent for the use of this term in a cardiac phenotype produced by mechanical strain²⁷. COVID-19 subjects without hot spots were termed “non-fibrotic”, and these groupings were maintained for all variables examined throughout the remainder of the study. HSP47⁺ myocardial cells were typically spindle-shaped and morphologically consistent with either activated fibroblast or myofibroblast. Higher power insets of HSP47⁺ clusters appear in Supplemental Figure S1. HSP47⁺ cells in much lower densities were interspersed in cardiac tissue from controls and non-fibrotic COVID-19 subjects (Figure 3a and b), particularly in the pericardium and pericardial fat. Hot spots were not observed in either controls or non-fibrotic COVID-19 subjects.

Figure 3.

Low power view showing representative images from control (a) and non-fibrotic subject (b) as well as hot spots of HSP47⁺ cells in cardiac tissue of a subject dying with SARS-CoV-2 (c). All displayed images were globally adjusted for white balance and color saturation, using Microsoft Powerpoint, to highlight the “hot spots”. Hot spots appear as irregularly shaped clusters of brown cells which were distributed throughout the left and right ventricles as well as septal region (c). Hot spots were not observed in control or in ~60% of subjects dying with COVID-19 (a,b). Given the median size of hot spots (~2 mm diameter), number of hot spots per tissue section (~5), hot spots covered between 3 and 6% of the tissue area.

Figure 3d-g. Representative IHC images from profibrotic hot spots comparing morphology of cells positive for HSP47, αSMA, FAP and POSTN in profibrotic cardiac tissue. HSP47 and αSMA appear largely intracellular, in spindle shaped cells (see green arrows), whereas FAP and POSTN staining is densest in extracellular regions (see yellow arrows). The appearance of cells positive for these 4 markers was similar in control and non-fibrotic cardiac tissue (data not shown), though positive cells for all of these markers were much less frequent in tissue from the latter two groups.

We next quantified the spatial and regional distribution of HSP47⁺ cells in separate sections of the heart. Table 4 shows the localization of HSP47⁺ cells with respect to the myocardium, endocardium, or epicardium. In controls, 36 sections from 14 subjects were analyzed. Seven sections exhibited scattered HSP47⁺ cells in the myocardium and endocardium, while 14 exhibited the same expression of cells in the pericardium. Out of 17 sections from 7 non-fibrotic subjects, two subjects exhibited isolated HSP47⁺ cells in the endocardium and 8 in the pericardium. None were localized to the myocardium in the non-fibrotic group. Out of 13 sections from profibrotic subjects, HSP47⁺ clusters were in LV, RV, and septal regions (Figure S2). HSP47⁺ clusters of cells were identified in the myocardium (10) and pericardium (10), with fewer hot spots in the endocardium (4). HSP47⁺ cells in all regions of either non-fibrotic subjects or controls were much lower than that of profibrotic subjects (p<0.01). No HSP47+ cells were noted in a perivascular location in any subjects.

To validate the specificity of HSP47⁺ for myofibroblasts in situ, we examined three complementary biomarkers including the intracellular α-SMA and two others associated with ECM remodeling, POSTN, and FAP. Figure 3d–g shows the IHC appearance of cells and ECMs positive for these individual markers. Both HSP47⁺ and α-SMA⁺ appear as antibody-positive spindle shaped cells interspersed among cardiomyocytes. As expected, uptake for both markers was largely either cytoplasmic or cytoskeletal. In contrast, staining for both FAP and POSTN appeared predominantly extracellular (i.e., perivascular and interstitial in location) but was also interspersed with cytoplasmic staining (Figure 3f–g). Such findings are consistent with COVID-19-induced shifts of these matrix proteins in ECM remodeling. POSTN expression exhibited either excellent or good correlation to HSP47⁺ cells in 5 of 5 samples with hot spots per pathologists’ (AP and PN) estimation (data not shown). However, POSTN was also increased in and around some cells with histological appearance of endothelium or pericardial cells in non-fibrotic COVID-19 and control samples, and in areas which were not positive for HSP47. These data demonstrate localized increases in products of activated fibroblasts or myofibroblasts among a subset of patients with COVID-19 using four separate, independent and complementary markers of this cell type during ECM remodeling. To begin to assess the unfolded protein/ER-stress response in COVID-19 samples, we compared GRP78 expression as detected by DAB in a pilot group of control and profibrotic COVID-19 samples (n=4 each group). Area positive for GRP78 was 2.7 ± 1.7% in controls compared to 17.3 ± 4.7% in hearts from COVID-19 subjects (p=0.027).

Figure 4 shows representative mIF images with overlap of some cells positive for three of these biomarkers, particularly in hot spots (i.e., from profibrotic subjects). Figure S3 for mIF images obtained in the absence of primary antibodies serves as our negative control. In quantified IHC images, we observed striking increases in the percent area of slides positive for HSP47 within hot spots of profibrotic (up to 3–6%) compared with either non-fibrotic or control subjects (less than 0.04%; Figure 5a, p=0.003 and p<0.001, respectively). These data support our primary contention of hot spots. Like HSP47, FAP was increased in COVID-19 subjects with differences being observed between non-fibrotic and profibrotic inside hot spots, as well as inside versus outside hotspots (Figure 5b). We have previously reported increased COLα1 expression in cardiac samples from patients with COVID-19 relative to controls¹². Figure 6a–c shows mIF images with overlap of COLα1 and HSP47 only in the profibrotic representative image (see figure legend; median of 16 cells per 40X image in a pilot study). Increases in COLα1 and POSTN inside (relative to outside) hot spots in profibrotic subject’s hearts were identified (Figure 6d–e; p<0.05). COLα1 increased with age of the subjects in hearts of all subjects (Figure S4; Spearman’s Rho 0.46; p=0.02). This relationship held true even amongst control subject values, with Spearman’s Rho for that group 0.66; p=0.014 (data not shown).

Figure 4.

Representative multiplexed immunofluorescent staining of control (a,d,g), non-fibrotic (b,e,h), and profibrotic COVID-19 hearts (c,f,I,j) show HSP47⁺(light blue), POSTN⁺(green), and FAP⁺ (red) cells sections from control, non-fibrotic or profibrotic subjects. Profibrotic images are taken inside a hot spot. The magnification of Figure 4J is increased relative to other images in order to demonstrate colocalization of the markers (see scale bar).

Cardiomyocytes are marked by grey (cTnT), and nuclei visualized by royal blue DAPI staining. Few individual HSP47⁺ cells were observed in cardiac tissue from control and 7/12 COVID-19 subjects. In contrast, clusters of HSP47⁺ cells were observed in 5/12 profibrotic COVID-19 patients.

Like HSP47⁺, POSTN⁺ and FAP⁺ were increased inside hot spots of profibrotic samples. Figures 4(j) shows overlap of HSP47⁺, POSTN⁺ and FAP⁺ cells, consistent with myofibroblasts.

Figure 5.

(a): Box plots with individual values of HSP47⁺ and FAP⁺ in control, non-fibrotic and profibrotic subjects, the latter divided into inside and outside of hot spot measurements. The numbers in each group appear in the graph. HSP47⁺ is shown on a log scale in order to appreciate the individual values in the control and non-fibrotic groups. The percentage of brown area was quantified using ImageJ (NIH) software with a custom developed macro for automated image analysis for the magnitude of brown staining. For profibrotic tissue, IHC density inside and outside hot spots was quantified separately. The percentage area for HSP47⁺ cells within hot spots of profibrotic subjects ranges from 0.2 to 2.2%, and only 0.015 to 0.047% in representative areas of controls were positive for HSP47⁺ cells. Statistical comparisons of groups appear in the figures. Note that values in control subjects were different from all groups except profibrotic outside (p=0.2), non-fibrotics are different from profibrotics outside and inside (p=0.01 and 0.003 respectively) and profibrotic outside expression is less than profibrotic inside (p=0.045).

Figure 5b. Box plots of individual values of FAP⁺ in 3 groups of subjects: control, non-fibrotic and profibrotic. The numbers in each group appear in the graph. FAP is increased in profibrotics inside versus outside (p<0.05), profibrotic inside versus non-fibrotics (p=0.03) and profibrotic inside versus controls (p=0.02).

Figure 6.

(a-e). Representative IF images for HSP47⁺ and COLα1⁺ are shown in control, non-fibrotic and profibrotic within hot spots(6a-c). Cardiomyocytes are marked with cTNT (grey) and nuclei with DAPI (blue). COLα1 and HSP47 are more highly expressed in hot spots. In a pilot study, we quantified cells from 10 control, 10 non-fibrotic and 10 profibrotic samples. These data showed no cells with either marker in 10 representative control images or in 10 non-fibrotic images. In contrast, 10 profibrotic images showed 8 to 25 (median 16) cells with both markers and DAPI. Thus, in our hands, with the antibodies and equipment available to us for IF studies, there is limited sensitivity in quantifying marker colocalization in tissue with lower expression of the markers.

Figures 6d and e depict quantification of COLα1 and POSTN inside and outside the hot spots using IHC images. Ns for each group and statistical comparisons appear in the figures. Both COLα1 and POSTN were higher inside than outside of the hot spots (p=0.04).

Because we observed localized increases in one form of collagen (COLα1), we next examined global trichrome staining to estimate deposition of collagen I and III. Representative examples of these stains are shown in Figure 7a–c, with grouped changes compared in Figure 7d. There were no differences in trichrome staining in any of the three subject groups.

Figure 7.

Representative trichrome images from three groups (control, non-fibrotic COVID-19 and profibrotic COVID-19) are shown (6a-c). Collagen appears blue. Because trichrome was heterogeneously distributed, it was quantified at a sufficiently low image magnification that precluded distinction between inside or outside hot spots in profibrotic subjects. Figure 6d shows comparisons of quantified trichrome blue staining in the three groups, with “n”s appearing within the graph. All P values are greater than 0.5.

Figure 8 shows HSP47⁺ expression as a function of trichrome stain density in individual subjects. Overall correlation analysis showed no significant relationship between trichrome and HSP47⁺ (Figure 8a). There was a negative relationship between HSP47⁺ and trichrome staining in tissue from control subjects (Figure 8b; Spearman’s Rho −0.57; p=0.03).

Figure 8.

a. HSP47⁺ percent area in control (n=13) and COVID-19 subjects (n=12) is shown as a function of trichrome stain density in individual subjects. There is no significant relationship between these variables in the cohort overall (p=0.54).

Figure 8b. When the controls were separated from the COVID-19 subjects, Spearman’s Rho= −0.57; p=0.03 for this group alone. In contrast, no significant correlation was found for COVID-19 subjects.

Figure 8c shows a positive correlation between Colα1 and HSP47+ across all subjects (Spearman’s Rho 0.61, p=0.002). In data not shown, the Spearman’s Rho for the five profibrotic subjects alone was 0.8, p=0.10.

Figure 8d. Colα1 is plotted as a function of subject’s age in control (n=12) and COVID-19 (n=12) subjects. There is a positive correlation between these variables across all groups (Spearman’s Rho =0.48; p=0.02). When separated into COVID-19 subjects and controls, a positive relationship holds for control subjects only (Figure 7e; Spearman’s Rho=0.61; p-0.02)

In contrast to no relationship between HSP47⁺ with trichrome positivity, there was a strong relationship between COLα1 and HSP47⁺ cells, both markers measured within hot spots for the profibrotic subjects (Figure 8c; Spearman’s Rho 0.61; p=0.02).

HSP47 is reported to increase with age in healthy subjects²⁸. We observed no relationship between HSP47⁺ cells and age in subjects overall (Figure S5; p=0.48). There was also no relationship between HSP47⁺ and symptom onset to death across all COVID-19 subjects (Figure S6; p=0.10). Of interest, there was a statistically significant increase in HSP47⁺ in non-fibrotic COVID-19 subjects as the SOTD increased (Figure S7; Spearman’s Rho 0.79; p=0.04).

Because macrophages are reported to partner with myofibroblasts in profibrotic or inflammatory process²⁹, we next assessed expression of total CD68⁺Mφs and selective CD163⁺ Mφs (M2). Like HSP47, CD163⁺ cells were distributed throughout the myocardium, with differences between control, non-fibrotic and profibrotic subjects evident in all regions (Figure S8). Representative mIF images are shown in Figure 9a–f. As expected, CD68⁺ cells are widely distributed in samples from control and COVID-19, both non-fibrotic and profibrotic tissues. In mIF images, colocalization of HSP47⁺ cells with CD163⁺ and some CD68⁺ cells is apparent.

Figure 9.

Representative IF images of 2 markers for macrophages (CD68 (green) and CD163 (red) from controls (a,d,g,j), non-fibrotic (b,e,h,k) and profibrotic (c,f,i,l), the latter within hot spots. Colocalization of these macrophages with HSP47⁺ cells (light blue) was identified. Juxtaposition of HSP47⁺ and CD163⁺ (Figure 9l) was subjectively observed more frequently than that of HSP47⁺ and CD68⁺ macrophages. Statistical comparisons appear in Figure 10.

To quantify CD68⁺ and CD163⁺ cells, IHC was employed. The lowest values in CD68⁺ macrophages were found in profibrotic hearts outside of hot spots, and expression was less than that from either non-fibrotic COVID-19 or control subjects (Figure 10a). Expression of CD68⁺ cells in profibrotic subjects was less outside than inside hot spots (p<0.05). Despite statistical differences, the overall CD68⁺ cell expression across groups was relatively more similar than different; median values for percent CD68⁺ areas in control, non-fibrotic, profibrotic outside and profibrotic inside were 2.05, 2.36, 1.32 and 1.58, respectively. CD163⁺ cells in profibrotic heart tissues were higher inside than outside hot spots (Figure 10b; p<0.05). CD163⁺ cells from non-fibrotic subjects were expressed at levels similar to those inside profibrotic hot spots. The ratio of CD163⁺/CD68⁺ cells was elevated in profibrotic hot spots (inside) compared with outside hot spots or those of controls (Figure 10c).

Figure 10.

Quantification of CD68 (a) and CD163 (b) counts and CD163/cd68 (c) in the groups as shown. “n”s appear in the figures as do statistical comparisons. CD68 percent positive area is statistically higher in profibrotic inside versus outside (p<0.05), and lower in profibrotic outside than controls (p=0.03), but there is overall overlap and not large differences amongst the groups. CD163 is increased in profibrotic hot spots versus outside the hot spots (p<0.05), but like CD68, there is overlap in the groups. The ratio of CD163/CD68 is highest in the profibrotic hot spots (inside) and higher inside than outside hot spots (p<0.05), and lower in controls versus inside hot spots (p=0.013), though there is overlap in individual values of these ratios as well.

We next explored the relationship between CD163⁺ and HSP47⁺. There is a non-significant relationship between CD163⁺ and HSP47⁺ cells across all subject groups (Figure S9; p=0.64 overall). We also examined the relationship between SOTD and age in CD163⁺ cells. Our data show no significant relationship between CD163⁺ cells and increasing SOTD in COVID-19 subjects (Spearman’s Rho −0.18; p=0.59; Figure S10). There was also no significant relationship between CD163⁺ cells and age (Figure S11).

Figure S12 shows a 7 color image with colocalization of markers including HSP47, CD173, COLα1, periostin, FAP in several cells.

Discussion:

In the present work, we have established that in situ hot spots containing both profibrotic HSP47⁺ myofibroblasts and profibrotic CD163+Mφ positive cells are enriched for ECM myocardial proteins in situ^24,29–31. These histomorphological hot spots were identified first by clusters of the most distinctive profibrotic molecular feature—HSP47⁺ cells (Figure 2C and 4a) but are also characterized by increased FAP⁺ and COLα1. We considered localized increases expression of HSP47⁺ cells as necessary but insufficient alone in identifying myofibroblasts. Staining with complementary biomarkers for αSMA, POSTN and FAP as well as subcellular location, shape and appearance of biomarkers support our assertion that hot spots are enriched with myofibroblasts (Figures 3). Accordingly, a distinct subgroup can be histologically and unambiguously identified using specific cellular biomarkers in subjects with median duration of symptom onset to death (SOTD) of only 21 days in ~40% fatal cases with COVID-19. Like profibrotic HSP47⁺ cells, CD163⁺ cells within hot spots were increased in all areas of the heart examined (Figure S8). Cardiac tissue with molecular characteristics of profibrotic hot spots did not exhibit globally enhanced extracellular matrix deposition in situ as determined by Masson trichrome staining (Figure 6). We interpret the differences in “fibrosis” based on trichrome and COLα1 or HSP47 as indicating that either (i) pre-existing global fibrosis does not augment the risk of localized, COVID-19 associated fibrogenesis or (ii) subjects with hot spots may be at risk of increased trichrome in the future because of the short time frame between onset of symptoms and death. Additional studies are needed to clarify this relationship. Either way, a ‘cardiac profibrotic phenotype’ for this distinct subgroup of autopsied COVID-19 hearts does not address the question of progression to trichrome positive fibrosis, given that each subject represents a “snapshot in time”. However, our observations raise unresolved public health questions for fibrosis pathogenesis with important implications COVID-19 survivors in the following years.

With a median age of 65 years and overrepresentation of black subjects, the demographic and comorbidity features of our COVID-19 subjects mirrors that of populations most afflicted by the pandemic in 2020 and 2021³². The high prevalence (~40%) of our novel profibrotic subgroup determined at postmortem examination mimics a similar percentage of COVID-19 involvement assessed by cardiac magnetic resonance imaging³³. We compared clinical features including APACHE-II severity of illness score, ICU admission, intubation or ECMO in non-fibrotic and profibrotic subjects, and found no differences. However, a study with higher sample size may be needed to determine whether differences in these endpoints between the COVID-19 subgroups exist or not.

Sudden cardiac death (SCD) due to arrhythmias have been causally linked to increased interstitial myocardial fibrosis among the HIV-positive persons compared with persons without known HIV¹¹. Known positive HIV status and serological evidence of recent Epstein-Barr virus (EBV) reactivation are independently associated with increased symptoms (e.g., fatigue, neurocognitive dysfunction) for long COVID-19 in a cohort of 280 adults with SARS-CoV-2 infections³⁴. Indeed, UPR upregulation in CD163⁺ macrophages has been reported during EBV infection and post-transplantation proliferative disorder³⁵.

Our study is noteworthy for absence of obvious myocardial injury including either myocyte necrosis or lymphocytic infiltrates based on H&E characteristics (see Table 3) despite evidence of increased HSP47⁺ and GRP78⁺, marked localized changes in ECM and cellular shifts. The absence of differences in global injury by H&E (Table 3) fits with the similarities in trichrome staining across groups. The clinical significance for cardiac profibrotic phenotype in situ is presently unknown for COVID-19 survivors.^36,37. Accordingly, we use the term “profibrotic” similar to a phenotype evoked by mechanical strain in human mitral valvular interstitial cells that upregulates ECM including COL1A1²⁷.

A question of intense interest is the underlying mechanism(s) for activation of the integrated stress response (ISR) and clustering of ER-dependent ECM biomarkers that characterize the cardiac profibrotic/-inflammatory phenotype. Pleiotropic effects such as protein misfolding and aggregation of the cytokine storm may induce collagen-specific HSP47^26,38,39, but does not seem to account for distinct hot spots such as we observed. Likewise, we observed induction of the secretory protein POSTN, which is synthesized in the ER and transported by the secretory pathway for subsequent assembly of the ECM architecture¹⁵. Endothelial barrier dysfunction and virion breakthrough is reported in hearts of individuals infected with COVID-19; hot spots could represent such foci⁴⁰. Our study has potential mechanistic implications for immune dependent cross-talk between CD163⁺Mφ and HSP47⁺, POSTN⁺, FAP⁺, αSMA myofibroblast cells^29,30. ISR-dependent pathways could lead to turnover in the ER-Golgi then ECM deposition characteristic of fibrotic remodeling of the heart ^15,16, but further studies to verify an ER-stress response in COVID-19 hearts are needed. As Adler and colleagues have suggested²⁹, self-amplifying cross talk between macrophages and myofibroblasts during tissue inflammation via ISR-dependent secretion of platelet derived growth factor (PDGF) and monocyte colony stimulating factor (M-CSF) could occur.

In COVID-19 hearts, we observed increases in the ratio of CD163⁺/CD68⁺ macrophages (Figure 9a–c) inside HSP47⁺ hot spots compared with outside. There was, however, no relationship between CD163⁺ and HSP47⁺ cells across all subjects (Figure S9). CD163⁺ cells have been reported to be increased in the lungs of subjects dying with SARS-CoV-2, though not influenza A, and SARS-CoV-2 triggers a fibrosis-associated transcriptional profile in monocytes³¹. To our knowledge, our report is the first describing clusters of cells with increased CD163⁺/CD68⁺ and HSP47⁺ myofibroblasts associated with ECM remodeling. Increased plasma levels of soluble CD163 have been observed in children with SARS-CoV-2⁴¹, all data consistent with a possible pathophysiological role for CD163⁺ cells in COVID-19 induced myocardial responses.

Our study has several limitations. Given the small numbers of patients (n=12 COVID-19), p values for many correlations between variables (such as correlations between HSP47⁺ and CD163⁺) do not reach significance. These are provided as hypothesis generating datasets based on Spearman’s Rho values ≤−0.3 or >0.3 in different subject groups. While our prior work is applicable to predominant delta variant of SARS-CoV-2, we are cautious to extrapolate to the sublineages of the B.1.1.529 (omicron), and more recent variants⁴². Likewise, our studies did not assess for the possibility that detectable cardiotropic, invasive virions might be implicated in the disease we identified in deceased COVID-19 subjects⁴³. A direct cause and effect relationship between SARS-CoV-2 infection and localized ISR response cannot be established based on observations in this work at a single point in time.

Future directions.

A formidable public health crisis looms on the horizon of the COVID-19 pandemic, including myocardial fibrosis with susceptible arrythmias^44–46, and injury from exuberant immunity^47,48, hyper-inflammatory responses^49,50 and multiorgan failure^5,51,52. In the most comprehensive analyses using the electronic health record (EHR) repositories, the cardio-renal system (34%) dominated among highly reproducible (e.g., respiratory, musculoskeletal, digestive) injury subphenotypes when assessed between 30–180 days after acute SARS-CoV-2 infection⁵³. Recent studies have demonstrated the sensitivity, specificity, accuracy, reproducibility, and clinical validation at single cell resolution for profibrotic sub-phenotyping. Studies to define other cell types within clusters of HSP47⁺ cells are of intense interest.

Non-invasive examination using cardiac magnetic resonance imaging (cMRI) with late gadolinium enhancement (LGE) has successfully identified a pathological phenotype with long-term outcomes of SCD from myocardial involvement with histologically diagnosed cardiac sarcoidosis when adjudicated with either human autopsy or transplantation^54,55. Clinical studies of plasma markers alongside cMRI evidence of cellular and matrix remodeling should be pursued especially for effective management of high-risk populations numbering in the millions for viral-mediated cardiac injury.

Highlights.

• COVID-19 has a profibrotic subphenotype.

• HSP47 is a profibrotic diagnostic biomarker.

• Collagen1α1 is increased in hot spots.

• Hot spots represent myofibroblasts.

• COVID-19 elicits host-pathogen interactions.