The human monocytic cell line, THP-1, from the ATCC was cultured in RPMI 1640/10% FBS. THP-1 cells (4 x 10⁵/mL) were differentiated into a macrophage-like phenotype in 6-well plates with 0.2 µM phorbol-12-myristate-13-acetate (PMA) for 3 days followed by 48 hours in growth medium without PMA (23, 24). PMA-differentiated THP-1 cells were transfected with 2.5 µg ssRNA hY3 (a noncoding Y ssRNA binding Ro60 and Toll-like receptor 7/8 agonist), as described (22) for 18 hours. Transfection was carried out with a commercial kit (DOTAP Liposomal Transfection).

Fibroblasts from healthy fetal human hearts (aged 16-24 weeks) were prepared as previously described (7). Using passages 2-5, fibroblasts were serum-starved for >2 hours then co-cultured at 1:5 dilution with the supernatants of PMA-differentiated THP-1 macrophages transfected with hY3 (“hY3 sup”) or with DOTAP alone (“NT sup”) for 72 hours in low serum (0.1% FBS) medium. Alternatively, fibroblasts were treated with activated Transforming Growth Factor-Beta 1 (TGF-β1) (2.7 ng/mL) (R & D Systems); IFNalpha (IFNα) (2000 U/mL) (PBL Science); a neutralizing antibody to IFNα (12 µM) (PBL Science, MMHA-2 clone); or recombinant S100A4 (0.5 µg/mL) (Abcam) added directly to the medium.

In immunofluorescence experiments, fibroblasts were initially plated (50,000/mL) on glass coverslips, serum-starved, treated, fixed with 4% paraformaldehyde and permeabilized with acetone. Coverslips were then blocked with 0.1% gelatin in PBS and stained with a Cy3-conjugated antibody to α-SMA (Sigma-Aldrich) (1:150) and a FITC-conjugated antibody to S100A4 (1:100) (BioLegend). DAPI (Sigma-Aldrich) was utilized to stain DNA content. Coverslips were mounted with Mowiol (Polysciences); images were captured using a ZEISS Axio Observer inverted light microscope at 20x magnification then processed with ImageJ software.

For flow cytometry assays, fibroblasts were plated (3 x 10⁵/mL) in 6-well plate(s), serum-starved, treated, permeabilized and stained using the same α-SMA and S100A4 conjugated antibodies as above. 10,000 events were acquired on an LSR II; cells were pretreated with Ghost Dye Violet 510 (Tonbo) to ensure cell viability. Data was analyzed with FlowJo software.

RNA was isolated from human fetal cardiac fibroblasts (variously treated as detailed above but for 24 hours) using RNeasy Mini kit (Qiagen). 1 µg total RNA was used to prepare cDNA template with iScript cDNA Synthesis kit (BioRad). Primers for IL6, MMP13, IL10 and Col1α1 were selected based on primer sequences in the PrimerBank database (25). GAPDH was selected as an endogenous control. PCR reactions were performed in the StepOne Real-Time PCR System with Fast SYBR Green Master Mix (Thermo Scientific). The cycling conditions for all the genes were: 95°C for 3 min, 40 cycles of 95°C for 30 sec, 60°C for 20 sec, and 72°C for 30 sec. Relative changes in gene transcript were quantified as fold change based on 2^-ΔΔCt with relative expression of transcript normalized to GAPDH.

Fibroblasts were spiked with collagen (12 µg/mL) as a positive control and to compensate for varying basal collagen secretion. After 30 hours of treatment with NT sup, hY3 sup, or recombinant S100A4, fibroblast supernatants were collected. The concentration of soluble collagen in the supernatants was measured using the Sircol soluble collagen assay (Biocolor). Absorbance was recorded with a multiwell plate reader set to 540 nm. The standard curve was obtained by running parallel 10, 25, 50 µg collagen standards.

Untreated fibroblasts were plated on glass coverslips and treated for 12 hours with hY3 sup, NT sup or recombinant S100A4. The media was then aspirated and the cells were washed 3 times with PBS. After which, fresh growth media was added and the fibroblasts were cultured for 24 hours. The supernatants (conditioned media) of these were harvested and added to a new row of untreated fibroblasts on glass coverslips at 1:5 dilution. After another 24 hours these fibroblasts (co-cultured with the conditioned medium of the first round of fibroblasts) were fixed, permeabilized, stained, mounted, and photographed in accordance with the prior indirect immunofluorescence experiments.

For evaluation of the transcriptome related to the different subtypes of fibroblasts, we examined a previously published data set (26) generated from human fetal fibroblasts plated at a density of 5 × 10⁵ cells/well (6-well plate), serum starved, and incubated for 16 hours with the following: no treatment, NT sup, hY3 sup, and TGF-β1. Similarly, we examined another previously published data set (26) wherein fibroblasts from cardiac-NL hearts were sorted using a cell-sorting flow cytometer yielding a DAPI-negative, CD45-negative, CD31-negative, and podoplanin-positive subset. Fold change (ratio of log₂(transcripts per million)) was calculated relative to healthy hearts or untreated fibroblasts. For evaluation of the CHB leukocyte transcriptome related to vascular calcification, a previously published data set (27) generated from three healthy hearts and three CHB hearts—flow-sorted to yield a DAPI-negative, CD45-positive, podoplanin-negative, and CD31-negative population—was consulted.

Formalin-fixed paraffin sections (6 μm section) were obtained from all hearts. Sections of right ventricle, left ventricle, and AV groove/conduction (septal) tissue were obtained from 4 fetuses dying with CHB (ranging from 20-27 weeks gestation) and a normal human fetus (no known cardiac disease) electively terminated at 24 weeks of gestation. Isotype control staining was conducted in sections from CHB Heart #2. Spleen tissue was used as a positive control for α-SMA and S100A4 immunostaining.

Deparaffinization was achieved by warming up paraffin sections for 1 hour at 60°C. Sections were sequentially treated with xylene and ethanol (30 min each). Endogenous peroxidase was inactivated by treatment with methanol + 1% H₂O₂ for 30 min. Sections were washed with PBS then epitopes were unmasked by hyaluronidase digestion (type IV-S, Sigma-Aldrich 1 mg/ml; 0.1 M Tris, pH 6.0) for 30 min. Slides were once again washed with PBS and incubated overnight at 4°C with primary Ab (anti-α-SMA, anti-S100A4, rabbit IgG isotype control) then incubated with alkaline phosphatase-conjugated anti-rabbit IgG (1 hour at room temperature). Sites of immunoreactivity were visualized using Fast Red TR/Naphthol AS-MX Tablet (Sigma-Aldrich) to report. The sections were counterstained with hematoxylin before photomicroscopy. Subsequent slides were stained with hematoxylin and eoisin or Masson’s Trichrome stain.

Supernatants from THP-1 macrophages transfected with hY3 (hY3 sup) or DOTAP alone (NT sup) were assayed for secreted S100A4 via an in-house sandwich ELISA (blocked with 3% NFDM in PBS, developed with alkaline phosphatase and read with a multiwell plate reader at 405 nm). Serial dilutions of recombinant S100A4 (starting at 50 ng/mL) were run in parallel to generate a standard curve. This ELISA protocol was later expanded to umbilical cord blood samples (run in triplicate at 1:8 dilution) from (n=49) cardiac-NL and (n=47) healthy (anti-SSA/Ro-exposed) neonates. 1:8 dilution was determined after assay optimization. S100A4 ELISAs were run blinded to cord blood subject identification/respective severity score. Severity scores of clinical disease in the affected neonates were evaluated using a previously established severity score index that prioritizes most highly extranodal manifestations of cardiac-NL (endocardial fibroelastosis, dilated cardiomyopathy, and hydrops fetalis) with or without advanced AV block. Classification was based on disease status at birth. One neonate with transient mild EFE exposed to dexamethasone with a normal EKG was included in the no disease group. Cord bloods from three pairs of twin siblings with discordant disease were evaluated by ELISA; these pairs were color-coordinated on the graph accordingly. Data points were formatted to reflect medications taken by the mothers during affected and unaffected pregnancies, as follows: dexamethasone/betamethasone, hydroxychloroquine, dexamethasone/hydroxychloroquine in combination, dexamethasone/terbutaline in combination, IVIG, and dexamethasone/IVIG in combination. Neonates were categorized if exposed to maternal medication for any duration of the pregnancy. Many of the healthy neonates were part of the PATCH (Preventative Approach to Congenital Heart Block with Hydroxychloroquine) trial (28).

A selection of the same umbilical cord bloods as above was assayed for soluble Frizzled-Related Protein 1 (FRP-1) via sandwich ELISA commercial kit (Invitrogen). Bloods were diluted 1:6, run in triplicate, developed with horseradish peroxidase, and read with a multiwell reader at 450 nm in accordance with kit instructions. The same three pairs of twin siblings with discordant disease were evaluated for sFRP-1; the severity scores of clinical disease in the affected neonates were evaluated using the same previously established severity score index as above.

Statistical significance was established via Kruskal-Wallis test for flow cytometry experiments; Mann-Whitney U test was applied for qPCR, soluble collagen assays, and ELISA experiments. Analyses were performed using GraphPad Prism version 10.7.2. Values of p <0.05 were considered significant.

Previously reported fibrosis and calcification of the AV node on histologic evaluation of fetal hearts dying with cardiac-NL (1) motivated the interrogation of 4 similarly affected hearts (clinical features described in Table 1 ). Staining of AV nodal tissue with anti-S100A4 antibodies revealed expression in clusters of cells ( Figures 1A–D ) present only in areas with dystrophic calcification ( Figures 1F–I ) (calcification being a property in common among the cases). Cell morphology suggests S100A4 expression in both fibroblasts (spindle-shaped, red arrow) and macrophages (plump/sphere, blue arrow) ( Figures 1B, D ). Staining of AV nodal tissue of a matched healthy heart did not reveal S100A4⁺ cells ( Figure 1E ) or calcification ( Figure 1J ). Control for antibody isotype was conducted, using CHB Heart #2, to ensure S100A4 staining specificity ( Supplementary Figure 1 ). Staining a section of membranous interventricular septum (CHB Heart #1) or AV node (CHB Hearts #2,3) with hematoxylin and eoisin ( Figure 2C ) and with Trichrome (Masson) ( Figure 2A ) presented a complex pattern of dystrophic calcification and collagen deposition respectively at distinct areas of the tissue, including portions of epi and endocardium. Staining with anti-α-SMA antibodies revealed diffuse expression ( Figure 2B ) consistent with the localization of classically-defined myofibroblasts in regions of fibrosis (the blue staining of Trichrome (Masson) used to identify collagen deposition). In contrast, staining of tissue with anti-S100A4 antibodies ( Figure 2D ) revealed expression in clusters of cells limited to areas with calcification, as in Figure 1 . These histologic findings support separate anatomic locations, albeit it is acknowledged that α-SMA and S100A4 expression are not absolutely mutually exclusive, particularly in fields where fibrosis and calcification foci coincide. The next set of experiments addressed the hypothesis that there is a bifurcation of fibroblast roles leading to an anatomical dichotomy. To test this hypothesis directly, subsequent studies leveraged an in vitro model of cardiac-NL to evaluate whether anti-SSA/Ro autoantibodies provoke injury via a pathway involving S100A4⁺ fibroblasts.

Given macrophages and giant cells are frequently identified at calcified foci, our prior studies have incorporated macrophage activation and fibrosis as an in vitro model of cardiac-NL: ssRNA hY3 associated with SSA/Ro60 was leveraged as a proxy for immune complex uptake by macrophages and the addition of supernatants from these activated macrophages to cultured human fetal fibroblasts as a measure of macrophage-fibroblast crosstalk. When cultured with supernatants from hY3-transfected macrophages (hY3 sup), the fetal cardiac fibroblasts differentiated into two distinct populations, distinguished by the expression of α-SMA and/or S100A4 ( Figures 3C, G, K ). In contrast, direct addition of activated TGF-β1 to cardiac fibroblasts resulted in expression of α-SMA but not S100A4 ( Figures 3D, H, L ). Fibroblasts cultured in the absence of any additions ( Figures 3A, E, I ) or following treatment with supernatants of macrophages transfected with DOTAP alone (NT sup) did not generate substantial expression of either α-SMA or S100A4 ( Figures 3B, F, J ).

Flow cytometry was used to confirm the differential effects of the variously generated supernatants. As with the indirect immunofluorescence, fibroblasts co-cultured with hY3 sup exhibited marked expression of both α-SMA (red box, red histogram) and S100A4 (green box, green histogram) whereas fibroblasts treated with TGF-β1 only demonstrated a shift in the expression of α-SMA ( Figure 3 , red box, red histogram). No addition and NT sup fibroblasts revealed marginal expression of either marker ( Figure 3 , red, green boxes). The sizeable shift in α-SMA expression in both hY3 sup (Mean Fluorescence Intensity: 8263 ± 5185) and TGF-β1 (9085 ± 5867) conditions was quantified relative to baseline (117.0 ± 48.3) and NT sup (207.3 ± 62.5) conditions and found to be significant ( Figure 3 , red box, bar graph, NT sup vs hY3 sup, p=0.002, n=4). The significant shift in S100A4 expression unique to fibroblasts co-cultured with hY3 sup (1316 ± 469) was similarly quantified versus the other three conditions ( Figure 3 , green box, bar graph plot, hY3 sup vs TGF-β1, p=0.030, n=4). Neither NT sup (122.8 ± 30.6) nor TGF-β1 (159.2 ± 56.9) resulted in any significant increase in S100A4 expression relative to untreated fibroblasts (64.8 ± 22.2) ( Figure 3 , green box, bar graph plot, n=4).

Since previous scRNA-seq of fetal heart(s) dying with cardiac-NL identified increased expression of type I IFN-stimulated genes (ISGs) in several cell types, including fibroblasts, (26, 29), we added human IFNα directly to fetal cardiac fibroblasts and again stained for α-SMA and S100A4 via indirect immunofluorescence. In contrast to the direct addition of TGF-β1, IFNα induced dual differentiation of the fibroblasts, with expression of both α-SMA and S100A4, paralleling the results obtained with hY3 sup ( Figures 4E, F ). Again, TGF-β1 induced myofibroblast transdifferentiation only ( Figure 4B ). The significant mean expression of α-SMA and S100A4 in response to IFNα (α-SMA: 16332 ± 9085; S100A4: 1698 ± 806.4) and hY3 sup (α-SMA: 17101 ± 4824; S100A4: 1939 ± 594.6), as assayed via flow cytometry, were quantified (n=3) compared to untreated fibroblasts (α-SMA: 117.0 ± 48.3; S100A4: 64.8 ± 22.2) ( Figure 4 , α-SMA, red box, bar graph; S100A4, green box, bar graph). Moreover, dual differentiation staining after treatment with hY3 sup and IFNα was completely abolished with a neutralizing IFNα antibody ( Figures 4G, H ), supporting that type I IFN is released by macrophages downstream of the uptake of SSA/Ro immune complexes. In contrast, the α-SMA expression induced by treatment with TGF-β1 was unaffected by neutralizing antibodies to IFNα ( Figure 4D ).

The next set of experiments addressed the cytokine expression profile of fibroblasts stimulated with hY3 sup or TGF-β1 directly, as agents that selectively induce S100A4⁺ dominant and α-SMA⁺ fibroblasts, respectively. After 24 hours in culture, fibroblasts exposed to hY3 sup significantly upregulated (mRNA) expression of the pro-angiogenic cytokine IL6 compared to treatment with TGF-β1 (5.7 vs. 1.5, p=0.0079, n=5) ( Figure 5A ). Likewise, expression of MMP13, a matrix metalloproteinase involved in ECM remodeling (crucial for subsequent neovascularization) and cell motility (14, 30, 31), was significantly increased after exposure to hY3 sup (hY3 sup vs. TGF-β1: 7.5 vs. 0.8, p=0.0079, n=5) ( Figure 5B ). Expression of IL10, an anti-inflammatory pro-angiogenic cytokine, was also upregulated in fibroblasts co-cultured with hY3 sup but not in fibroblasts treated with TGF-β1 (hY3 sup vs. TGF-β1: 5.1 vs. 0.96, p=0.032, n=3) ( Figure 5C ). In contrast, expression indicative of collagen/ECM deposition (Col1α1) was upregulated by TGF-β1 compared to hY3 sup (4.8 vs. 1.8, p=0.028, n=4) ( Figure 5D ). Notably, the qPCR data harmonized with bulk RNA sequencing conducted on fibroblasts treated with the same conditions (NT sup, hY3 sup, TGF-β1). Specifically, fibroblasts treated with hY3 sup uniquely upregulated ECM-degrading matrix metalloproteinases, including MMP13 ( Figure 5E ), and pro-inflammatory, pro-angiogenic cytokines, including IL6, as well as angiogenic factors previously reported in the S100A4⁺ cardiac fibroblast transcriptome (21). When flow-sorted cardiac fibroblasts from 3 CHB hearts and 3 healthy hearts were similarly sequenced, the same pattern of upregulated MMPs and pro-angiogenic cytokines was observed in CHB fibroblasts relative to the healthy control ( Figure 5F ). Transcripts common among qPCR and RNAseq experiments (MMP13, IL6, IL10) are bolded.

Given the upregulation of ECM-degrading transcripts in CHB fibroblasts and those exposed to hY3 sup, further phenotypic distinction between the fibroblast populations was generated using the SirCol soluble collagen assay. In considering that there might be functional differences between α-SMA⁺ fibroblasts that secrete collagen and S100A4⁺ fibroblasts, we tested whether the direct addition of recombinant human S100A4 to cardiac fibroblasts would minimize the level of soluble collagen in the cell medium. In order to compensate for variable basal production of collagen in cardiac fibroblasts, fibroblasts were treated with collagen (12 µg/mL) to ensure a positive control. Incubation with S100A4 for 30 hours resulted in marked collagen degradation (Collagen spike vs. S100A4+spike: 14.1 vs. 0.6 µg/mL, p=0.026, n=5) ( Figure 5G ). Fibroblasts treated with hY3 sup similarly exhibited collagenase activity, to a lesser degree (Collagen spike vs. hY3 sup+spike: 10.4 vs. 4.2 µg/mL, p=0.046, n=5) ( Figure 5H ).

Given the distinct effects of hY3 sup, IFNα and TGF-β1 on fibroblast expression and behavior, the next set of experiments addressed the secretion of S100A4 by the stimulated macrophage supernatants. Evaluation of THP-1 supernatants for S100A4 via enzyme-linked immunosorbent assay highlighted the secreted protein’s presence in hY3-stimulated supernatants over DOTAP-transfected (13 vs. 3 ng/mL, n=4, p<0.01) ( Figure 6A ). Similarly, THP-1 macrophages treated with hY3, but not vehicle control, highly expressed S100A4 ( Figures 6B, C ). An indirect approach to evaluate S100A4 in fibroblast cultures involved a ‘two-tier’ double round culture condition. Briefly, at stage one, untreated cardiac fibroblasts were exposed to no addition, NT sup, hY3 sup, or recombinant S100A4 (added directly) ( Figures 6D–G ). At the second stage, untreated fibroblasts received the previous stage’s fibroblast supernatants; whereby conditioned media involving exposure to hY3 sup or recombinant S100A4, resulted in the untreated fibroblasts themselves expressing S100A4 and α-SMA, evidencing the feedforward amplification of the S100A4 signal in vitro ( Figures 6J, K ). No 'conditioning' related to basal fibroblasts and those treated with NT sup was observed ( Figures 6D, E, H, I ).

Clinical translation of the newly-identified S100A4⁺ fibroblast phenotype was actualized with ELISA measurement of 96 umbilical cord samples from anti-SSA/Ro-exposed neonates (healthy, n=47 (49%); cardiac-NL, n=49 (51%)). Soluble S100A4 was significantly elevated in the cord bloods of neonates with cardiac-NL compared to those with no evidence of injury (61.6 ng/mL vs. 12.0 ng/mL respectively, p<0.0001) ( Figure 7 ). Cord blood levels of S100A4 tracked with a previously established severity index of cardiac-NL disease (32). Specifically, 12 of 14 (86%) neonates with the most severe scores—advanced AV block and/or at least one extranodal manifestation—have the highest serum levels of S100A4 (>86 ng/mL); the lowest of the 14 being a milder case (endocardial fibroelastosis with no block) ( Figure 7 ). Remarkably, for 3 twin pairs (all with discordant disease), the level of S100A4 was consistently higher in the affected twin compared to the unaffected twin. Treatments given to the mothers of affected and unaffected neonates at any time during the pregnancy are reflected in the data symbols. For the affected neonates, the most common treatment was dexamethasone, given to 31 of 49 (63%) mothers. There was no significant relationship between cumulative dose of dexamethasone and the level of S100A4 ( Supplementary Figure 2 ). Another secreted protein, Frizzled-like Receptor Protein-1 (sFRP-1), which acts as an antagonist of the Wnt/β-catenin signaling pathway, was evaluated with the expectation that it would track with severity (33, 34). In contrast to S100A4, sFRP-1 showed no association with degree of injury and the levels were equal to those seen in the healthy neonates ( Supplementary Figure 3 ).