T cell-derived IFNγ instructs ECM crosslinking by cardiac fibroblasts through LOXL3 in experimental cardiometabolic HFpEF

27

28 Heart failure with preserved ejection fraction (HFpEF) is a major clinical challenge characterized 29 by diastolic dysfunction. Left ventricular stiffening and inflammation are hallmarks of HFpEF, yet 30 the contribution of extracellular matrix ( ECM) stiffness and the immune–stromal mechanisms 31 driving ECM stiffening in cardiometabolic HFpEF remain poorly understood. 32

33 We used the murine “2-hit model” of cardiometabolic HFpEF, in which the combination of high fat 34 diet and hypertension induced by L-NAME causes diastolic dysfunction. We evaluated diastolic 35 function by echocardiography and ECM mechanics by uniaxial tensile testing of decellularized 36 cardiac tissue. Functional in vivo studies included genetic depletion of T cells, interferon- (IFNγ) 37 knockout mice, and pharmacological lysyl oxidase inhibition. We combined co-cultures of CD4+ T 38 cells and cardiac fibroblasts (CFB) with mechanical testing of cardiac ECM and molecular biology 39 to elucidate cellular and molecular mechanisms. 40

41 Left ventricular ECM stiffness strongly correlated with impaired diastolic function in experimental 42 cardiometabolic HFpEF. Cardiac CD4 ⁺ T cell infiltration was required for ECM stiffening and 43 upregulation of lysyl oxidase enzymes in CFB. CD4+ T cell-derived IFNγ was both necessary and 44 sufficient to induce LOXL3 in CFB, which increased ECM stiffness in vitro. Mechanistically, IFNγ 45 signaling activated hypoxia -inducible factor-1α (HIF1α) in CFB, driving LOXL3 expression and 46 subsequent collagen crosslinking. Genetic or pharmacologic disruption of this IFN γ–HIF1α–47 LOXL3 axis in vivo attenuated adverse ECM remodeling and improved diastolic function. 48

49 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint CD4⁺ T cells promote pathological ECM stiffening in cardiometabolic HFpEF through IFN γ-50 mediated, LOXL3-dependent ECM crosslinking by CFB. Targeting this immune–stromal pathway 51 may offer a novel therapeutic strategy for HFpEF. 52 53

54 Extracellular matrix crosslinking, Immune-stromal, cardioimmunology, cardiac mechanics 55 56 Non-standard Abbreviations and Acronyms 57 BAPN ß-aminopropionitrile 58 CFB Cardiac fibroblast 59 Echi Echinomycin 60 ECM Extracellular matrix 61 E/A Early/atrial mitral valve inflow velocity 62 HFD High fat diet 63 HFpEF Heart failure with preserved ejection fraction 64 HIF1α Hypoxia-inducible factor-1α 65 IFNγ Interferon-γ 66 LOXL3 Lysyl oxidase-like 3 67 LV Left ventricle 68 L-NAME L-nitro-arginine-methyl ester 69 STD Standard diet 70 TGFß Transforming growth factor ß 71 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint

72 Heart failure with preserved ejection fraction (HFpEF) accounts for more than 50% of all 73 heart failure cases and is strongly associated with cardiometabolic comorbidities such as obesity, 74 hypertension, and diabetes (1). Despite its growing prevalence, the pathophysiological 75 mechanisms underlying HFpEF remain incompletely understood, and effective therapies are 76 lacking. A defining feature of HFpEF is diastolic dysfunction, which has been linked to myocardial 77 extracellular matrix (ECM) remodeling and increased tissue stiffness (2). However, the cellular 78 and molecular drivers of this maladaptive mechanical remodeling in the context of cardiometabolic 79 disease remain elusive. 80 Emerging evidence suggests that chronic low -grade inflammation contributes to HFpEF 81 pathogenesis (3,4), yet the specific immune cell subsets and their interactions with cardiac 82 stromal cells have not been fully delineated. CD4⁺ T cells, which are enriched in the myocardium 83 during cardiometabolic stress induced by obesity and hypertension (5), have been implicated in 84 fibrotic remodeling through crosstalk with cardiac fibroblasts (CFB) in other etiologies of HF (6,7). 85 CFB are the primary cell type maintaining and remodeling the ECM . They e xhibit profound, 86 dynamic phenotypic changes in response to mechanical and biochemical cues, including 87 upregulation of enzymes that mediate collagen crosslinking and tissue stiffening. Among these, 88 lysyl oxidase family members have emerged as critical regulators of ECM mechanics through 89 collagen crosslinking , yet their role in cardiometabolic HFpEF has not been defined (8–10). 90 Whether T cell interactions with CFB increase the expression of ECM crosslinking enzymes and 91 thereby promote myocardial stiffening is unknown. 92 Here, using a murine model of cardiometabolic HFpEF, we measured cardiac stiffening 93 and ECM mechanical remodeling using uniaxial tensile testing and performed a combination of 94 immunological, pharmacological and physiological approaches . We demonstrate that elevated 95 left ventricular (LV) ECM stiffness in cardiometabolic HFpEF is driven by CD4+ T cells through 96 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint induction of lysyl oxidase family member lysyl oxidase -like 3 (LOXL3) in CFB . We find that 97 interferon-γ ( IFNγ) activates hypoxia-inducible factor -1α ( HIF1α) signaling and is c ritical for 98 LOXL3 mediated ECM crosslinking by CFB. Genetic and pharmacologic perturbation of this 99 pathway in vivo attenuates ECM stiffening and improves diastolic function. 100

101 Animal models 102 Male w ild-type (WT) (C57Bl/6J, 000664), Tcra-/- (B6.129S2-Tcratm1Mom/J, 002116) or Ifng-/- 103 (B6.129S7-Ifngtm1Ts/J, 002287, all from Jackson Labs ) mice were 8 -12 weeks old at start of 104 experiments. Mice were given a high fat diet (HFD) with 60% of caloric intake from lard (Research 105 Diets Inc. D12492) and 0.5 g/L of the hypertension inducing agent L-nitro-arginine-methyl-esther 106 (L-NAME, SigmaAldrich N5751) in drinking water for 3 weeks or 5 weeks as indicated. Control 107 groups received standard chow and drinking water (STD). A subpopulation of WT mice received 108 daily i ntra-peritoneal injections of the pan -lysyl oxidase inhibitor β-aminopropionitrile (BAPN , 109 SigmaAldrich A3134, dissolved in PBS) of 100 mg/kg body weight from week 3-5 of HFD+L-NAME 110 treatment. Other mice were subjected to intra -peritoneal injections of recombinant IFN γ (25 111 kU/injection in 100 µL PBS, Peprotech, 315-05). Control mice for BAPN and IFN γ treated mice 112 received PBS injections of equivalent volume. Mice had ad libitum access to food and water and 113 were kept at 12h day-night cycle. All animal experiments were approved by the local authorities 114 and carried out in compliance with Institutional Animal Care and Use Committee ( IACUC) 115 requirements. 116 Echocardiography 117 Systolic and diastolic cardiac function were assessed using transthoracic echocardiography in 118 anesthetized mice as previously described (7). Mice were kept on a heated stage in supine 119 position, with heart and respiratory rates continuously monitored via stage electrodes. Heart rate 120 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint was kept between 450 and 550 bpm. Depilatory cream (Nair) was used to remove fur on the 121 chest, and ultrasonic coupling gel was applied onto the chest for imaging with a 22-55 MHz 122 echocardiography transducer (MS550D; Vevo 2100, FUJIFILM VisualSonics). Once the LV was 123 clearly visualized in short axis view, LV end‐systolic and end‐diastolic dimensions (M‐mode) were 124 measured, and the LV ejection fraction was calculated. LV geometry was assessed using LV 125 weights, LV anterior and posterior wall thickness, and end-diastolic volume. Parameters for 126 diastolic function, including the ratio of early-to-late mitral valve inflow velocity (E/A) were derived 127 from pulsed wave Doppler of transmitral flow in an apical four chamber view. Parameters of 128 cardiac systolic and diastolic function were measured by averaging of values obtained from 8 129 cardiac cycles. 130 Cardiac tissue decellularization and mechanical testing 131 Transverse sections of the LV of 1-2 mm thickness were incubated in decellularization buffer (1% 132 w/v sodium-dodecyl sulfate, 1% v/v Triton-X100 in PBS) for 3 -5 days at room temperature on a 133 rotator until the tissue was translucent. After washing in PBS, the decellularized tissue was 134 submerged in PBS with 0.05% w/v sodium azide for storage. 135 For mechanical testing, a 1 -3 mm long strip corresponding to the LV free wall was 136 dissected from the decellularized tissue and mounted on a custom uniaxial tensile testing system 137 (11). On one end, it was glued to a static lever, while the other end was fixed to a Dual Mode 138 Lever System with a 1 N load cell (AuroraScientic #6350*358). For control and data digitization, 139 the system was coupled to a digital input/output instrument (National Instruments USB-6221). 140 Using an in-house LabVIEW script (LabVIEW 2011, National Instruments) uniaxial tension was 141 applied in displacement control mode. Output was recorded at 30 samples/s. Before starting a 142 strain protocol, the membrane was brought into a straight planar position with 5 mN pre -load 143 applied. Then, a cyclic strain protocol was executed and at least 20 cycles were recorded. During 144 measurements, the tissue was submerged in saline. Analysis of the stress -strain data was 145 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint performed using a custom MatLab script (MatLab 202 5b). Briefly, the recorded force and 146 displacement were converted into stress and strain, respectively. Then, a linear relation was 147 computed for the stress-strain relation during tissue stretch (not relaxation) at 10-15% strain. The 148 slope of the linear relation is referred to as Elastic modulus. 149 To assess the effects of the CFB secretome on ECM mechanical properties, right 150 ventricular tissue was harvested from 6–12-week-old C57Bl/6J mice and decellularized as above. 151 Decellularized tissues preparations (dECM) were subjected to uniaxial tensile testing, then 152 sterilized using ethanol, and incubated in conditioned media from CFB for 16-20h before repeating 153 the tensile test. CFB conditioned media was supplemented with the lysyl oxidase inhibitor BAPN 154 (500 µM, SigmaAldrich A3134) where indicated. Data was analyzed as above and displayed as 155 fold-change of the Elastic modulus between before and after the incubation. 156 Picrosirius Red staining 157 For assessment of LV fibrosis, one third of the LV was fixed in 10% formalin, embedded in paraffin, 158 sectioned (5 µm thickness) and mounted onto microscopy slides. After deparaffinization , the 159 slides were stained in Picrosirius Red staining solution ( 1 g/L Direct Red 80, SigmaAldrich 160 365548) for 60 min followed by two washing steps in acidified water. The slides were then 161 dehydrated and mounted using non-aqueous mounting medium (Depex, EMS 13514). Per heart, 162 five representative fields of view not containing vessels w ere imaged, and the collagen area 163 fraction was quantified using FIJI (12). Each data point represents the average of five images 164 from the same section. 165 ECM quantification 166 To quantitatively assess ECM content in LV tissue, insoluble matrix proteins were enriched from 167 ~20 mg of frozen cardiac tissue using the Compartment Protein Extraction Kit (EMD Millipore 168 2145) according to the manufacturer’s protocol. The remaining insoluble pellet was washed twice 169 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint in PBS containing protease inhibitors. Pellets were dried overnight at room temperature and dry 170 pellet weight was normalized to the weight of the input tissue. 171 Heart digestion for flow cytometric analysis 172 LV myocardium was harvested from anesthetized mice after terminal blood draw, finely minced 173 using a razor blade and digested using 0.895 mg/mL of Collagenase Type-II (Gibco 17101015) in 174 phosphate buffered saline (PBS) to achieve a single cell suspension. After 20 min, the tissue was 175 dissociated mechanically using a 19 Gauge stainless steel cannula. After a total digestion time of 176 30 min, the suspension was filtered through a 100 µm cell strainer to remove remaining tissue 177 chunks. The resulting single cell suspension was then stained for flow cytometry using fluorophore 178 coupled antibodies (Table S1) at the indicated dilutions in FACS buffer (PBS +2% heat-inactivated 179 fetal bovine serum) at 4 C in the dark. After 20 min, the cells were washed in FACS buffer, and 50 180 μL/sample Precision Count Beads (BioLegend 424902) were added to quantify absolute cell 181 numbers. Spectral flow cytometry was performed on the Cytek® Aurora flow cytometer. After 182 spectral unmixing, flow cytometry data was analyzed using FloJo (BD BioScience v10.10.0). 183 Human bulk RNA sequencing analysis 184 Hahn et al performed bulk RNA sequencing on LV septum biopsies from control patients and 185 patients with HFpEF (13). We downloaded the list of differentially expressed genes and extracted 186 those with significant differences (adjusted p<0.05) between healthy controls and HFpEF patients. 187 Genes with significantly higher expression in HFpEF patients were subjected to GO term analysis. 188 Then, raw reads per patient were downloaded and data from control patients as well as patients 189 in HFpEF was extracted. Using an in-house Python script generated with support from Microsoft 190 Co-pilot, we computed Pearson’s correlation coefficient (r) and p-value for the correlation between 191 CD4 and all collagen or lysyl oxidase encoding genes. 192 Single cell RNA sequencing analysis 193 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint We downloaded raw single cell RNA sequencing data from a publicly available source (14). Count 194 data was processed sample wise with application of the following filter cut -offs: >200 Features, 195 <25% mitochondrial genes, 500 RNA counts using Seurat version 196 5.3.1. Data was log -normalized, and s amples were clustered by principal component analysis 197 using 30 dimensions followed by FindCluster function within the Seurat package with a resolution 198 of 0.7. Cell type markers were calculated using the FindMarkers function with default parameters 199 (Wilcoxon test) in Seurat and cell types were manually annotated based on known canonical 200 markers. CFB were subsetted and re -clustered by principal component analysis using 30 201 dimensions and the FindCluster function with a resolution of 0.5. Percent contribution to the CFB 202 population for each cluster was calculated based on the number of single cells in each cluster in 203 control and HFpEF groups, respectively. Subcluster markers were calculated using the 204 FindMarkers function with default parameters. Significant marker genes of cluster 0 (adjusted p-205 value<0.05) were subjected to GO analysis using Panther. Lastly, a lysyl oxidase score was 206 calculated as the sum of the expression of all lysyl oxidase family members ( Lox, Loxl1, Loxl2, 207 Loxl3, Loxl4) per cell. 208 LOXL3 ELISA 209 For protein analyses, snap-frozen LV tissue samples were thawed on ice and mechanically 210 disrupted in 100 μL RIPA buffer (ThermoFisher J62524.AE) containing protease (ThermoFisher 211 A32953) and phosphatase inhibitors (ThermoFisher A32957). The lysates were cleared from 212 debris by centrifugation (10,000 rpm, 5 min, 4 C) and total protein content was determined using 213 a bicinchoninic acid assay (ThermoFisher 23225) according to the manufacturer’s instructions. 214 Lysates were diluted to 10 μg/mL and applied to enzyme-linked immunosorbent assay targeting 215 LOXL3 (LS Bioscience LS-F14541-1). 216 Splenic CD4+ T cell isolation, culture and polarization 217 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint Splenic CD4+ T cells were isolated by positive selection using CD4+ magnetic beads ( Miltenyi 218 Biotec 130 -117-043) from spleens of 7–12-week-old C57Bl/6J mice according to previously 219 established protocols (15,16). As basal T cell cul ture media we used RPMI supplemented with 220 10% (v/v) heat-inactivated fetal bovine serum ( Gemini Bio S11550), 1.2 mM sodium pyruvate 221 (Gibco, 11360-070), 0.1% (w/v) sodium bicarbonate (Gibco 25080-094), 1x GlutaMAX ( Gibco 222 A12860-01), 1x Pencillin/Streptomycin ( Gibco 15070-063) and 0.0005 (v/v) β-mercaptoethanol 223 (Sigma M3148). For in vitro experiments, cells were cultured at 2 mi llion cells per mL in the 224 presence of plate-bound αCD3 (coating with 2.5 μg/mL, BioLegend 100253) and soluble αCD28 225 (1 μ g/mL, BioLegend 102102) and IL -2 (25 U/mL, Peprotech 212 -12) for 3 days at 37°C to 226 generate activated CD4+ T cell blasts. After 3 days, cells were expanded at 1:2 in media containing 227 IL-2 (25 U/mL). After 24h, cells were centrifuged at 10,000 g for 5 min at room temperature and 228 supernatants were collected under sterile conditions. These conditioned media were used to treat 229 CFB as described below. 230 CFB isolation, culture and treatments 231 To isolate CFB, LV of 4–8-week-old WT C57Bl/6J mice were minced and digested using 2 mg/mL 232 Collagenase Type I (Gibco 17100017) for 25 min (including mechanical disruption by cannulation 233 at 20 min digestion time). The suspension was filtered through a 100 µm cell strainer and plated 234 onto gelatine-coated tissue culture plates. At 90% confluency, CFB were detached using Trypsin-235 EDTA (Gibco 25300054) and expanded. Experiments were performed at passage 2 after isolation. 236 For RNA isolation, CFB were plated at 50,000 cells/well in 12 well plates. Basic CFB culture media 237 was DMEM supplemented with 10% (v/v) fetal bovine serum (Gemini Bio S11550), 1% (v/) Insulin-238 Transferrin-Selenium (Gibco 41400-045) and 1% (v/v) Penicillin/Streptomycin (Gibco 15070-063). 239 For immunofluorescence, CFB were plated at 5,000 cells/well onto gelatin -coated glass -240 coverslips in 24 well plates. 24h after plating, cells were starved in culture media containing 2% 241 fetal bovine serum (FBS) for another 24h. Then, the indicated treatments were applied as follows: 242 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint conditioned media from activated CD4+ T cell blasts (1:4 in CFB culture media), recombinant 243 interferon-γ (IFNγ, 100 U/mL, Peprotech, 315-05), recombinant IL-2 (25 U/mL), recombinant IL-4 244 (500 ng/mL, Peprotech 214-14), recombinant transforming growth factor -β (TGFβ, 100 ng/mL, 245 Peprotech 100-21). In selected experiments, CFB were additionally treated with the following 246 inhibitors: αIFNγ (10 ug/mL, BioLegend 505834), Echinomycin (HIF-1α inhibitor, 5 nM, Tocris 247 Bioscience 5520). 248 qPCR 249 RNA from cultured CFB was isolated using the RNeasy kit (Qiagen 74106) according to the 250 manufacturer’s instructions and RNA yield was assessed spectrometrically. Adjusted amounts of 251 RNA were reverse transcribed into complementary DNA (cDNA) using High -Capacity cDNA kit 252 (Fisher Scientific 43-874-06). A total of 9 ng/reaction was then subjected to quantitative real-time 253 polymerase chain reaction using the primer sequences in Table S2. Expression of target genes 254 was normalized by the expression of the reference gene Rpl19 and is expressed as fold change 255 with respect to the untreated control condition. 256 Collagen contraction assay 257 To assess the ir contractile properties, primary murine CFB were cultured in a neutral Type1 258 Collagen solution (1 mg/mL, Advanced Biomatrix 5074) at 150,000 cells in 500 µL per well of a 259 24 well plate. After an initial solidification period of 1h, 600 µL of CFB culture media with the 260 indicated treatments were added to each well followed by immediate release of the collagen 261 hydrogel from the edges of the well. Images of the collagen disks were taken after 24 h. Disk area 262 was quantified from these images and expressed as percentage of the area of the entire well. 263 Immunofluorescence 264 CFB cultured on glass coverslips and treated as indicated were chemically fixed using 4% 265 paraformaldehyde for 10 min. After permeabilization with 0.1% Triton-X100 in PBS (15 min) and 266 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint blocking in 5% normal goat serum in PBS (1h), primary antibodies were applied overnight at 4 C. 267 Appropriate secondary antibodies were applied for 1h at room temperature. Coverslips were 268 mounted onto microscopy slides using Permafluor mounting medium containing DAPI (Southern 269 Biotech 0100 -20) for nuclear counterstain. Cells were imaged within 48 h of staining. For 270 quantification of smooth muscle actin (SMA) and collagen-I (COL1A1), the mean fluorescence 271 intensity across the entire image was averaged. For quantification of nuclear HIF1 α, a region of 272 interest was defined based on the nuclear signal and the average intensity of the HIF1α signal in 273 this region was computed. Five representative images were acquired per coverslip, and each data 274 point represents the average of those. 275 Statistical analysis 276 Statistical analysis was performed using GraphPad Prism 10. All data are presented as 277 mean±standard error of the mean. Individual data points represent data from individual mice (for 278 in vivo studies) or independent biological replicates (for in vitro studies). Normal distribution was 279 assessed using Shapiro-Wilk test. Two group comparisons were done by unpaired Student’s t-280 test for normally distributed data or Mann-Whitney test for non-normally distributed data. Multiple 281 group comparisons were done using 1way or 2way ANOVA with Tukey’s multiple comparison test 282 as appropriate. Differences were considered statistically significant at p<0.05. 283 284 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint

285 Mechanical ECM remodeling correlates with diastolic dysfunction in experimental 286 cardiometabolic HFpEF 287 We subjected 8-12-week-old WT mice to the combination of HFD and L-NAME for 5 weeks which 288 induced diastolic dysfunction , while ejection fraction was preserved (Figure 1A, Table S3). 289 Doppler echocardiography confirmed a higher ratio of early/late mitral valve inflow velocity which 290 is indicative of diastolic dysfunction , as expected based on recent publications (Figure 1B and 291 1C) (5,17). As diastolic function depends on the extensibility of the ECM of the LV, we next aimed 292 to assess the mechanical properties of the LV ECM. To this end, we prepared transverse sections 293 of the LV free wall and subjected them to decellularization. We then assessed the elastic modulus 294 as a measure of stiffness of the ECM (Figure 1D and 1E). The LV-ECM derived from the hearts 295 of mice treated with HFD+L-NAME was significantly higher than that of mice receiving STD chow 296 (Figure 1F), while we did not detect significant differences in the stiffness of the right ventricular 297 ECM between both groups (Fig S1A). Importantly, only the combination of HFD and L -NAME 298 treatment, but not either treatment alone, caused high LV-ECM stiffness after 5 weeks (Figure 299 1G). Importantly, the stiffness of the LV-ECM correlated significantly with the E/A ratio across STD 300 and H/L groups (Pearson’s R2: 0.512, p value: 0.009), highlighting the importance of mechanical 301 ECM remodeling for diastolic function (Figure 1H). Noteworthy, the higher stiffness of the LV-ECM 302 was not associated with significant collagen deposition that could be detected by picrosirius red 303 staining (Figure 1I). Similarly, we did not find significant differences in the abundance of insoluble 304 ECM proteins (expressed percent of insoluble ECM weight over tissue weight, Figure 1J). Flow 305 cytometry results showed no significant differences in total CD45-CD31-MESFK4+ CFB numbers 306 (Figure S1B and S1C). Together, these data demonstrate significant stiffening of the LV-ECM in 307 response to HFD/L-NAME treatment without increasing ECM deposition , and suggests that 308 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint mechanical remodeling of the ECM , rather than ECM abundance , is a critical determinant of 309 diastolic dysfunction in response to HFD/LNAME. 310 Mechanical ECM remodeling in cardiometabolic HFpEF is T cell-dependent 311 We previously demonstrated that HFD/L-NAME causes a cardiotropic response in inflammatory 312 CD4+ T cells and that T cell-deficient mice are protected from diastolic dysfunction in this model 313 (5). Thus, we investigated the temporal relationship between cardiac CD4 + T cell infiltration and 314 ECM stiffening. Using flow cytometry to identify CD4+ T cells in the LV of mice treated with STD 315 chow or HFD+L-NAME for 3 or 5 weeks (Figure S2), we found increased LV CD4+ T cell numbers 316 at 5 weeks compared to STD fed controls , while no significant differences were observed at 3 317 weeks. (Fig 2A). Remarkably, the time course of increase in LV-ECM stiffness correlated with 318 CD4+ T cell infiltration. Specifically, LV-ECM stiffness was significantly higher after 5 weeks of 319 HFD/L-NAME compared to STD mice, while no significant differences were identified at 3 weeks 320 of HFD/L-NAME, a time point at which CD4+ T cells are not yet significantly increased in the heart 321 (Figure 2B). To assess if the increase in CD4 + T cells in the hearts was causally related to the 322 stiffness of the ECM and diastolic function, we treated T cell deficient mice (Tcra-/-) alongside WT 323 mice with HFD/L-NAME or STD chow for 5 weeks. Tcra-/- mice did not show significant diastolic 324 dysfunction in response to HFD/L-NAME treatment (Figure 2C and 2D, Table S4). Strikingly, LV-325 ECM stiffness from HFD/L-NAME treated Tcra-/- mice did not increase compared to STD. In fact, 326 LV-ECM stiffness was significantly lower in HFD/L -LAME treated Tcra-/- mice compared to WT 327 mice (Figure 2E and 2 F). These data confirm the importance of CD4 + T cells for diastolic 328 dysfunction and support that CD4+ T cells are functionally involved in ECM stiffening in response 329 to HFD/L-NAME. 330 Given the important role of T cell s in murine cardiometabolic HFpEF, w e next asked if 331 similar features were present in human HFpEF patients using a publicly available bulk RNA 332 sequencing data set from human HFpEF patients and healthy control patients (13). We found that 333 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint terms related to (CD4 +) T cell activation and differentiation were significantly overrepresented 334 within genes upregulated in HFpEF patients (Figure 2G). We also found the cellular compartment 335 Collagen trimer to be overrepresented in HFpEF patients suggesting that collagen crosslinking 336 might be involved in HFpEF (Figure 2H). To examine a potential relation between the presence 337 of CD4+ T cells and collagen remodeling, we compared the correlation coefficient and p-value of 338 all collagen and lysyl oxidase genes, central to collagen crosslinking and ECM remodeling, with 339 CD4 gene expression, a proxy for the number of CD4+ T cells in the tissue. This analysis identified 340 the strongest and most significant correlation occurs between CD4 and the lysyl oxidase family 341 member LOXL3 (Pearson’s R: 0.63, p value: <0.0001, Figure 2I). The correlation between CD4 342 and LOXL3 was statistically significant across patient groups (Figure 2J). Thus, we hypothesized 343 that lysyl oxidation by LOXL3 might represent a mechanism by which the stiffness of existing ECM 344 is increased in the absence of additional ECM production. 345 CFB express higher levels of lysyl oxidases in cardiometabolic HFpEF 346 We next investigated the cellular source of lysyl oxidases in the myocardium by mining publicly 347 available single cell RNA sequencing data from mice receiving HFD/L-NAME compared to STD 348 chow (14). We identified all major cardiac non-myocyte cell types (Figure S3A and S3B) and found 349 fibroblasts as the dominant source of lysyl oxidases in the heart ( Figure 3A). Next, we 350 subclustered the isolated CFB population (Figure S3C) and found that Cluster 0, which made up 351 35% of the CFB population in control mice, expanded to 50% in HFpEF mice (Figure 3B and 3C). 352 To better understand the characteristics of this expanded cluster, we performed gene ontology 353 (GO) analysis of genes which were significantly enriched in this cluster ( i.e. its marker genes). 354 The term peptidyl lysine oxidation stood out by being 50-fold overrepresented within the cluster 0 355 marker genes (Figure 3D). Consequently, we found higher lysyl oxidase expression across the 356 entire CFB population in HFpEF mice compared to control mice (Figure 3E). Based on the findings 357 that lysyl oxidase family member LOXL3 correlated significantly with CD4 (Figure 2I and 2J) and 358 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint that T cells are causally in volved in the development of ECM stiffening (Figure 2E and 2F), we 359 next assessed the abundance of LOXL3 protein in LV tissue of mice after 3-5 weeks of HFD/L-360 NAME treatment. In line with the timeline of CD4 + T cell infiltration and ECM stiffening, the 361 abundance of LOXL3 was significantly higher after 5 weeks, but not 3 weeks of HFD/L -NAME 362 treatment compared to STD controls (Figure 3F). These data support that CFB are a main source 363 of Loxl3 in the onset of HFD/L-NAME-induced ECM remodeling when CD4+ T cells are present in 364 the LV and prompted us to further investigate the relevance of this axis in myocardial stiffening 365 and diastolic dysfunction. 366 CD4+ T cell-derived IFNγ induces LOXL3 expression in CFB through HIF1α which 367 causes ECM stiffening in vitro 368 To gain insight into the mechanism by which CD4 + T cells could stimulate Loxl3 expression in 369 CFB, we next investigated whether the secretome derived from CD4+ T cells was able to induce 370 Loxl3 expression in CFB in vitro. We treated primary murine CFBs with the secretome collected 371 from activated CD4+ T cell blasts (Figure 4A). After 24h of treatment, the expression of Loxl3 was 372 significantly higher in CFB treated with the CD4+ T cell secretome compared to non-treated control 373 CFB (NC, Figure 4B). Consequently, lysyl oxidase activity in the cu lture media from CFB was 374 significantly elevated after treatment with CD4+ T cell secretome . This demonstrates that, in 375 addition to transcriptional Loxl3 upregulation, higher levels of active LOXL3 are secreted by CFB 376 in response to CD4+ T cell secretome treatment (Figure 4C). We also found that the CD4+ T cell 377 secretome did not alter the protein levels of COL1A1 (Figure S4B and S4E), in line with the 378 absence of significant collagen deposition we observed in vivo . Thus, we next investigated 379 whether the CD4+ T cell secretome increases the ability of CFB to remodel existing collagen 380 instead of producing collagen de novo. We first measured the contraction of collagen by CFB 381 disks in vitro in the presence or absence of CD4 + T cell secretome using T GFβ treatment as a 382 positive control. Treatment of CFB with CD4 + secretome resulted in higher collagen gel 383 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint contraction by CFB compared to non-treated CFB, albeit not to the same extent as observed in 384 TGFβ treated CFB (Figure 4D). We also evaluated the expression levels of the contractile protein 385 αSMA in CFB and found that gene and protein expression were enhanced by TGFβ and not by 386 the activated CD4+ T cell secretome, in line with the stronger disc contraction observed by TGFβ 387 compared to the T cell secretome treatment (Figure S4A and S4 D). Next, w e collected the 388 conditioned media from CFB treated with CD4+ T cell secretome (containing active LOXL3) and 389 treated native cardiac dECM to determine its effect on ECM stiffness by uniaxial tensile testing 390 (Figure 4E). Conditioned media from CFB previously exposed to the CD4 + T cell secretome 391 resulted in increased stiffness of dECM, compared to dECM treated with control CFB conditioned 392 media (not pre-treated with CD4+ T cell secretome and therefore low in LOXL3). This effect was 393 completely abolished in the presence of the lysyl oxidase inhibitor BAPN (Figure 4E and 4F). 394 Taken together, our results demonstrate that a paracrine mediator derived from CD4 + T cells 395 induces the expression and secretion of LOXL3 in CFB which subsequently increases the 396 stiffness of native cardiac ECM. 397 To identify the CD4+ T cell derived signal that causes the upregulation of LOXL3 in CFB, 398 we treated CFB with cytokines which are well -known to be produced at high levels by CD4+ T 399 cells (activated by TCR stimulation with αCD3 and co-stimulation with αCD28 in culture but not 400 polarized towards a specific subset) in culture. TGFβ was used as a profibrotic signal for CFB and 401 positive control . We found that IFN γ induced Loxl3 expression, whereas IL -2 or IL -4 had no 402 significant effect (Figure 4G). Moreover, neutralizing IFNγ with a selective αIFNγ neutralizing 403 antibody in the CD4+ T cell secretome resulted in abrogation of Loxl3 expression by CFB (αIFNγ, 404 Figure 4H). The ECM stiffening ability of CFB-derived conditioned media was blunted when IFNγ 405 was neutralized during treatment with the CD4+ T cell secretome, whereas IFNγ treatment alone 406 reconstituted ECM stiffening (Figure 4I). Taken t ogether, th ese results demonstrate the 407 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint requirement of CD4+ T cell IFNγ for Loxl3 expression in CFB and subsequent ECM stiffening by 408 CFB in vitro. 409 To understand the mechanism by which IFNγ induces Loxl3 expression in CFB, we focused 410 on HIF1α signaling, as lysyl oxidases have been shown to be induced by hypoxia in cancer -411 associated fibroblasts (18) (Figure 4J). Immunocytochemistry revealed that the CD4 + T cell 412 secretome increased the nuclear abundance of HIF1 α in CFB. More over, HIF1 α nuclear 413 localization was dependent on IFNγ, as it was abrogated when IFNγ was neutralized in the 414 secretome (Figure 4K and 4L), demonstrating that CD4+ T cell derived IFNγ drives HIF1α nuclear 415 localization in CFB . To further test the requiremen t of HIF1 α in LoxL3 expression and ECM 416 stiffening, we performed similar studies in the presence of the HIF1α inhibitor Echinomycin (Echi). 417 Echi completely abrogated Loxl3 induction by the CD4+ T cell secretome (Figure 4M), as well as 418 ECM stiffening (Figure 4N). 419 Taken together, these data demonstrate that CD4+ T cell IFNγ is necessary for HIF1α nuclear 420 localization and subsequent Loxl3 expression in CFB, and that inhibition of HIF1α or IFNγ prevent 421 ECM stiffening induced by T cells. 422 IFNγ and lysyl oxidation are required for mechanical ECM remodeling and 423 diastolic dysfunction in cardiometabolic HFpEF 424 To assess the relevance of IFNγ-mediated HIF1α activation and subsequent LOXL3 upregulation 425 in vivo, WT mice were injected with recombinant IFN γ for 5 consecutive days ( Figure 5A). We 426 found higher levels of Hif1a and Loxl3 mRNA (Figure 5B and 5C) as well as LOXL3 protein (Figure 427 5D) in whole LV lysates of IFNγ-treated mice compared to PBS-treated controls. Consequently, 428 we asked if the absence of IFN γ would change the pathological outcome of HFD/L -NAME 429 treatment in vivo. We treated IFNγ-deficient (Ifng-/-) mice alongside WT mice with HFD/L-NAME 430 for 5 weeks (Figure 5E). Strikingly, the LV-ECM of Ifng-/- mice was significantly softer than that of 431 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint WT mice, as determined by lower elastic modulu s (Figure 5F). Moreover, Ifng-/- mice were 432 protected from diastolic dysfunction in response to HFD/L-NAME, supporting the requirement of 433 IFNγ for ECM stiffening and diastolic dysfunction in cardiometabolic HFpEF (Figure 5G). 434 To investigate the importance of lysyl oxid ase-mediated ECM stiffening for diastolic 435 dysfunction in response to HFD/L-NAME in vivo, we treated WT mice with BAPN, starting 3 weeks 436 after switching their diet to HFD/L-NAME (i.e. prior to CD4 + T cell infiltration ( Figure 2A), ECM 437 stiffening (Figure 2B) and cardiac LOXL3 upregulation (Figure 3F)) and continued until the end of 438 the experiment at 5 weeks of HFD/L-NAME (Figure 5H). Strikingly, BAPN treatment resulted in 439 significantly lower ECM stiffness in BAPN treated mice compared to PBS treated mice ( Figure 440 5I). Consequently, BAPN treated mice did not develop significant diastolic dysfunction (Figure 5J). 441 Taken together, we show that IFNγ is sufficient to induce cardiac LOXL3 expression in vivo. Both, 442 IFNγ and lysyl oxidation are required for cardiac stiffening and diastolic dysfunction in response 443 to HFD/L-NAME. 444 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint

445 Here, we investigated the role of mechanical ECM remodeling in the pathology of cardiometabolic 446 HFpEF and assessed the contribution of CD4 ⁺ T cells in steering mechanical ECM remodeling 447 through collagen crosslinking. We identified elevated ECM stiffness as a crucial contributor to 448 diastolic dysfunction and describe a novel mechanism of T cell IFNγ instructed upregulation of 449 LOXL3 in CFB that drives this stiffening. Cardiac tissue fibrosis , defined as excessive 450 accumulation of ECM, is associated with the majority of cardiac diseases. In HFpEF patients the 451 occurrence of fibrosis appears to be variable. While a number of studies show mild but significant 452 fibrosis in patients as well as animal models (17,19,20), other studies suggest that fibrosis plays 453 a minor role in regard to diastolic function. In fact, HFpEF patients with the best-preserved ejection 454 fractions may be the least likely to show significant fibrosis (21). Our results d emonstrate that 455 enhanced crosslinking of ECM by lysyl oxidases results in ECM stiffening in experimental 456 cardiometabolic HFpEF in the absence of histologically detectable fibrosis . Pharmacological 457 inhibition of lysyl oxidases prevented both ECM stiffening and diastolic dysfunction. This supports 458 that enhanced ECM crosslinking is sufficient to cause diastolic dysfunction and does not require 459 de novo ECM accumulation. This is highlighted by our functional in vitro studies, in which LOXL3 460 enrichment in culture media was solely responsible for the stiffening of native cardiac ECM. Using 461 mass spectrometry based in -depth analysis of the cardiac EC M, we previously identified sub-462 histological increases of ECM deposition in the LV of mice with hypertensive HFpEF , a 463 phenomenon termed hidden fibrosis (22). Our data in cardiometabolic HFpEF supports the 464

that ECM remodeling does not need to be visible as collagen deposition to be 465 functionally relevant. In fact, our data collected after only 5 weeks of HFD/L-NAME treatment 466 suggests that sub-histological mechanical remodeling may precede the development of 467 prominent fibrosis characterized by collagen deposition. 468 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint We demonstrate the causal involvement of CD4 + T cells in mechanical ECM remodeling 469 in HFD/L-NAME-induced cardiometabolic HFpEF. This is in line with our previous work which 470 identified systemic CD4 + T cell activation and CD4 + T cell cardiotropism as crucial features of 471 cardiometabolic HFpEF (5). Sequencing data from human HFpEF patients suggests a strong 472 correlation between the presence of CD4 + cells and the expression of LOXL3. While we did not 473 observe a significant difference in LOXL3 expression between control and HFpEF patients, the 474 correlation between CD4 and LOXL3 was statistically significant across groups suggesting that 475 the abundance of CD4 + T cells in patient tissue relates to the extent of collagen crosslinking 476 through LOXL3. The lysyl oxidase LOXL2 has emerged as a promising therapeutic target in HF 477 patients with hypertension and aortic stenosis (24). Our work suggests that targeting LOXL3 may 478 serve as a candidate novel therapeutic strategy in cardiometabolic HFpEF patients. 479 Mechanistically, we discovered a novel, contact-independent communication pathway 480 between cardiac infiltrated CD4+ T cells and CFB, mediated by the canonical Th1 cytokine IFNγ. 481 Our in vitro studies demonstrate that T cell derived IFNγ is required for CFB expression of LOXL3, 482 and our in vivo studies using recombinant IFNγ and Ifng-/- mice support this and its importance for 483 diastolic function. However, in vivo, T cells are not the exclusive source of IFNγ in the heart as it 484 is also produced by natural killer (NK) and NKT cells. NKT cells may be of particular interest in 485 the context of hyperlipidemia as they are activated by lipid antigens presented by the non-classical 486 MHC-I-like molecule CD1d (23). Whether IFN γ-producing NK(T) cells are also enriched in the 487 myocardium and how they may contribute to T cell -CFB communication and ECM crosslinking 488 through LOXL3 in this model will be the subject of future studies. 489 The effects IFNγ on CFB are manifold. Our data paint a complex picture, in which on the 490 one hand, IFNγ upregulates Loxl3 similarly to TGFβ, and on the other hand does not enhance the 491 fibrotic marker genes Acta2 and Col1a1. Even though seemingly opposing , b oth effects are 492 supported by previous studies. Our lab demonstrated that the integrin α4-dependent interaction 493 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint between IFNγ+ Th1 cells and CFB leads to TGF β production and subsequent transformation of 494 CFB into fibrotic myofibroblasts in pressure overload -induced HF (7). Others showed that IFNγ 495 counteracts the effects of TGFß in CFB and favors an inflammatory phenotype over a fibrotic 496 phenotype (23). Altogether, this highlights the complexity of the CFB phenotype and the need for 497 delineating CFB responses to stimulation in a disease-specific context. CFB heterogeneity and 498 disease-specificity have been explored using single cell sequencing and computational 499 approaches to compare their transcriptomic signatures in multiple murine HF models. In line with 500 our data showing the requirement for HIF1 α for IFN γ-driven LOXL3 expression, s ingle cell 501 sequencing identified a hypoxia signaling signature as specific feature of HFpEF fibroblast s 502 compared to those from HFrEF models like transverse aortic constriction or long term remodeling 503 after myocardial infarction (14). 504 While our work demonstrates that mechanical ECM remodeling in cardiometabolic HFpEF 505 is triggered by CD4+ T cell derived IFNγ and executed by CFB releasing LOXL3, there are some 506

that need to be acknowledged. Our in vivo studies have been performed in male mice 507 as female mice are more resistant to diastolic dysfunction in the 2 -hit model of cardiometabolic 508 HFpEF even up to 15 weeks of HFD/L-NAME (25). However, the IFN γ-LOXL3 axis we report 509 herein for CD4+ T cell-CFB crosstalk may show sex specific differences. Further, we focused our 510 current study on the HFD/L-NAME HFpEF model, so the relevance of our findings to other models 511 of HFpEF , and therefore potentially to other patient subgroups , remains untested . In a 512 hypertensive HFpEF model using DOCA-salt treatment and uninephrectomy, we also found that 513 ECM enrichment was not detectable by histology (22). Whether such small increases of ECM 514 deposition alone are sufficient to cause diastolic dysfunction or whether additional crosslinking 515 such as through the mechanisms described here are at play requires further studies. The IFNγ-516 mediated transcriptional upregulation of LOXL3 we describe here may involve changes in 517 chromatin accessibility, consistent with the mechanism we described in the hypertensive HFpEF 518 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint model (22). In addition to T cells, other immune cell types including macrophages are important 519 contributors to pathology in the 2-hit model of cardiometabolic HFpEF (26), and it is possible that 520 T cell-macrophage interactions also take place. Lastly, beyond the mechanical properties of the 521 ECM, cardiomyocyte relaxation is a critical contributor to diastolic function. Our previous study 522 suggests that impaired cardiomyocyte relaxation is T cell dependent (5). However, how T cells, 523 CFB and potentially macrophages interact with cardiomyocytes in the context of cardiometabolic 524 HFpEF and contribute to ECM mechanical remodeling remains to be explored. 525 Taken together, our work identifies a previously unrecognized immune -stromal axis by 526 which T cell inflammation directly affects mechanical ECM remodeling and diastolic dysfunction 527 in cardiometabolic HFpEF. We demonstrate a novel mechanism of IFN γ-dependent 528 communication between cardiac infiltrated CD4 + T cells and CFB and establish a novel 529 mechanistic link between adaptive immunity and cardiac mechanics in HFpEF . The resulting 530 increase in LOXL3 expression and collagen crosslinking is a crucial driver of diastolic dysfunction 531 in cardiometabolic HFpEF. This novel immune-stromal interaction reveals potential new targets 532 to mitigate cardiac stiffening and diastolic dysfunction in cardiometabolic HFpEF. 533 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint Novelty and Significance 534 What is known? 535 - Stiffness of the extracellular matrix is an important determinant of diastolic function 536 - Cardiac fibroblasts are the major cell type responsible for cardiac extracellular matrix 537 homeostasis and remodeling, and are sensitive to cytokine stimulation 538 - CD4+ T cells infiltrate the myocardium and are required for diastolic dysfunction in 539 cardiometabolic HFpEF 540 What new information does this article contribute? 541 - Enhanced crosslinking by lysyl oxidation causes diastolic dysfunction in the absence of 542 histological fibrosis in cardiometabolic HFpEF 543 - CD4+ T cell-derived IFNγ drives the expression of the lysyl oxidase LOXL3 in cardiac 544 fibroblasts through HIF1α 545 - Interference with this axis by pharmacologic inhibition of lysyl oxidation or genetic knock-546 out of IFNγ protects mice from pathologic ECM stiffening and diastolic dysfunction 547 Rising prevalence and limited treatment options render heart failure with preserved ejection 548 fraction (HFpEF) one of the biggest unmet needs of modern medicine. Our work demonstrates a 549 novel mechanism by which cardiac CD4+ T cells interact with cardiac fibroblasts to increase 550 extracellular matrix stiffness and cause diastolic dysfunction in experimental cardiometabolic 551 HFpEF. This suggests the lysyl oxidase LOXL3 and the inflammatory cytokine IFNγ as novel 552 therapeutic targets for patients with cardiometabolic HFpEF. 553 554 555 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint Sources of funding 556 This work was supported by National Institute of Health (NIH) Grants R01 HL144477 and 557 HL165725 (P.A.), Tufts Springboard Tier 1 Grant (P.A.). The German Research Foundation 558 (DFG) within the Walter Benjamin program 539486371 (R.E.) and the Collaborative Research 559 Center CRC1550 (C.K.). NIH F31 Grant HL159907A and AHA Predoctoral grant 906561 (SS). 560 NIH Grants HL171711 and HL127240 (T.A.M.), American Heart Association Collaborative 561 Sciences Award 24CSA1255857 (T.A.M.). NIH Grants HL147463 and HL166708 (J.G.T.). 562 563 Disclosures 564 T.A.M. is a co-founder of Myracle Therapeutics and is on the scientific advisory boards of 565 Eikonizo Therapeutics and Revier Therapeutics. 566 567 Supplemental Material 568 Tables S1-6 569 Figures S1-4 570 571 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint

572 1. Fonarow GC, Ahmad FS, Ahmad T, Albert NM, Alexander KM, Baker WL, et al. HF STATS 573 2025: Heart Failure Epidemiology and Outcomes Statistics An Updated 2025 Report from 574 the Heart Failure Society of America. J Card Fail [Internet]. 2025 Sep 22 [cited 2025 Oct 575 8];0(0). Available from: https://onlinejcf.com/article/S1071-9164(25)00326-4/fulltext 576 2. Zile MR, Baicu CF, S. Ikonomidis J, Stroud RE, Nietert PJ, Bradshaw AD, et al. Myocardial 577 Stiffness in Patients With Heart Failure and a Preserved Ejection Fraction. Circulation. 2015 578 Apr 7;131(14):1247–59. 579 3. Alcaide P, Kallikourdis M, Emig R, Prabhu SD. Myocardial Inflammation in Heart Failure 580 With Reduced and Preserved Ejection Fraction. Circ Res. 2024 Jun 7;134(12):1752–66. 581 4. Schiattarella GG, Alcaide P , Condorelli G, Gillette TG, Heymans S, Jones EAV, et al. 582 Immunometabolic mechanisms of heart failure with preserved ejection fraction. Nat 583 Cardiovasc Res. 2022 Mar;1(3):211–22. 584 5. Smolgovsky S, Bayer AL, Kaur K, Sanders E, Aronovitz M, Filipp ME, et al. Impaired T cell 585 IRE1α/XBP1 signaling directs inflammation in experimental heart failure with preserved 586 ejection fraction. J Clin Invest [Internet]. 2023 Dec 15 [cited 2025 Oct 12];133(24). Available 587 from: https://www.jci.org/articles/view/171874 588 6. Nevers T, Salvador AM, Grodecki-Pena A, Knapp A, Velázquez F, Aronovitz M, et al. Left 589 Ventricular T-Cell Recruitment Contributes to the Pathogenesis of Heart Failure. Circ Heart 590 Fail. 2015 Jul;8(4):776–87. 591 7. Nevers T, Salvador AM, Velazquez F, Ngwenyama N, Carrillo-Salinas FJ, Aronovitz M, et al. 592 Th1 effector T cells selectively orchestrate cardiac fibrosis in nonischemic heart failure. J 593 Exp Med. 2017 Oct 2;214(11):3311–29. 594 8. Yang J, Savvatis K, Kang JS, Fan P, Zhong H, Schwartz K, et al. Targeting LOXL2 for 595 cardiac interstitial fibrosis and heart failure treatment. Nat Commun. 2016 Dec 596 14;7(1):13710. 597 9. El Hajj EC, El Hajj MC, Ninh VK, Gardner JD. Inhibitor of lysyl oxidase improves cardiac 598 function and the collagen/MMP profile in response to volume overload. Am J Physiol-Heart 599 Circ Physiol. 2018 Sep;315(3):H463–73. 600 10. López B, Querejeta R, González A, Beaumont J, Larman M, Díez J. Impact of Treatment on 601 Myocardial Lysyl Oxidase Expression and Collagen Cross-Linking in Patients With Heart 602 Failure. Hypertension. 2009 Feb;53(2):236–42. 603 11. Black LD, Brewer KK, Morris SM, Schreiber BM, Toselli P, Nugent MA, et al. Effects of 604 elastase on the mechanical and failure properties of engineered elastin-rich matrices. J Appl 605 Physiol. 2005 Apr;98(4):1434–41. 606 12. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, et al. Fiji: an 607 open-source platform for biological-image analysis. Nat Methods. 2012 Jul;9(7):676–82. 608 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint 13. Hahn VS, Knutsdottir H, Luo X, Bedi K, Margulies KB, Haldar SM, et al. Myocardial Gene 609 Expression Signatures in Human Heart Failure With Preserved Ejection Fraction. 610 Circulation. 2021 Jan 12;143(2):120–34. 611 14. Lanzer JD, Wienecke LM, Ramirez Flores RO, Zylla MM, Kley C, Hartmann N, et al. Single-612 cell transcriptomics reveal distinctive patterns of fibroblast activation in heart failure with 613 preserved ejection fraction. Basic Res Cardiol. 2024 Dec 1;119(6):1001–28. 614 15. Bayer AL, Zambrano MA, Smolgovsky S, Robbe ZL, Ariza A, Kaur K, et al. Cytotoxic T cells 615 drive doxorubicin-induced cardiac fibrosis and systolic dysfunction. Nat Cardiovasc Res. 616 2024 Aug;3(8):970–86. 617 16. Bayer AL, Magri Z, Muendlein H, Poltorak A, Alcaide P . MyD88 determines T cell fate 618 through BCAP-PI3K signaling. J Immunol. 2025 Mar 1;214(3):433–45. 619 17. Schiattarella GG, Altamirano F, Tong D, French KM, Villalobos E, Kim SY, et al. Nitrosative 620 stress drives heart failure with preserved ejection fraction. Nature. 2019 Apr;568(7752):351–621 6. 622 18. Bigos KJ, Quiles CG, Lunj S, Smith DJ, Krause M, Troost EG, et al. Tumour response to 623 hypoxia: understanding the hypoxic tumour microenvironment to improve treatment 624 outcome in solid tumours. Front Oncol [Internet]. 2024 Jan 30 [cited 2025 Dec 2];14. 625 Available from: 626 https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2024.1331355/full 627 19. Mascherbauer J, Marzluf BA, Tufaro C, Pfaffenberger S, Graf A, Wexberg P , et al. Cardiac 628 Magnetic Resonance Postcontrast T1 Time Is Associated With Outcome in Patients With 629 Heart Failure and Preserved Ejection Fraction. Circ Cardiovasc Imaging. 2013 630 Nov;6(6):1056–65. 631 20. Hahn VS, Yanek LR, Vaishnav J, Ying W, Vaidya D, Lee YZJ, et al. Endomyocardial Biopsy 632 Characterization of Heart Failure With Preserved Ejection Fraction and Prevalence of 633 Cardiac Amyloidosis. JACC Heart Fail. 2020 Sep;8(9):712–24. 634 21. Rosch S, Kresoja KP, Besler C, Fengler K, Schöber AR, von Roeder M, et al. 635 Characteristics of Heart Failure With Preserved Ejection Fraction Across the Range of Left 636 Ventricular Ejection Fraction. Circulation. 2022 Aug 16;146(7):506–18. 637 22. Travers JG, Wennersten SA, Peña B, Bagchi RA, Smith HE, Hirsch RA, et al. HDAC 638 Inhibition Reverses Preexisting Diastolic Dysfunction and Blocks Covert Extracellular Matrix 639 Remodeling. Circulation. 2021 May 11;143(19):1874–90. 640 23. Ikeda M, Ide T, Matsushima S, Ikeda S, Okabe K, Ishikita A, et al. Immunomodulatory Cell 641 Therapy Using αGalCer-Pulsed Dendritic Cells Ameliorates Heart Failure in a Murine 642 Dilated Cardiomyopathy Model. Circ Heart Fail. 2022 Dec;15(12):e009366. 643 24. Perryman L, Findlay A, Baskar J, Charlton B, Foot J, Hamilton R, et al. The small molecule 644 LOXL2 inhibitor SNT-5382 reduces cardiac fibrosis and achieves strong clinical target 645 engagement. Sci Rep. 2025 Jul 2;15(1):22653. 646 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint 25. Tong D, Schiattarella GG, Jiang N, May HI, Lavandero S, Gillette TG, et al. Female Sex Is 647 Protective in a Preclinical Model of Heart Failure With Preserved Ejection Fraction. 648 Circulation. 2019 Nov 19;140(21):1769–71. 649 26. Filipp M, Ge ZD, DeBerge M, Lantz C, Glinton K, Gao P, et al. Myeloid Fatty Acid 650 Metabolism Activates Neighboring Hematopoietic Stem Cells to Promote Heart Failure With 651 Preserved Ejection Fraction. Circulation. 2025 May 20;151(20):1451–66. 652 653 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint Figures654 655 Figure 1: Mechanical ECM remodeling correlates with diastolic dysfunction in 656 cardiometabolic HFpEF. A: 8-12-week-old wild-type C57Bl/6J mice received standard chow and 657 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint water (STD) or the combination of high fat diet (HFD, 60% fat) and drinking water containing 0.5 658 g/L of the hypertension-inducing agent L-nitro-arginine-methyl-ester (L-NAME) for 5 weeks. B+C: 659 Diastolic function assessed by Doppler -mode echocardiography: representative recordings of 660 mitral valve inflow velocity (B) and ratio of early/late mitral valve inflow velocity (E/A, N= 11-661 12/group, C). D: Preparation of decellularized left ventricular (LV) ECM samples for uniaxial tensile 662 test. E: Representative recordings of uniaxial tensile test on LV-ECM samples from STD vs H/L-663 treated mice. F: Elastic modulus of LV -ECM (N=12/group). G: Elastic modulus of LV -ECM from 664 mice on STD diet, HFD, L-NAME in drinking water, or the combination of HFD and L-NAME (H/L) 665 for 5 weeks. H: Simple linear regression between E/A and LV-ECM Elastic modulus after 5 weeks 666 of STD- or H/L-treatment. I: Representative images of Picrosirius red staining of LV sections of 667 STD- or H/L-treated mice with quantification of collagen area fraction (N=3/group). J: Weight of 668 dry, insoluble ECM peptides as a fraction of whole tissue weight for LV from STD- vs H/L-treated 669 mice (N=6-7/group). 670 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint 671 Figure 2: Mechanical ECM remodeling in murine cardiometabolic HFpEF is T cell -672 dependent and human HFpEF transcriptome is characterized by a T cell signature. A: 673 Number of CD4+ T cells identified by flow cytometry in mice fed a STD diet for 3 weeks or H/L for 674 3 or 5 weeks (N=12/group). B: Elastic modulus of LV-ECM of STD- (5 weeks) or H/L-treated mice 675 (3 or 5 weeks, N=12/group). C+D: Diastolic function assessed by Doppler -mode 676 echocardiography in wild -type or Tcra-/- C57Bl/6J mice after 5 weeks of STD or H/L treatment: 677 representative recordings of mitral valve inflow velocity (C) and ratio of early/late mitral valve 678 inflow velocity (E/A, D, N=6 -8/group). E+F: Uniaxial tensile testing of decellularized LV -ECM 679 derived from WT or Tcra-/- mice after 5 weeks of STD- or H/L-treatment: Representative recordings 680 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint (E) and Elastic modulus (F, N=6-8/group). G+H: Bulk RNA sequencing of human LV tissue from 681 control and HFpEF patients. GO analysis of significantly upregulated genes in HFpEF compared 682 to control samples. Shown are the 10 most significantly enriched biological processes (G) and 683 cellular compartments (H). I: Correlation analysis of collagen (green) and lysyl oxidase (blue) 684 encoding genes with CD4. J: Simple linear regression and 95% confidence interval between gene 685 expression of the lysyl oxidase LOXL3 and CD4 across control and HFpEF samples. 686 687 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint 688 Figure 3: Cardiac fibroblasts express higher levels of lysyl oxidases in cardiometabolic 689 HFpEF. A: Lysyl oxidase expression score across all identified cell types including data from STD- 690 and H/L-treated mice. B: UMAP of CFB isolated from the full data set reveals 6 distinct CFB 691 subclusters. C: Relative abundance of each CFB subcluster in STD - compared to H/L -treated 692 mice. D: Gene ontology (GO) analysis of CFB -cluster 0 marker genes. Shown are the most 693 significantly enriched biological processes. E: Lysyl oxidase expression score across CFB in STD 694 vs H/L-treated mice. F: Cardiac LOXL3 protein levels assessed by ELISA in wild -type C57Bl/6J 695 mice after 3 or 5 weeks of STD- or H/L-treatment (N=12/group). 696 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint 697 Figure 4: CD4+ T cell-derived IFNγ induces LOXL3 expression in CFB through HIF1α which 698 is sufficient for ECM stiffening in vitro. A: Cardiac fibroblasts (CFB) isolated from C57B l/6J 699 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint mice were treated with the secretome of activated CD4+ T cells. B: Loxl3 mRNA expression levels 700 in CFB treated with the secretome of activated CD4+ T cells for the indicated durations (expressed 701 as fold change compared to non -treated CFB (NC), N=6). C: Relative lysyl oxidase activity in 702 culture media from CFB after treatment with CD4 + T cell secretome for 24 h (N=6). D: In vitro 703 collagen gel contraction by CFB treated with control (NC), CD4+ T cell secretome or recombinant 704 TGFβ (100 ng/mL) after 24 h (N=4). E+F: Uniaxial tensile testing of dECM after incubation with 705 CFB secretome. Representative recordings (E) and fold change of Elastic modulus (F) of dECM 706 preparations after incubation with CFB secretome after the indicated treatments in presence or 707 absence of β-aminopropionitrile (BAPN, 500 uM, N=4). G+H: Relative Loxl3 mRNA levels in CFB 708 after 24 h of treatment with the indicated recombinant cytokines (IL-2: 25 U/mL, IFNγ: 100 U/mL, 709 IL-4: 500 ng/mL, N=5, G) or CD4+ T cell secretome in presence or absence of αIFNγ (10 µg/mL, 710 N=3, H) for 24 h. I: Fold change of the Elastic modulus of dECM after incubation with secretome 711 of CFB subjected to the indicated treatments. J: CD4+ T cell derived IFNγ triggers HIF1α signaling 712 in CFB to induce the expression LOXL3 and subsequent ECM remodeling. K: Nucleus (DAPI, 713 blue) and HIF1 α ( green) staining in CFB after 24 h of treatment with CD4 + secretome. L: 714 Quantification of the nuclear HIF1α signal intensity from K (N=3-4). M: Relative Loxl3 mRNA levels 715 after 24 h of treatment with CD4 + secretome in the presence or absence of HIF1 α inhibitor 716 Echinomycin (Echi, 5 nM, N=3). N: Fold change of the Elastic modulus of dECM after incubation 717 with secretome of CFB treated as indicated (N=4). 718 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint 719 Figure 5: In vivo, IFNγ and lysyl oxidation are required for ECM stiffening and diastolic 720 dysfunction. A: Wild -type C57Bl/6J mice received 25 kU recombinant IFN γ or PBS for 5 721 consecutive days by i.p. injection (N=6). B-C: Relative Hif1α (B) and Loxl3 (C) mRNA as well as 722 LOXL3 protein (D) levels in the LV of PBS or IFN γ treated mice. E: Wild-type or Ifng-/- C57Bl/6J 723 mice were subjected to H/L treatment for 5 weeks. F -G: Elastic modulus of LV -ECM (F) and 724 diastolic function expressed as E/A ratio (G) in wild-type and Ifng-/- mice after 5 weeks of HFD/L-725 NAME treatment (N=12/group). H: Wild-type C57Bl/6J were subjected to HFD/L-NAME treatment 726 for 5 weeks with daily PBS or β-aminopropionitrile (BAPN, 100 mg/kg) injections from week 3 to 727 5. I-J: Elastic modulus of LV-ECM (I) and diastolic function expressed as E/A ratio (H) in wild-type 728 mice receiving PBS or BAPN during weeks 3-5 of HFD/L-NAME treatment (N=12/group). 729 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 18, 2026. ; https://doi.org/10.64898/2026.03.16.712110doi: bioRxiv preprint