Experiments were approved by the Animal Ethics Committee at the Alfred Research Alliance, Melbourne, Australia. All mice were on a C57BL/6 genetic background. NKG2D knockout (NKG2D^-/-) mice were from Jackson Laboratory, Bar Harbor, Maine, USA; perforin knockout (pfp^-/-) mice were from Mark Smyth, Peter MacCallum Cancer Centre, Melbourne, Australia, CD8β knockout mice were from David Davis, Australian National University, Canberra, Australia with approval from Dan Littman, New York School of Medicine, USA and wild-type C57BL/6 mice were from WEHI, Melbourne, Australia. Mixed chimeric mice whose CD8+ T cells were deficient in perforin or NKG2D were generated using our previously published procedure (24). Briefly, 6-week-old male wildtype C57BL/6 mice were irradiated with 1100 Gy. To generate mixed chimeric mice with perforin-deficiency limited to CD8+ T cells, irradiated mice were transplanted with 5 million mixed bone marrow cells from CD8β-knockout (80%) and perforin-deficient mice (20%); control mice received bone marrow from CD8β-knockout mice (80%) and wild type mice (20%). An identical strategy was used to generate chimeric mice with NKG2D-deficiency limited to CD8+ T cells. Four weeks after transplantation mice were subjected to TAC surgery. Bone marrow reconstitution and specific perforin/NKG2D deficiencies were confirmed by FACS/immunofluorescence.

Ketamine 80mg/Kg, Xylazine 20mg/Kg and Atropine 0.06mg/Kg (via intraperitoneal injection) were used to induce a short anesthesia for surgical procedures. Hypertension/pressure overload was induced by trans-aortic constriction (TAC) of the thoracic aorta reducing aortic lumen to 0.44mm as previously described by us (12). In Sham mice the aortic arch was not narrowed. Two-kidney one-clip (2K1C) hypertension was also induced in anaesthetized mice. The left renal artery exposed and a silver clip (120 µm gap) placed around the artery (25). Sham operated mice did not undergo clipping. For unilateral nephrectomy-DOCA-salt (1K-DOCA-salt) hypertension the left kidney was removed and a deoxycorticosterone acetate (DOCA) pellet (2.4 mg/day, 21 days; Innovative Research of America, USA) implanted subcutaneously (26). Sham mice did not receive DOCA. These mice were given 0.9% saline as drinking water.

Blood pressure and left ventricular diastolic function were measured by micromanometry under anaesthesia [ketamine, xylazine and atropine (80/20/0.06 mg/kg i.p.)] whilst mice were ventilated 28 days post-surgeries using a 1.4F Millar microtipped transducer catheter (Millar Instruments, Houston, TX, USA). The Miller catheter was then inserted into the aorta and the aortic blood pressure recorded using a computer and AD Instrument software (AD Instruments, Australia). The catheter was then advanced into the left ventricle to assess left ventricular diastolic function; –dp/dt (min), relaxation time constant (Tau), and left ventricular end-diastolic pressure (LVEDP) as previously described (27).

For CD8+ T cell depletion, the rat anti-mouse CD8β (lyt-3) monoclonal antibody YTS 156.7 prepared from a hybridoma provided by Steve Cobbold, Oxford University, UK. Mice were administered 0.5 mg antibody/mouse/week i.p, for 4 weeks; IgG2b isotype was administered to control mice. CD4+ T cells were depleted using an anti-CD4 monoclonal antibody (GK1.5) 0.3mg/week i.p. and controls were given non-immune IgG2b. For STING inhibition mice were administered H-151 (0.21mg/day i.p.) commencing one day after surgery for 4 weeks; H-151 covalently blocks STING palmitoylation preventing TBK1 activation and STING phosphorylation (28, 29).

At the end of experiment, mice were euthanized using the gradual-fill method of carbon dioxide overdose followed by cervical dislocation. However, mice anesthetised for hemodynamic measurements were euthanized via cervical dislocation.

Flow cytometry was performed as previously described (24), using APC-conjugated rat anti-mouse CD4 mAb (clone: RM4-5; BD Biosciences), PerCP conjugated anti-mouse CD8α mAb (clone: 53-6.7; BD Biosciences), APC-Cy7conjugated Armenian hamster anti-mouse TCR-β mAb (clone: H57-597; Biolegend) and FITC conjugated rat anti-mouse CD19 antibody (BD Pharmingen). For perforin staining PE conjugated rat anti-mouse perforin (clone: eBioOMAK-D; eBiosciences) was used. For intracellular perforin, spleen cells were stimulated for 12 hours with phorbol 12-myristate 13-acetate, ionomycin, brefeldin A and monensin (eBioscience, San Diego, CA). A BD LSRFortessa was used to acquire flow cytometry data, analysed using BD FACSDiva.

The left ventricle (LV) was dissected into three equal transverse LV slices and frozen in OTC at -70°C; 6 µm-thick sections were cut for histology/immunofluorescence stainings and 30 µm-thick sections for ROS assessment.

For collagen deposition sections were stained with Picrosirius Red F3BA, (0.1% solution in saturated aqueous picric acid) and analysed using a computer –interfaced colour Imaging system (Optimus Bioscan 2. Thomas Optical Measurement System, Inc). Ten randomly selected fields from each section were analysed and expressed as percent of the stained area (12, 14).

Six µm sections were allowed to thaw then fixed in cold (-20°C) acetone for 20 min. Sections were sequentially incubated in 3% hydrogen peroxide in PBS (20 min), 10% normal serum/PBS (30 min) and biotin/avidin-blocking reagents (15 min each reagent) (Vector Laboratories). These sections were then incubated (1hr) with primary antibodies in serum: rat anti-mouse CD4 (BD: cat#550280), rat anti-mouse CD8 (BD: cat#550281), rat anti-mouse CD68 (BioRad: cat#MCA1957), rabbit anti-mouse TGFβ1 (Novus Biological: cat#NB100-91995), or corresponding non-immune IgG’s. Then, sections were washed and incubated with the appropriate secondary antibody (biotinylated mouse anti-rat: BD Pharmingen cat#550325 or biotinylated anti-rabbit; Vector Labs: cat#BA-1000)] for 40 min, followed by incubation with streptavidin horseradish peroxidase complex (30 min) (Vector Laboratories) and counterstaining with haematoxylin. Antigens were visualised using 3, 3-diamonobenzidine (Sigma). Cell expressing antigens were quantitated by cell counting in ten randomly selected fields per section.

Sections were fixed in acetone, washed and permeabilized in 0.1% Triton-X-100/PBS. After washings, sections were incubated in 10% normal serum/PBS before being incubated with primary antibodies in serum: rabbit anti-mouse CD8 (Sino Biological: cat#50389-R20), mouse anti-mouse NK1.1 (Biolegend: cat#108701), Armenian Hamster- anti-mouse TCR γ/δ (Biolegend: cat#108101), rat anti-mouse Perforin (LS Biosciences: cat#LS-C18632), biotin rat anti-mouse NKG2D (R&D Systems; cat#BAM1547), rabbit anti-mouse TGFβ1 (Novus Biological: cat#NB100-91995), rat anti-mouse CD68 (BioRad: cat#MCA1957), rabbit anti-mouse troponin (Abcam: cat#ab125266), goat anti-mouse RAE-1 (R&D Systems: cat#AF1136), rabbit anti-mouse α-Actinin (Acris Antibodies: cat#BP210), rabbit anti-mouse phosphor-STING (Invitrogen: cat# PA5-105674), goat anti-mouse CX3CR1 (R & D Systems: cat#AF5825), mouse anti-mouse phospho p38MAPK (Biolegend: cat#690201), rabbit anti-mouse phospho-IRF3 (Cell Signaling: cat#29047), rabbit anti-mouse phospho-TBK1 (Imgenex: cat# IMX-5194), mouse anti-mouse NF-kB (Santa Cruz: cat# sc-109), rabbit anti-mouse MCP-1 (Abcam: cat #ab7202), rabbit anti-mouse α-Actinin Biotin (Bioss antibodies: cat# BS-10367R-Biotin) or corresponding isotype matched control IgG antibodies. Then sections were then washed and incubated with secondary antibodies/conjugates (goat anti-rat Alexa 488, Invitrogen: cat#A11006; goat anti-rabbit Alexa 488, Invitrogen: cat#A11034; goat anti-rabbit TRITC, Novex: cat# A24542; goat anti-rat TRITC, Novex: cat#A18870; donkey anti-goat TRITC, Invitrogen: cat#A16004 or Streptavidin Alexa 488, Invitrogen: cat#S11223; goat anti-mouse Alexa 488, Invitrogen: cat# A11029; goat anti-rabbit Alexa 568, Invitrogen: cat#A11036 or streptavidin TRITC, Serotec: cat# STAT3B) for 30 min, washed in PBS and counterstained with 4’,6-diamidino-2-phenylindole dihydrochloride (DAPI). For dual fluorescence stainings sections were first incubated with the first primary antibody/secondary antibody and after multiple washings were incubated with the second primary antibody and its corresponding secondary antibody. Antigens were assessed using an Olympus BX61 microscope with a FV11 CCD camera using cellSens imaging software (Olympus).

Apoptotic cells were identified using a deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) system [In Situ Cell Death Detection Kit-POD (Roche: cat#11-684-817-910)]. Briefly, sections were incubated in 4% diethyl pyrocarbonate (DEPC)/100% ethanol at 4C, washed and fixed in acetone. Sections were permeabilized in 0.1% Triton-X-100 and incubated in 15% FCS/2% BSA/0.1M Tris-HCl buffer and after washing, they were incubated with a TUNEL reaction mixture at 37C (according to the manufacturer’s instructions), washed and counterstained with DAPI. TUNEL positive cells were visualized using Olympus BX61 microscope. For double immunofluorescence to identify apoptotic cardiomyocytes frozen sections were stained by TUNEL followed by rabbit anti-mouse α-Actinin (Acris Antibodies, Cat#BP210). Then goat anti-rabbit Alexa 568 (Invitrogen, cat#A11036) was used to detect Actinin. Nuclei were counterstained with DAPI and apoptotic cardiomyocytes visualised using a Nikon A1r confocal microscope. Ten randomly selected fields per section were analysed.

Cytosolic DNA and cardiomyocytes were stained with Pico488 (picogreen), a highly sensitive fluorescent dye that can detect double stranded cytoplasmic DNA (30, 31) and anti-α-Actinin antibody. Six µm formalin-fixed permeabilized sections of left ventricle were incubated with Pico488 DNA quantification solution (Lumiprobe, Cat#12010) followed by polyclonal rabbit anti-mouse α-Actinin antibody (see above). Then goat anti-rabbit Alexa 568 secondary antibody (see above) was used to detect Actinin. After counterstaining with DAPI the sections were visualised using a Nikon A1r confocal microscope.

Reactive oxygen species (ROS) were measured using dihydroethidium (DHE). Sections were incubated with DHE (10 μM, Krebs bicarbonate buffer), protected from light at 37°C for 30 min. Images were obtained using a fluorescence microscope measuring fluorescence via a 585 nm long-pass filter (12).

Randomized and coded animals and samples were processed and analyzed by blinded investigators. All data was analyzed statistically using Prism 8.0 (GraphPad Software, La Jolla, CA, USA). All results are expressed as man ± SEM. n represents mouse numbers in each experiment, as detailed in figure legends. Comparison between two groups were determined by two-tailed Student’s t-test, and three or more groups were compared by one-way ANOVA followed by Tukey-Kramer post hoc analysis. P < 0.05 was considered statistically significant.

We used three different hypertensive mouse models to assess the importance of CD8+ T cells for development of cardiac fibrosis: 1) mice whose systolic blood pressure is elevated by trans-aortic constriction (TAC) inducing left ventricular pressure overload 2) mice with high-renin salt-independent 2K1C renovascular hypertension and 3) mice with salt-sensitive low-renin deoxycorticosterone (DOCA)-salt hypertension. We used highly specific anti-CD8β antibodies to deplete CD8αβ+ T (CD8+ T) cells (32). Their depletion during the 4-week study period reduced ventricular CD8+ T cells in all three models ( Figures 1A , S1A, G ). Compared to normotensive sham-operated mice (Systolic Blood Pressure (SBP) = 103 ± 5 mmHg; n=10), mice subjected to TAC, 2K1C or DOCA procedure were hypertensive ( Figures 1B , S1B, H ). Neither hypertension ( Figures 1B , S1B, H ), left ventricular hypertrophy ( Figures 1C , S1C, I ) nor cardiac CD4+ T cells were affected ( Figure 1D ) but left ventricular fibrosis was significantly reduced ( Figures 1E , S1D, S1J ). As fibrosis is associated with poor cardiac function due to impaired ventricular relaxation, we also assessed using micromanometry how CD8+ T cell depletion affected left ventricular diastolic function and relaxation (33). CD8+ T cell depletion greatly improved left ventricular relaxation (-dp/dt) and reduced isovolumic relaxation constant (tau) and left ventricular end diastolic pressure (LVEDP) ( Figures 1F , S1E, K ). Since CD8+ T cells can be cytotoxic, we assessed effects on apoptotic cells. CD8+ T Cell depletion reduced cardiac apoptotic cell numbers by 70-79% ( Figures 1G , S1F, L ). As macrophages are also important for fibrosis (34), and known to remove apoptotic cells by phagocytosis (35), thus increasing their biosynthesis of TGF-β1 (14, 21), we also assessed macrophage TGF-β1 expression in the TAC mice. TGF-β1-positive macrophages were most apparent in fibrotic regions and greatly reduced following CD8+ T cell depletion ( Figure 1H ). These findings indicate that TGF-β1 produced by macrophages is an early signaling event; this TGF-β1, by activating fibroblasts to produce collagen can further elevate TGF-β1 levels by positively regulating its own expression in fibroblasts (22, 23), greatly increasing the pro-fibrotic signal. To determine whether the CD8+ T cells exhibited a memory phenotype, we assessed their expression of CX₃CR1. The majority of CD8+ T cells in fibrotic areas of TAC hearts expressed CX₃CR1 ( Figure 1I ), consistent with a memory CD8+ T cell subset that possesses less proliferative ability, but higher cytotoxic effector function (36). Given the commonality between cardiac fibrosis, CD8+ T cells and cardiac apoptosis in the different hypertensive mice, we focused our subsequent mechanistic studies only on TAC mice.

As CD4+ T cells contribute to development of cardiac fibrosis (16), we examined whether CD4+ T cells provide help to CD8+ T cells. CD8+ T cells initiating low inflammatory responses are known to be highly dependent on CD4+ T cell help (37). We depleted CD4+ T cells in TAC mice using anti-CD4 cell depleting antibodies and 28 days later examined effects on CD8+ T cells and pro-fibrotic responses. Treatment with depleting antibodies reduced spleen CD4+ cell numbers by approximately 95% without effecting CD8+ cells ( Figure 2A ). In contrast cardiac CD8+ T cells were reduced by 55% ( Figure 2B ) and fibrosis by 60% ( Figure 2C ); apoptotic cell numbers were also reduced, by 45% ( Figure 2D ). Cardiac macrophage numbers were reduced ( Figure 2E ), as were cells expressing TGF-β1 ( Figure 2F ). These findings indicate that CD4+ T cells enhance the profibrotic actions of CD8+ T cells, mostly likely by increasing their survival (38).

Given the relationship between CD8+ T cells and cardiac apoptotic cells and the earlier proposal suggesting that CD8+ T cells promote fibrosis via innate mechanisms (17), we examined whether cardiac CD8+ T cells were cytotoxic and expressed innate NKG2D receptors; NKG2D receptors recognize stress-induced surface ligands and are sufficient to initiate a killing response (38). The majority of the CD8+ T cells in ventricular fibrotic areas expressed perforin ( Figure 3A ) and NKG2D ( Figure 4A ) in TAC mice. To determine if perforin was required for fibrosis, we generated mixed chimeric (CD8T-Pfp^-/-) mice in which CD8+ T cells were made deficient in perforin (24, 39), transferring mixed bone marrow [80% from CD8β-deficient and 20% from perforin-deficient mice] into irradiated C57BL/6 mice; control (CD8T-Pfp^+/+) mice received mixed bone marrow from CD8β-deficient (80%) and 20% from wild type mice (24, 39). In these mice CD8+ T cells are derived from bone marrow of either the perforin-deficient or wild type mice. Four weeks after bone marrow transplantation major lymphocyte populations in the spleen were comparable to those in non-irradiated C57BL/6 mice ( Figure 3B ). FACS analysis confirmed that in CD8T-Pfp^-/- mice, TCRβ+ CD8+ T cells did not express perforin ( Figure 3C ). Mixed chimeric mice were subjected to TAC surgery and effects on cardiac fibrosis examined 4-weeks later. Cardiac fibrosis was nearly abolished in mice with perforin-deficient CD8+ T cells ( Figure 3D ). Also cardiomyocyte apoptosis was markedly attenuated ( Figure 3E ). We assessed effects on ventricular macrophages and TGF-β1. Both macrophages and TGF-β1 were greatly reduced when perforin is deleted from CD8+ T cells ( Figures 3F, G ). In accordance with these findings, previous studies have reported that macrophages are the major apoptotic cell-clearing phagocytes in tissue homeostasis (40, 41) and their production of TGF-β1 is increased following phagocytosis of apoptotic cells (21, 42). To assess the importance of NKG2D we also generated mixed chimeric mice using bone marrow from CD8β-deficient, NKG2D-deficient and wild type mice and subjected the mice to TAC surgery. We confirmed that cardiac CD8+ T cells in hearts of chimeric mice did not express NKG2D ( Figure 4B ), however, NK/NKT and TCRγδ+ T cells accumulated in myocardial fibrotic areas expressed NKG2D receptor in TAC mice with NKG2D deficiency limited to CD8 T cells ( Figure S2 ). Cardiac fibrosis in these mice was nearly abolished ( Figure 4C ); apoptotic cardiomyocytes were greatly reduced ( Figure 4D ) as were macrophage numbers and TGF-β1 expressing cells ( Figures 4E, F ). These findings indicate that CD8+ T cells are activated locally within the left ventricle by mechanisms dependent on the NKG2D receptor and use perforin to promote fibrosis.

We have previously shown that during development of cardiac fibrosis reactive oxygen species (ROS) are greatly increased (12). Increases in ROS promote mitochondrial DNA damage and its leakage into the cytoplasm (43) thus activating the cytoplasmic STING-TBK1-IRF3-IFN-β signaling pathway (44). As RAE-1 ligands which activate the NKG2D receptor are regulated via IRF3 (45), we investigated whether cardiac RAE-1 and the cGAS-STING pathway might be important for CD8+ T cell mediated interstitial fibrosis in the hypertensive heart. We confirm that high ROS levels in hypertensive hearts ( Figure 5A ) and using fluorescent DNA binding molecule pico488 (PicoGreen) (46), demonstrate the presence of cytosolic DNA in cardiomyocytes of hypertensive hearts ( Figure 5B ). Cardiomyocyte RAE-1 was also greatly elevated ( Figure 5C ); similarly, phosphoSTING was detectable in stressed cardiomyocytes ( Figure 5D ), indicating cGAS-STING DNA sensor pathway activation (43, 44). To determine whether this pathway was responsible for the RAE-1 expression, CD8+ T cell mediated apoptosis and fibrosis we treated TAC mice with H-151, a small molecule inhibitor that blocks STING plamitoylation, which is required for STING signaling (28, 47). Treatment inhibited cardiac STING signaling, indicated by the absence of phosphoSTING ( Figure 5E ); treatment also prevented RAE-1 expression ( Figure 5F ) and cardiac cell apoptosis ( Figure 6A ), resulting in reduced cardiac fibrosis ( Figure 6B ). Macrophages and TGF-β1 expressing cells were also greatly reduced ( Figures 6C, D ). Neither left ventricular hypertrophy nor blood pressure were affected ( Figures 6E, F ); left ventricular micromanometry indicated markedly improved left ventricular relaxation and diastolic function in H-151 treated mice ( Figure 6G ).

To confirm involvement of the STING, TBK-IRF3-NFkB signaling cascade we assessed the phosphorylation status of these molecules in relation to their intracellular location within cardiomyocytes. Phospho-TBK was apparent, localized in the perinuclear region ( Figure 7A ), whilst phospho-IRF3 appeared mostly nuclear ( Figure 7B ), as was NF-kB ( Figure 7C ) and phospho-p38 MAPK ( Figure 8A ). Because NF-kB and p38MAPK have been associated with increased MCP-1 (48, 49), a chemokine that not only attracts macrophages but also enhances their phagocytic activity (50), we examined its expression in cardiomyocytes. MCP-1 was increased during development of fibrosis and completely prevented by treatment with H-151 ( Figure 8B ) as were all other phosphorylation signals in the TBK-IRF3-NF-kB signaling cascade ( Figures 7 , 8A ); STING signaling was not apparent in macrophages associated with fibrosis ( Figure 8C ).