Sorting nexin 3 induces myocardial injury via promoting retromer-dependent nuclear tracking of STAT3

Sorting nexins (SNXs), the retromer-associated cargo binding proteins, have emerged as critical regulators of the tracking of proteins involved in the pathogenesis of diverse diseases. However, studies of SNXs in the development of cardiovascular diseases, especially pathological cardiac hypertrophy, are lacking. Here, we asked whether SNX3, the simplest structured isoform in the SNXs family, may act as a key inducer of myocardial injury. We reported that increased level of SNX3 was observed in failing hearts from human patients and mice. Cardiac-specic Snx3 knockout (Snx3-cKO) in mice signicantly protected against isoproterenol (ISO) -induced cardiac injury at 12 weeks. Conversely, cardiac-specic Snx3 transgenic (Snx3-cTg) mice were more susceptible to ISO-induced cardiac injury at 12 weeks, and showed aggravated cardiac injury even heart failure at 24 weeks. Immunoprecipitation-based mass spectrometry (IP-MS), immunouorescent staining, co-immunoprecipitation and localized surface plasmon resonance (LSPR) were performed to examine the direct interaction of SNX3-retromer with STAT3. STAT3 was discovered as a new interacting partner of SNX3-retromer, and SNX3-retromer served as an essential platform for assembling gp130/JAK2/STAT3 complexes and subsequent phosphorylation of STAT3 by direct combination at early endosomes. SNX3-retromer and STAT3 complexes were transiently imported into the nucleus after hypertrophic stimuli. Moreover, SNX3-retromer promoted importin α3-mediated STAT3 nuclear tracking and ultimately leading to cardiac injury. Also, pharmacological inhibition of STAT3 reversed SNX3 overexpression-induced myocardial injury in vivo and in vitro. Taken together, our study reveals that SNX3 plays a key role in cardiac function and implicates SNX3 as a potential therapeutic target for cardiac hypertrophy and heart failure.


Introduction
By responding to various patho-physiological stresses, the heart undergoes a transition from compensatory hypertrophy to decompensatory phase, in which left ventricular remodeling, systolic and diastolic dysfunction and eventual heart failure (HF) occur [1][2][3] . The abnormal activation of multiple cellular signaling pathways, such as protein kinase C (PKC) -mitogen-activated protein kinase (MAPK), calcineurin-nuclear factor of activated T cells (NFATs) and janus kinase (JAK) -signal transducer and activator of transcription protein (STAT), is the pathological basis for development of heart failure 3,4 .
The physiological functions of signaling proteins are closed linked to their intracellular tra cking and subcellular localization, which is determined by sorting nexins (SNXs)-retromer complex and endosomes [5][6][7][8] . SNXs are featured by a highly conserved Phox homology (PX)-domain and comprise 33 members in mammals 9 . Structurally, SNXs directly or indirectly bind to vacuolar protein sorting (VPS) family, VPS26 VPS35 and VPS29, to constitute retromer complex; SNXs also target to endosomal membrane through the PX-domain 5,6 . Functionally, SNXs recruit retromer to early endosomes (EE) or recycling endosomes (RE), and subsequently mediate retrograde transport of cargo proteins via vesicle budding from EE/RE to trans-Golgi network (TGN), plasma membrane (PM) or cell nucleus [5][6][7] . SNXs have been reported to regulate diverse disease processes, including Alzheimer's disease, cancer, heart failure and arthritis [10][11][12][13] . For instance, we previously reported that de ciency of SNX10 prevents in ammation and bone erosion in a mouse model of rheumatoid arthritis through promoting NFATc1 degradation 10 ; SNX13 profoundly affects cardiac performance through apoptosis repressor with caspase recruitment domain (ARC) -caspase signaling pathway 11 . Recently, it is found that SNX3 (the simplest structured isoform in the SNXs family)-retromer is required for retrograde recycling of Wntless 14 , PC1 and PC2 15 , the cation-independent mannose 6-phosphate receptor (CI-M6PR) 16 and iron transporters 17 . However, the role of SNX3 in the pathogenesis of cardiovascular diseases, especially cardiac hypertrophy and heart failure, remains unknown.
Signal transducer and activator of transcription 3 (STAT3), a subtype of STAT family, participates in the pathological process of cardiac hypertrophy and heart failure 4,18,19 . Cardiomyocyte-speci c overexpression of STAT3 (Stat3-Tg) in mice causes spontaneous concentric cardiac hypertrophy 4 . In response to stimulation with pro-in ammatory cytokines and growth factors, STAT3 is phosphorylated at tyrosine 705 (Y705) by receptor-associated JAK2, then form homo-or hetero-dimers, and translocate to cell nucleus where they act as transcription activators [18][19][20] . Importin α3 (also called as Kpna3), a nuclear import factor, is responsible for STAT3 nuclear import independent of tyrosine phosphorylation 21,22 . Upon axons injury exposure, the newly-synthesized STAT3 combined with importin α5 and dynein (a retrograde molecular motor), and transported back to the cell body [23][24][25] . STAT3 co-localizes with endocytic vesicles in transit from the cell membrane to the perinuclear region, the perinuclear endosomal compartment to sustain phosphorylated STAT3 (Y705) in nucleus [26][27][28][29] . To date, it remains largely unknown whether the retromer-dependent mechanism is involved in the phosphorylation and nuclear tra cking of STAT3.
Here, we demonstrate that the mRNA and protein levels of SNX3 were increased in end-stage failing human hearts and cardiac tissues from isoproterenol (ISO) -induced cardiac injury mouse model. Besides, cardiac-speci c Snx3-cKO in mice e caciously protected against ISO-induced cardiac injury at 12 weeks. Conversely, cardiac-speci c Snx3-cTg mice were hypersensitive to ISO-induced cardiac injury at 12 weeks, and resulted in cardiac hypertrophy and dysfunction even heart failure at 24 weeks. Our results reveal the importance of SNX3-retromer complex for assembling gp130/JAK2/STAT3 complexes and subsequent phosphorylation of STAT3 by direct combination at early endosomes. Hypertrophic stimulation-induced nuclear translocation of STAT3 was facilitated by importin α3-mediated SNX3-retromer importing. SNX3retromer promoted STAT3 nuclear tra cking and ultimately leading to abnormal energy metabolism and cardiac injury. Also, pharmacological inhibition of STAT3 by stattic reversed SNX3 overexpressioninduced myocardial injury in vivo and in vitro. Together, our study identi es SNX3 as a potential target for cardiac hypertrophy and heart failure.

SNX3 expression was up-regulated in human and mouse failing hearts
To explore the potential role of SNX3 in the development of HF, we analyzed heart samples from 9 patients with HF in end-stage and 3 non-failing healthy controls. The SNX3 mRNA expression was increased in failing human heart tissues compared with that in the normal control tissues as showed by qPCR (Supplementary Figure S1a). The results from immuno uorescence (IF) and western blot analysis suggest that the protein expression of SNX3 was signi cantly higher in the failing human heart tissues ( Fig. 1a and b). Similarly, the mRNA and protein levels of SNX3 were clearly induced in the myocardium of mouse HF model by subcutaneous (s.c.) injection of ISO (3 mg/kg/day, a classic hypertrophic agonist [30][31][32] for four weeks (Supplementary Figure S1b-d). These results suggest that SNX3 was associated with cardiac hypertrophy and HF in human and mice.
Cardiac-speci c Snx3 knockout attenuated ISO-induced myocardial injury in mice To examine the function of SNX3 in myocardial injury, cardiac-speci c Snx3-cKO mice model was generated by breeding Snx3oxed mice with C57BL/6-Myh6 em1(IRES-Cre)Smoc line (Supplementary Figure  S2a). A cDNA encoding internal ribosome entry site (IRES) and Cre recombinase is inserted into the 3'UTR region of α-myosin heavy chain gene to generate this Myh6-IRES-Cre mouse line 33 . As shown in Fig. 1c and Supplementary Figure. S2f, endogenous SNX3 protein expression was speci cally depleted in heart tissues of Snx3-cKO mice, compared with their littermate negative controls (CTL). Snx3-cKO mice showed no obvious cardiac structural or functional defects at basal conditions. To assess whether Snx3 deletion in uences cardiac dysfunction under stressed conditions, at 11 weeks, both Snx3-cKO mice and CTL mice were injected with ISO (3 mg/kg/day, s.c.) for one week to establish cardiac injury model. Compared with saline group, ISO treatment signi cantly induced cardiac hypertrophy, brosis and heart dysfunction in CTL mice, evidenced by gross morphologic examination, histological staining, hypertrophic biomarkers (Anp, Bnp and β-Mhc), heart weight to the tibia length ratio (HW/TL) and echocardiography ( Fig. 1d-i, Supplementary Figure S3). The results of heart morphology, wheat germ agglutinin (WGA) staining, hematoxylin-eosin (HE) staining, HW/TL ratio, as well as the mRNA levels of Anp, Bnp and β-Mhc, demonstrated that ISO treatment induced cardiomyocyte hypertrophy in CTL mice ( Fig. 1d-  Cardiac-speci c Snx3 transgene leaded to myocardial injury in mice To further determine whether SNX3 contributes to myocardial injury, cardiac-speci c Snx3 transgenic (Snx3-cTg) mice were generated by crossing Snx3 transgenic mice with B6.FVB-Tg(Myh6-Cre)2182Mds/J mice (Fig. 2a, Supplementary Figure S4). This transgenic strain uses a Cre/loxP approach in which transgenic Cre expression is driven by the mouse Myh6 promoter 34 . Compared with littermate negative controls (N-Tg), the expression of SNX3 protein is robustly expressed in the heart of Snx3-cTg mice, as indicated by western blot, immuno uorescence (IF) assay and the luciferase activity by live imaging (Fig.  2b, Supplementary Figure S4e-j).
At 12 weeks of age, Snx3-cTg mice showed a mild degree of myocardial hypertrophy, as indicated by the following observations: (1) the larger heart/cardiac size by gross morphological examination, the increased hypertrophic biomarkers and HW/TL ratio, WGA staining and HE staining ( We also investigated whether overexpression of SNX3 aggravates ISO-induced cardiac hypertrophy in Snx3-cTg mice (12-weeks-old). There were signs of increase of the cardiomyocytes size in both N-Tg mice and Snx3-cTg mice; however, Snx3-cTg mice have larger cardiomyocyte size than in N-Tg mice ( Fig. 2c-e, Supplementary Figure S5a and b). In addition, ISO-induced cardiac brosis was aggravated when SNX3 was overexpressed (Fig. 2f). After ISO infusion, severe heart dysfunction detected by echocardiography was noted in Snx3-cTg mice more than N-Tg mice ( Fig. 2g-j, Supplementary Figure S5c-h). These results suggest that Snx3-cTg mice were more susceptible to ISO-induced cardiac injury.
After 24 weeks, Snx3-cTg mice displayed signi cant hypertrophic growth of cardiomyocytes, increases in HW/TL ratios and hypertrophic biomarkers, myocardial brosis, compared with N-Tg mice (  Figure S6g), implying that Snx3-cTg mice spontaneously developed severe heart dysfunction. These results suggest that the cardiac structure was disorganized and systole diastolic performances tend to be worsened with advancing age in Snx3-cTg mice.

SNX3-retromer directly interacted with STAT3 at early endosomes
The prominent effects of SNX3 on myocardial injury ( Fig. 1-2, Supplementary Figure S1-6) prompt us to investigate its underlying mechanisms. Since SNX3 is a key retromer-associated protein 5,6 , we thus examined whether retromer is involved in SNX3-mediated myocardial injury. By immunoprecipitationbased mass spectrometry (IP-MS), we identi ed STAT3, as a new interacting partner of SNX3, in addition to known interacting proteins, such as retromer proteins VPS26 and VPS35 (Supplementary Figure S7a and b). To validate the interaction of SNX3 and STAT3, neonatal rat cardiomyocytes (NRCMs) were infected with Ad-STAT3 (HA-tagged) or Ad-SNX3 (Flag-tagged). Protein interaction was evaluated by IF staining and co-IP assays (Fig. 3a-c). Indeed, IF staining results indicate that HA-labeled STAT3 and retromer proteins (SNX3, VPS26 and VPS35) were located in the same cellular compartments (Fig. 3a). The co-IP results further suggest that SNX3-retromer interacted with STAT3 in NRCMs and mouse heart tissues ( Fig. 3b  Besides, NRCMs were precipitated by anti-early endosome antigen 1 (EEA1), Rab5 and clathrin heavy chain (CHC) (markers of early endosome), SNX3 and STAT3 were found in the precipitation of early endosome by co-IP assays. (Fig. 3e, Supplementary Figure S7m and n). The early endosome fraction was puri ed from NRCMs using density gradient centrifugation, and detected by western blot analysis ( Fig.   3f). We observed that both STAT3 and SNX3 mainly associated with early endosomes in NRCMs, though some STAT3 positive staining was present in late endosomes and lysosomes, as suggested by STAT3 combining with the endosomal compartments markers (EEA1, Rab5, Rab7, CHC and Lamp-2) (Fig. 3b, c, e, f, Supplementary Figure S7m-p). These results imply that STAT3 directly interacted with SNX3-retromer at early endosome in cardiomyocytes.

SNX3-retromer served as a platform for STAT3 activation
It is well-established that JAK2/STAT3 pathway is involved in the pathogenesis of cardiac hypertrophy and heart failure 4 . To investigate whether SNX3 cooperates with STAT3 to mediate cardiac hypertrophy, we established three classical cardiac hypertrophic models (induced by ISO, angiotensin (Ang ) and phenylephrine (PE), respectively) in NRCMs. According to the increase of cell surface area and increased mRNA expression of hypertrophic markers, three models of cardiomyocyte hypertrophy were successfully established ( Fig. 4a, b). The phosphorylation levels of JAK2 (Y1007 and Y1008) and STAT3 (Y705), and the protein expression of SNX3 were enhanced in these hypertrophic models, whereas the protein levels of gp130 and STAT3 remained unchanged (Fig. 4c, Supplementary Figure S8a). The homo-/heterodimerization of STAT3 and STAT1 was also increased by treatment with hypertrophy stimuli (Supplementary Figure S8b).
Given that SNX3 is localized in retromer-containing endosomes 6 , we investigated whether SNX3-retromer is involved in JAK2/STAT3 signaling. The results of co-IP and IF assay showed that ISO increased the association of STAT3 with SNX3, JAK2 and gp130, whereas knockdown of SNX3 decreased these interactions ( Fig. 4d-f). Consistently, knockdown of SNX3 significantly inhibited ISO-induced phosphorylation and dimerization of STAT3 ( Fig. 4g and h), whereas overexpression of SNX3 by infecting with Ad-SNX3 had the opposite effects (Supplementary Figure S8c-e). Collectively, these results suggest that SNX3-retromer serves as a platform for the assembly of gp130/JAK2/STAT3 complexes and the subsequent phosphorylation of STAT3 in ISO-induced cardiac hypertrophy.

SNX3-retromer promoted the nuclear localization of STAT3
After treatment with stimuli of cardiac hypertrophy, cytoplasmic STAT3 shuttles to the nucleus, where STAT3 orchastreates hypertrophy-related target gene expression (such as Anf, c-fos, c-myc, aGT) 4,35,36 . Given that SNX3-retromer complex plays a vital role in sorting cargoes from endosomes to the intracellular tra cking 6 , we examined the effects of SNX3-retromer on the subcellular distribution of STAT3 in the process of cardiac hypertrophy. NRCMs were treated with ISO, Ang or PE, or were infected with Ad-SNX3 or Ad-sh-SNX3 before ISO treatment. The nuclear and cytoplasmic proteins were extracted from NRCMs, and were detected by western blot analysis. The results showed that transient nuclear import of STAT3, SNX3, VPS35 and VPS26, after treatment with hypertrophic stimuli for 1 h (Supplementary Figure S9a). By contrast, SNX3 silencing reversed ISO-induced nuclear import of STAT3 using IF and western blot analysis ( Fig. 5a and b, Supplementary Figure S9b). Similarly, the ISO-induced increase of transcriptional activity and target gene expression of STAT3 was inhibited by knockdown of SNX3 using luciferase reporter assay and qPCR ( Fig. 5c and d). The above ndings revealed that nuclear translocation of STAT3 was depends on SNX3-retromer complex.
Previous studies have shown that importin α3, a nuclear importing factor, is responsible for STAT3 nuclear import induced by hormones or intracellular tyrosine kinases 21,22 . The endogenous importin α3 was knocked down using the appropriate shRNA in NRCMs (Supplementary Figure S9c and d). By IF staining, we observed that Ad-SNX3 treatment signi cantly increased ISO-induced interaction of importin α3 and STAT3-SNX3-retromer complexes (Fig. 5e). However, importin α3 knockdown attenuated Ad-SNX3triggered the nuclear importing of STAT3 and SNX3-retromer ( Fig. 5f and g, Supplementary Figure S9e), suggesting that SNX3-retromer promoted the importin α3-mediated nuclear localization of STAT3 in cardiomyocytes.

Involvement of STAT3 in SNX3-mediated cardiomyocyte injury
We then asked whether SNX3 induces cardiomyocyte injury through STAT3 activation. To this end, NRCMs were infected with Ad-sh-SNX3 before ISO treatment, or were infected with Ad-SNX3 with or without stattic (a speci c inhibitor of STAT3) 28 . We observed that knockdown of SNX3 signi cantly suppressed the hypertrophic growth of cardiomyocytes induced by ISO (Fig. 6a, b). Conversely, Ad-SNX3 led to increased hypertrophic responses, which was partly inhibited by stattic, implying that STAT3 was involved in SNX3-mediated cardiomyocyte hypertrophy (Fig. 6a, b). Given that the progression of cardiac hypertrophy is characterized by disturbed mitochondrial energy metabolism 1 , we then measured the rate of oxygen consumption (OCR) with the high-throughput analyzer Seahorse XF96. The results from Fig. 6c showed that OCR was signi cantly increased in ISO or Ad-SNX3-stimulated cardiomyocytes, both of which were inhibited by Ad-sh-SNX3 or stattic. Interestingly, Ad-SNX3 increased basal respiration, ATP production, proton leak, spare respiratory capacity, maximum respiration and non-mitochondrial oxygen consumption, which was partly inhibited by stattic (Fig. 6c, Supplementary Figure S10a-f). Similarly, Adsh-SNX3 decreased the ISO-induced OCR indices (Fig. 6c, Supplementary Figure S10g-l). These data pinpointed a critical role of STAT3 and SNX3 in ISO-induced cardiac hypertrophy and abnormal mitochondrial energy metabolism.
To further con rm the role of STAT3 in SNX3-induced cardiac injury in vivo, Snx3-cTg mice were treated with stattic (40 mg/kg/d, i.p.) or an equal volume of vehicle for 2 weeks. The phosphorylation level of STAT3 at the tyrosine 705 (pY705) was signi cantly inhibited by stattic, suggesting that the activation STAT3 was inhibited by stattic (Fig. 6e, f). After treatment with stattic, N-Tg mice showed no noticeable changes in cardiomyocytes ( Fig. 6g-j, Supplementary Figure S11a-l). However, the degree of cardiac hypertrophy (determined by gross observation of heart morphology, hypertrophic biomarkers, HW/TL ratio, WGA staining and HE staining) of Snx3-cTg mice was signi cantly alleviated in stattic treatment group compared with that in vehicle-treated group (Fig. 6g, Supplementary Figure S11a, b, d, e). Besides, Snx3-cTg mice showed increased cardiomyocyte interstitial brosis (determined by PSR staining and Masson staining), while that effect was widely suppressed upon stattic treatment (Fig. 6h,  Supplementary Figure S11c). The echocardiography data showed that Snx3-cTg reduced EF, FS, CO, LVV, LVID and SV, increased LVPW, LVM and IVS, which could be antagonized by stattic in different degree (Fig. 6j, Supplementary Figure S11f-m).
These results suggest that STAT3 was involved SNX3 overexpression-induced myocardial injury in cardiomyocytes and in Snx3-cTg mice.

Discussion
SNX3 plays a crucial role in the pathogenesis of diverse diseases, including Parkinson's disease 17 , Alzheimer disease 37 , autosomal dominant polycystic kidney disease 15 . However, it remains unknown whether SNX3 plays a role in cardiac diseases. In this study, we observed that SNX3 expression was increased in end-stage failing human hearts and cardiac tissues from mouse model of ISO-induced cardiac injury. By using gain-and loss-of-function experiments in vivo, we revealed a novel role of SNX3 in the development of cardiac hypertrophy and heart failure. This was evidenced by the fact that cardiacspeci c Snx3-cKO protected mice against ISO-induced cardiac injury, conversely, Snx3-cTg mice were hypersensitive to ISO-induced cardiac hypertrophy and aggravated cardiac injury with advancing age.
It has been reported that SNXs family affect a wide range of biological processes by regulating the intracellular tra cking of diverse signaling proteins [10][11][12][13]38 . For instance, Deficiency of SNX10 prevents inflammation and bone erosion in rheumatoid mouse arthritis through promoting NFATc1 degradation 10 ; SNX13 profoundly affects cardiac performance through the SNX13-PXA-ARC-caspase signaling pathway 11 ; Deletion of SNX27 reverses epithelial-mesenchymal-transition in highly aggressive breast cancer cells 13 ; SNX27 serves as an essential adaptor protein, mediates beta 2ARs to the retromer tubule and endosome-to-plasma membrane tra cking 38 . However, the effect of SNX3 on the intracellular transportation of cargo proteins, which is tightly associated with cardiovascular diseases, has not been evaluated yet. Here, we identify the speci c role of SNX3-STAT3 interaction in cardiomyocytes. SNX3 was required for STAT3 activation, inhibition of STAT3 could reduce the detrimental role of overexpressed SNX3 in cardiomyocytes and Snx3-cTg mice.
STAT3 activity is tightly regulated for proper physiological processes, and its aberrant and persistent activation will result in pathological cardiac hypertrophy 4,18 . Over-activation of STAT3 could be mainly attributed to either the stimulation of STAT3 activators (such as ISO and IL-6), or the aberrantly activated upstream tyrosine kinases (such as JAK2 and EGFR) [18][19][20] . The present work suggest that SNX3-retromer acts as an essential platform for the assembly of gp130/JAK2/STAT3 complexes induced by ISO, and subsequent phosphorylation of STAT3 by direct combining at early endosomes.
To date, the main research of SNXs-mediated intracellular transportation is focused on the subcellular tra cking from the endosomes to TGN or PM 6 . It is rarely reported that the nuclear transport of intracellular proteins mediated by SNXs, except that SNX11 is required for the translocation of factor II receptor-like 1 (F2rl1) from the PM to the cell nucleus in retinal ganglion cells 7 . We report that SNX3retromer complex was transiently imported into the nucleus after hypertrophic stimuli, and promoted importin α3-dependent nuclear translocation of STAT3 in the cardiac hypertrophy model. According to our work, it is plausible that SNX3-retromer promotes both STAT3 activation and nuclear translocation. The detailed mechanism whereby the complex is recruited to membrane receptor to be activated and subsequently directed to nuclear after activation is not clear yet. There may be another protein or posttranslational modi cation determines the fate of the complex for intracellular tra cking.
In conclusion, the present study demonstrates that SNX3-retromer plays a critical role in cardiac function and suggests that SNX3 could be exploited as a potentially new therapeutic target for cardiac hypertrophy and heart failure.

Human heart samples
The study conforms to the principles that govern the use of human tissues outlined in the Declaration of Helsinki. Approval was obtained from the human ethics committee of First A liated Hospital of Sun Yatsen University. All patients or the family of prospective heart donors gave written informed consent prior to participation. Failing human heart samples were collected from 9 patients undergoing heart transplantation because of end-stage heart failure (table S1). Three non-failing control heart samples were obtained from prospective multi-organ donors, which did not exhibit cardiovascular pathology but were unable to be transplanted due to technical reasons. Tissue samples were collected at the time of explantation and rapidly frozen in liquid nitrogen or xed with 4 % paraformaldehyde.

Animal Studies
Snx3oxed mice were constructed in the Shanghai Model Organisms Center by CRISPR/Cas9 technology. Brie y, the donor vector containing four guide RNAs and Cas9 mRNA targeting Snx3 introns 2 and 3 was microinjected into C57BL/6 mouse fertilized eggs. The positive founder mice were backcrossed with wild-type C57BL/6 mice to obtain heterozygous Snx3 ox/+ mice with germ line transmission. Snx3 ox/+ mice were self-crossed to generate homozygous Snx3 ox/ ox mice. Snx3 transgenic mice were constructed in the Shanghai Model Organisms Center using standard methods. Brie y, the Piggybac vector harboring mouse SNX3 cDNA was microinjected into the fertilized egg of C57BL/6 mouse, and the transgenic founder mice were obtained. 3 generations of backcross between each Snx3 transgenic mice and wild-type (C57BL/6) mice were adopted to breed two independent Snx3 transgenic lines. The two transgenic lines were respectively crossed with B6.FVB-Tg(Myh6-Cre)2182Mds/J mice (Jackson Laboratory (Bar Harbor, ME), Stock No: 011038. MGI ID: 2386742.) to generate cardiac-speci c overexpressed Snx3 (Snx3-cTg) mice.
The mice were genotyped by PCR and further con rmed by western blot analysis (table S2-S4). All animal protocols were conducted under the institutional guidelines of Animal Care and Use Committee, and were approved by the Research Ethics Committee, Sun Yat-sen University. Experimental animals were housed, bred, and maintained in the speci c pathogen-free (SPF) facility of the Experimental Animal Center of Sun Yat-sen University. As we previously described 30 , the transthoracic 2D-guided M-mode echocardiography (such as heart function and global cardiac volumes) was assessed using by a Technos MPX ultrasound system (ESAOTE, SpAESAOTE SpA, Italy) equipped with an 40-MHz scan probe. The Vevo 2100 imaging software was used for measurements and calculations. Then, the treated mice were sacri ced, and hearts were rapidly sampled for further experiments.

Bioluminescence Imaging 47
XenoLight TM D-luciferin potassisum salt (PerkinElmer, P/N 122799) was diluted in sterile PBS to 15 mg/mL, and was respectively injected into the luciferase-labeled Snx3-cTg mice or N-Tg mice (150 mg/kg). A few minutes after injection, mice were anesthetized using 2% iso urane inhalation (with a 2 L/min oxygen ow rate), and were placed inside the camera box of the IVIS Lumina XR small animal optical imaging system (PerkinElmer). The sequential images of the mice were run every two minutes.
Lumazone Version 2.0 software was used to analyze the intensity of the uorescence (intensity/second) in mice.

Histological Analysis
Myocardial tissue samples were xed in 10% paraformaldehyde and embedded in para n for sectioning.
Recombinant adenoviral vectors expressing rats Snx3 cDNA (Ad-SNX3, with Flag-tag), rats STAT3 cDNA (Ad-STAT3, with HA-tag), sh-SNX3 sequence (Ad-sh-SNX3, with Flag-tag) and control vectors were generated by standard procedures. Brie y, the pAdTrack plasmids containing corresponding genes were linearized by using restriction endonuclease Pme , and were transformed into Escherichia coli strain BJ5183 cells carrying the Ad-Easy-1 plasmid. The successful recombinant plasmids were digested with restriction endonuclease Pac , and were transfected into HEK293A cells to generate Ad-SNX3, Ad-STAT3 or Ad-sh-SNX3. The vectors were puri ed by plaque, cultured in a large scale, and puri ed by CsCl stepand isopycnic-gradient centrifugation.
GST-STAT3, GST-SNX3, GST-VPS35 and GST-VPS26 protein were respectively expressed in E. coli BL21 (DE3) maintained in luria-bertani (LB) medium (including 50 ug/mL ampicillin) at 37℃. By the addition of isopropyl-β-D-thiogalactopyranoside (IPTG, 0.5 mmol/L), protein expression was induced at 16℃ overnight to an OD600 value of 0.8. Cells were re-suspended in 20 mmol/L Tris + 200 mmol/L NaCl + 1 mmol/L PMSF, were broken using ultrasonic instrument in ice, and were collected supernatant after centrifugation (1,2000 rpm for 3 h at 4°C). These recombinant protein were puri ed by GST-Se nose Gravity Column, were enzyme digested at room temperature overnight. All puri ed protein were dissolved in sterile phosphate-buffered saline (PBS), were measured their concentration and stored at -80°C.

Localized surface plasmon resonance (LSPR) Assays 48
To examine the direct interaction between STAT3 and SNX3-retromer, LSPR assays were conducted on an OpenSPR system (Nicoya Lifesciences, Waterloo, Canada). Recombinant proteins STAT3 (117-770aa, 576-678aa), SNX3 and VPS35 served as the ligand and were respectively immobilized on a gold nanoparticle sensor chip via capture-coupling. Subsequently, the recombinant protein SNX3, VPS35 and VPS26 at different concentrations were sequentially injected into the chamber in running buffer ( ltered PBS) with a constant ow rate of 20 μL/min, and were passed over the sensor (about 5 min) for the association of two protein. Following each recombinant protein injection (all concentrations were performed in triplicate), the chip was completely dissociated with the complex and regenerated by injecting hydrochloric acid (pH 2.0). As recommended by the manufacturer, the results were analyzed by Trace Drawer software (Ridgeview Instruments AB). The kinetic parameters, including the association constant (ka), dissociation constant (kd) and a nity (KD, KD=kd/ka), were calculated by a simple 1: 1 diffusion corrected model, which adjusted to the wavelength shifts consisting with the varied concentration of protein.
Primary culture of neonatal rat cardiomyocytes (NRCMs) As reported before 30 , NRCMs were isolated from the hearts of one to three-day-old SD rats.

Plasmid Transfection and virus infections 30
NRCMs were transfected with sh-SNX3 plasmid together with lipofectamine 3000 reagent (Invitrogen, USA) in OptiMEM medium as per the manufacturer's instructions. The medium was changed to DMEM complete medium after eight to seventy two hours of transfection. NRCMs were infected with recombinant adenoviruses (including Flag-tagged SNX3, Flag-tagged sh-SNX3 or HA-tagged STAT3) at a multiplicity of infection (MOI) of 20 particles per 5 cells. Western blot and/or Quantitative RT-PCR were performed to con rm the e ciency of overexpression or depletion.

Mitochondrial respiration assay by Seahorse XF96 49
The prepared cardiomyocytes were seeded in a Seahorse XF96 polystyrene cell culture microplate (Seahorse Bioscience, North Billerica, USA), and incubated in a cell culture incubator at 37°C overnight. The next day, the medium was changed to XF base Medium (Seahorse Bioscience) containing 10 mmol/L glucose (Sigma), 1 mmol/L pyruvate (Sigma), and 2 mmol/L glutamine (Sigma) for the detection of basal respiration. Some reagents were serially injected to detect ATP production (mitochondria respiration), maximal respiration and non-mitochondrial respiration, including oligomycin (2 μmol/L, inhibits ATP synthase (complex ), FCCP (1 μmol/L, uncouples oxygen consumption from ATP production), and a mix of rotenone (0.5 μmol/L, inhibits complexes ) and antimycin A (0.5 μmol/L, inhibits complexes ). Proton leak and spare respiratory capacity were calculated by these parameters using a high-throughput XF e 96 extracellular ux analyzer (Seahorse Bioscience).
Measurement of the cell surface area 30,50,51 NRCMs were seeded in 24-well microplates, xed with paraformaldehyde (4%, Beyotime, #P0099) for 10 min at room temperature. After permeabilizing with Triton X-100 (0.3%, Beyotime, #P0096) and blocking with goat serum (Beyotime, #C0265), the cells were incubated with primary antibody myosin light chain 2 (MLC2, Proteintech, #10906-1-AP) overnight at 4℃, treated with secondary antibody anti-alexa uor 488 (Proteintech, #SA00013-6, for 2h) and rhodamine-phalloidin (0.1%, Invitrogen #R415, for 30 min) at room temperature. After washing with ltered PBS, NRCMs were mounted by using DAPI (CST, #4083). The images were taken using the High Content Screening System (Thermo Fisher Scienti c, USA), and the cell surface area from randomly selected elds ( fty for each group) was analyzed by the built-in image analysis software. Isolation of early endosomes by continuous density gradient centrifugation 52,53 The nuclear and cytoplasmic protein was successively extracted from NRCMs using a kit (SC-003, Inventbiotech, MN, USA). The postnuclear fraction was suspended in buffer, and was used to a continuous sucrose density gradient. After centrifugation (210,000 g for 3 h at 4°C), a milky band should be visible at each interface. 24 consecutive fractions were collected from each interface into tubes, and were subjected to western blot analysis for detection of protein EEA1, which is an early endosomal marker.
Low temperature SDS-PAGE, Western blot and co-immunoprecipitation (co-IP) analysis Low-temperature SDS-PAGE was conducted to investigate the homodimers (STAT3/STAT3) or heterodimer (STAT3/STAT1). In short, total protein from NRCMs were incubated in loading buffer (without 2-mercaptoethanol) for 5 min at 37°C, were separated by 8% SDS-PAGE for 4-5 h at a constant current of Page 15/28 40 mA, and were transferred to a PVDF membranes (Millipore). The whole process of the experiment must be made under the low temperature. After blocking with 10% blocking buffer (Beyotime, #P0023B), the membranes were incubated with individual antibodies at 4°C overnight 54 .

Immuno uorescence (IF) assay
NRCMs were cultured in chamber slides (ThermoFisher Scienti c). After treatment, cells were washed with ltered PBS for 3 times, xed with 4% paraformaldehyde for 10 min, permeabilized with 0.3% Triton X-100 for 5 min and followed by blocking with goat serum for 1 h at room temperature. Fluorescence emitted by uorescence-conjugated secondary antibodies (Proteintech, #SA00013-6 and #SA00013-8) at room temperature for 2 h. The slides were mounted with DAPI (CST, #4083) and were observed by a confocal microscope (Zeiss, Germany) or EVOS FL Auto (Life Technologies).
Total RNA isolation, cDNA Synthesis and real-time polymerase chain reaction (qPCR) Total RNA was extracted from snap-frozen cardiac tissues or NRCMs by using Trizol reagent (Invitrogen, #15596026), and its concentration was measured with a Nanodrop 2000 (Thermo Fisher Scienti c, USA).
The RNA extract (1000 ng) was reversely transcribed to rst strand cDNA using the One-step Reverse Transcription (RT) Kit (Thermo Fisher Scienti c, USA). Quantitative SYBR Green-based PCR (TOYOBO, Janpan) was conducted on the iCycler iQ system (Bio-Rad, USA). GAPDH was used as reference gene. All PCR assays were performed in triplicate. Data were analyzed using the 2 -ΔΔCT method. The oligonucleotide sequences are synthesized by Sangon (Shanghai, China), and listed in table S7.

Dual-luciferase reporter gene assay
The conserved DNA binding sequence of rats STAT3 (TTCCGGGAA) were subcloned into pGL3 Basic plasmid (Promega, #E1751) 55 . NRCMs were seeded at 5×10 4 cells per well into 96-well microplates, and were transiently co-transfected with the luciferase reporter (100 ng/well) and pRL-TK reporter constructs (Promega, E2241) at 20 ng/well. The total content of transfected DNA was normalized by empty vector. After the indicated treatments, the luciferase activity was determined by the dual-luciferase reporter assay system (Promega, #E1980) on a microplate reader (TECAN In nite M1000).

Statistical Analysis
Graph Pad Prism 6.0 (Graph Pad software) or SPSS Version 21 was used for statistical analysis. Normality of the obtained data was assessed using a Shapiro-Wilk test. When normality was con rmed, statistical differences among groups were analyzed using either Student's t test (for two groups) or one (or two) -way analysis of variance (ANOVA, for more than two groups). Otherwise, the non-parametric test Kruskal-Wallis test followed by the Dunn's post-hot test was used to correct for multiple comparisons. In all cases, differences were considered statistically signi cant at a P value less than 0.05. Figure 1 SNX3 expression was increased in failing human hearts, and cardiac-speci c Snx3-cKO mice were protected from ISO-induced cardiac injury. a and b A signi cant induction of SNX3 protein expression in end-stage failing human heart tissues (n = 9 per HF, n = 3 per normal control), which was identi ed by IF assay (Scale bar: 200 μm) and western blot analysis. Male C57BL/6 Snx3-cKO mice and their respective CTL at 11 weeks were injected with ISO (3 mg/kg/day, s.c.) or an equal volume of sterile normal saline for one week. c The protein expression of SNX3 was measured bys IF assay (Scale bar: 100 μm). d and e Gross observation of heart morphology, WGA staining (Scale bar: 50 μm) -stained cross sections of heart tissues were shown. f The HW/TL ratio was calculated. g Masson staining (Scale bar: 100 μm) -stained cross sections of heart tissues were shown. h The representative echocardiographic graphs were presented. i The echocardiographic parameters IVS were measured. Representative images of ve independent experiments were presented. The data were presented as the means ± SEM. *P < 0.05 vs. Normal or CTL + Saline group, #P < 0.05 vs. CTL + ISO group, n = 8-12. Abbreviations: CTL, littermate negative control; HF, heart failure; HW/TL, the heart weight to the tibia length ratio; IF, immuno uorescence; ISO, isoproterenol; IVS, interventricular septum; ns, no statistical difference; s.c., subcutaneously; Snx3-cKO, cardiac-speci c Snx3 knockout; WGA, wheat germ agglutinin; 1w, 1 week. See also Supplementary Figure S1-S3.

Figure 2
Cardiac-speci c Snx3 transgene leaded to myocardial injury in mice. We investigated the changes of cardiac structure and function in Snx3-cTg mice at 12 and 24 weeks. Besides, Snx3-cTg mice (at 11 weeks) were injected with 3 mg/kg/d ISO (s.c.) or an equal volume of sterile normal saline for one week.