Postnatal expression of cell cycle promoter Fam64a causes heart dysfunction by inhibiting cardiomyocyte differentiation through repression of Klf15

Summary Introduction of fetal cell cycle genes into damaged adult hearts has emerged as a promising strategy for stimulating proliferation and regeneration of postmitotic adult cardiomyocytes. We have recently identified Fam64a as a fetal-specific cell cycle promoter in cardiomyocytes. Here, we analyzed transgenic mice maintaining cardiomyocyte-specific postnatal expression of Fam64a when endogenous expression was abolished. Despite an enhancement of cardiomyocyte proliferation, these mice showed impaired cardiomyocyte differentiation during postnatal development, resulting in cardiac dysfunction in later life. Mechanistically, Fam64a inhibited cardiomyocyte differentiation by repressing Klf15, leading to the accumulation of undifferentiated cardiomyocytes. In contrast, introduction of Fam64a in differentiated adult wildtype hearts improved functional recovery upon injury with augmented cell cycle and no dedifferentiation in cardiomyocytes. These data demonstrate that Fam64a inhibits cardiomyocyte differentiation during early development, but does not induce de-differentiation in once differentiated cardiomyocytes, illustrating a promising potential of Fam64a as a cell cycle promoter to attain heart regeneration.


INTRODUCTION
The limited proliferation potential of adult cardiomyocytes (CMs) is a major obstacle hindering regeneration of myocardium lost following injury. Introducing fetal-specific signatures into damaged adult hearts is one of the promising strategies that could stimulate CM proliferation (Pö ling et al., 2012). This is because fetal CMs are highly proliferative and reveal a striking regenerative capacity following ablation of up to 60% of CMs (Sturzu et al., 2015). This strategy has recently been addressed with regard to various aspects of the fetal signatures, including cell cycle promoting genes (Mohamed et al., 2018), microRNAs (Borden et al., 2019), epigenetics , metabolic profiles (Honkoop et al., 2019), and hypoxic environments (Nakada et al., 2017).
We have recently identified family with sequence similarity 64, member A (Fam64a; also known as Pimreg, Cats, or Rcs1) as a fetal-specific cell cycle promoter in CMs . The strong nuclear expression of Fam64a in fetal CMs was almost completely lost in postnatal CMs from mice  and sheep (Locatelli et al., 2020). Fam64a knockdown inhibited and its overexpression enhanced fetal CM proliferation in vitro . This proliferation promoting function was also noted in various cancer cell lines (Jiang et al., 2019;Yao et al., 2019). In addition to its proliferation role, Fam64a also enhanced cell migration in several cell lines (Jiang et al., 2019;Yao et al., 2019), and this enhancement was coupled to the epithelial-to-mesenchymal transition (Yao et al., 2019;Zhang et al., 2019), a process that is closely related to cellular dedifferentiation. Fam64a expression was higher in cancer patients with metastasis than in non-metastasis patients (Wei et al., 2019). Fam64a also enhanced stemness features in breast cancer cells . These data prompted us to hypothesize the additional role of Fam64a in maintaining immature undifferentiated states in cells by promoting dedifferentiation or inhibiting differentiation. Thus, the aim of this study is (1) to explore the additional function of Fam64a in CMs, and (2) to test how these functions are implicated in cardiac regeneration.
Here, we analyzed transgenic (TG) mice expressing CM-specific Fam64a driven by alpha myosin heavy chain promoter. These mice maintained long-term Fam64a expression after birth when endogenous expression was abolished . Despite an enhancement of CM proliferation as expected, the TG mice showed impaired CM differentiation during postnatal development, resulting in cardiac dysfunction in later life characterized by increased expression of immature fetal markers and perturbation of the cardiac rhythm. Rhythmic activity of an organism is tightly coupled to cellular differentiation. The circadian clock is absent in undifferentiated cells, such as zygotes and early embryos, and is gradually established during differentiation (Umemura et al., 2017;Yagita et al., 2010). The established rhythmicity is abolished when differentiated cells are reprogrammed to regain pluripotency (Yagita et al., 2010). Thus, the rhythm disturbance and the impaired differentiation observed in Fam64a TG mice could be regulated by the common mechanisms.
We focused on Krü ppel-like factor 15 (Klf15) as a candidate molecule responsible for such mechanisms, because this transcription factor is reportedly involved both in cellular differentiation and the establishment of cardiac rhythmicity. Klf15 has been reported to promote differentiation in several cells including skeletal muscle cells (Dmitriev et al., 2011;Wu et al., 2014), adipocytes (Asada et al., 2011;Lee et al., 2016), and podocytes (Mallipattu et al., 2012). Klf15 is also a principal regulator that establishes cardiac rhythmicity (Jeyaraj et al., 2012;Zhang et al., 2015). A deficiency or excess of Klf15 perturbs rhythmic CM electrical activity and increases susceptibility to ventricular arrhythmias (Jeyaraj et al., 2012). It also controls other rhythmic biological processes, such as bile acid synthesis (Han et al., 2015). Klf15 inhibits pathological cardiac remodeling by repressing the process called a fetal gene program, which involves reactivation of immature fetal genes with an enhancement of CM proliferation (Cui et al., 2018;Fisch et al., 2007;Leenders et al., 2010Leenders et al., , 2012. In the present study, we demonstrate that Fam64a transcriptionally inhibits Klf15, thereby impairing CM differentiation during postnatal development, which leads to cardiac dysfunction coupled with rhythm disturbance in adult TG mice, despite an enhancement of CM proliferation. Thus, we propose a previously unknown function of Fam64a in inhibiting CM differentiation through repression of Klf15, in addition to the role as a cell cycle promoter. In contrast, introduction of Fam64a in differentiated adult wildtype (WT) hearts improved functional recovery upon injury with augmentation of the cell cycle and no apparent dedifferentiation in CMs. These data demonstrate that Fam64a inhibits CM differentiation during early development, but does not induce dedifferentiation in once differentiated adult CMs, which would contribute to the functional recovery upon injury, illustrating a promising potential of Fam64a as a cell cycle promoter to attain heart regeneration.

Enhanced CM proliferation in CM-specific Fam64a TG mice
We have established CM-specific Fam64a TG mice under the control of alpha myosin heavy chain promoter with a C-terminal FLAG tag (Figures S1A-S1C). Expressed protein was confirmed to localize in the CM nuclei, in the same location as an endogenous protein   (Figure S1D). We found that the cell cycle promoting genes were slightly, but consistently, upregulated in the hearts of TG mice compared to WT mice at the neonatal (postnatal day, P12-P15), adult (6-7 weeks), and aged (>25 weeks) stages ( Figures 1A-1C). We then assessed the CM cell cycle activity by staining for Ki67, a cell cycle marker, and phospho-histone H3 (pH3), a mitosis marker. This revealed that the numbers of both Ki67-positive and pH3-positive CMs were significantly increased at the neonatal stage in TG mice ( Figure 1D), but only Ki67positive CM were increased at the adult stage ( Figure 1E), and no increase occurred in either Ki67-positive or pH3-positive CM at the aged stage ( Figure 1F). In TG mice, the total CM count per ventricle was increased at 3 weeks ( Figure 1G). These data demonstrated an enhanced CM proliferation in TG mice at the neonatal and the juvenile stages, but not at the later stages. These in vivo results are in agreement with our previous in vitro analyses identifying Fam64a as a CM cell cycle promoter .

Fam64a TG mice unexpectedly show cardiac dysfunction with poor survival
Echocardiography demonstrated that TG mice showed progressive left ventricular dilation both at diastole and systole, leading to a severe decline in cardiac contractile function as estimated by fractional shortening when compared to WT mice (Figures 2A-2D). Histological assessment indicated that although there was no apparent difference at the neonatal stage, chamber dilation and wall thinning in left ventricle was progressively observed in TG mice in the adult and aged stages ( Figures 2E-2G iScience Article Figure 1. Enhanced CM proliferation in CM-specific Fam64a TG mice (A-C) qPCR analysis of cell cycle promoting genes in WT and TG mice hearts at neonatal (A, P12-P15), adult (B, 6-7 weeks), and aged (C, > 25 weeks) stages. Data were shown as normalized to WT. n = 3-12 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001 as compared to WT by Student's two-tailed unpaired t-test. Error bar = SEM. (D-F) Immunofluorescence for Ki67 and phospho-histone H3 (pH3) observed in sarcomeric a-actinin (as a CM marker) and DAPI in WT and TG mice heart sections at the neonatal (D, P12-P15), adult (E, 6-7 weeks), and aged (F, > 25 weeks) stages. Quantitative analysis for the percentage of Ki67-positive and pH3-positive CMs were shown. n= 3-4 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001 as compared to WT by Student's two-tailed unpaired t-test. Error bar = SEM. Scale bar = 10 mm. iScience Article marked drop in survival rate in TG mice ( Figure 2H). These data show that Fam64a TG mice develop agerelated cardiac dysfunction with poor survival despite their enhanced CM proliferation.
CM differentiation is impaired during postnatal development, leading to cardiac dysfunction in later life with increased expression of immature fetal markers in Fam64a TG mice qPCR analysis demonstrated that Ca 2+ handling genes that are important for mature differentiated CMs, including Ryr2, Cacna1c, and Atp2a2, were strongly downregulated in TG mice already at neonatal stages ( Figure 3A). Thyroid hormone receptor a (Thra), a receptor for thyroid hormone T3 that is a strong inducer of postnatal CM differentiation (Karbassi et al., 2020), was also downregulated ( Figure 3A). These reductions were similarly seen at adult and aged stages ( Figures 3B and 3C). Similar downregulations were observed in primary cultures of isolated CMs overexpressing Fam64a ( Figure S2). Moreover, genes encoding several K + channel subunits, which are involved in electrical activity in mature differentiated CMs (Karbassi et al., 2020), were consistently repressed ( Figure S3).
The Ca 2+ transient measurements in isolated CMs from aged mice (29-32 weeks) revealed a reduction in the peak amplitude and a delay in the time to peak, indicating impaired Ca 2+ mobilization in TG mice as compared to WT mice ( Figures 3D-3F). Although not statistically significant, a tendency was observed toward an increased time constant during the decay phase ( Figure 3G) and a decreased sarcoplasmic reticulum (SR) Ca 2+ content ( Figure 3H), suggesting impaired Ca 2+ re-uptake into the SR in TG mice. Cell shortening in response to electrical stimuli was decreased in TG mice at all the frequencies tested, indicating impairment of the CM contractile properties ( Figure 3I). Analyses using tissue sections ( Figure 3J) and isolated CMs ( Figure 3K) revealed disorganization of the sarcomere structures in TG mice. We also found a decreased CM cell size in TG mice, which is recognized as a less differentiated phenotype ( Figure 3L).
Collectively, these data show that in Fam64a TG mice, CM differentiation is impaired during postnatal development, leading to cardiac dysfunction in later life with increased expression of immature fetal markers.
Perturbed cardiac rhythmicity and locomotor activity in Fam64a TG mice Surprisingly, RNA-seq analysis of heart samples identified circadian rhythm as the most differentially altered pathway in TG mice, as indicated by an enrichment score far greater than other pathways like cardiac muscle contraction, hypertrophic cardiomyopathy, and dilated cardiomyopathy ( Figure 4A, see STAR Methods for details; data have been deposited in DDBJ sequencing read archive, DRA009818). Some of the principal genes relating to circadian rhythm, including Arntl (known as Bmal1), Cry1, Per2, Npas2, and Dbp, were dysregulated in TG hearts at both the mRNA ( Figure 4B) and the protein ( Figure S5) levels, although the changes were small. Telemetric measurements using freely moving conscious mice revealed perturbed heart rate regulation in TG mice: (1) Heart rate was consistently low in TG mice, irrespective of daytime or nighttime, throughout the course of measurements of up to 8 days ( Figure 4C).
(2) Nighttime-todaytime ratios of heart rate in TG mice were slightly, but significantly, lower than in WT mice, and had values of less than 1, indicating an abnormal daytime (inactive phase)-dominant heart rate regulation ( Figure 4D). In addition, TG mice frequently developed premature ventricular contraction, either as a single form or more hazardous serial forms, in sharp contrast to WT mice that displayed virtually no such arrhythmias ( Figures 4E and 4F). Decreased expression of connexin 43 ( Figure 3M) and K + channel genes ( Figure S3) may partially explain these aberrant phenotypes. (A-C) qPCR analysis of Ca 2+ handling genes important for mature differentiated CMs in WT and TG mice hearts at neonatal (A, P12-P15), adult (B, 6-7 weeks), and aged (C, > 25 weeks) stages. Data were shown as normalized to WT. n = 3-6 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001 as compared to WT by Student's two-tailed unpaired t-test. Error bar = SEM.
(D-I) Ca 2+ transients and cell shortening were measured in isolated CMs from WT and TG mice at aged stages (29-32 weeks). Representative Fura-2 ratio tracings of CMs (WT: blue, TG: red) stimulated at 0.5 Hz were shown as normalized to the peak value in WT set at 100% (D). Quantitative analysis for peak Ca 2+ amplitude (% normalized to WT) (E), time to peak (F), time constant (G), and sarcoplasmic reticulum (SR) Ca 2+ content (% normalized to WT) (H) were shown. Cell shortening (% of initial cell length) stimulated at indicated frequencies were shown in (I). WT: filled bar, TG: open bar. n = 9-28 CMs from 3 WT mice and 23-50 CMs from 2 TG mice. In (E-H), *p < 0.05 as compared to WT by Student's two-tailed unpaired t-test. In (I), *p < 0.05 and **p < 0.01 as compared to WT under the same stimulating frequency by Student's two-tailed unpaired t-test. Error bar = SEM. iScience Article Locomotor activity analysis of mice using the infrared motion detector also demonstrated perturbation of rhythmic behavior in TG mice: Whereas the nighttime (active phase)-dominant activity was observed in both WT and TG mice, the activity in TG mice was decreased during nighttime and increased during daytime when compared to WT mice ( Figure S6). The abnormal daytime-dominant heart rate regulation (Figure 4D) might be responsible for this phenotype. These data indicate that rhythmic CM electrical activity and locomotor activity were perturbed in Fam64a TG mice.

Impaired CM differentiation and enhanced CM proliferation observed in TG mice are mediated through transcriptional inhibition of Klf15 by Fam64a
Because Fam64a TG mice showed characteristic phenotypes such as the impaired CM differentiation coupled with rhythm disturbance, we focused on Klf15, a key transcription factor that is reportedly involved in these processes (Fisch et al., 2007;Jeyaraj et al., 2012;Leenders et al., 2010Leenders et al., , 2012Zhang et al., 2015). We found that mRNA expression of Klf15 and its downstream target Kv channel-interacting protein 2 (Kcnip2; also known as KChIP2) (Jeyaraj et al., 2012) was strongly upregulated during the course of differentiation in WT hearts but was severely depressed in Fam64a-overexpressing TG mice hearts, suggesting that Fam64a inhibits Klf15 expression at the transcriptional level ( Figure 5A). The extent of the inhibition was correlated with the expression level of Fam64a ( Figure S7). We conducted a comprehensive search for interacting partners of Fam64a that could mediate the inhibition of Klf15 by immunoprecipitation ( Figure S8A), followed by mass spectrometry (see STAR Methods for details; data have been deposited in ProteomeXchange Consortium via jPOST: PXD020570 and JPST000921.). This analysis led us to focus on glucocorticoid receptor (GR) ( Figure 5B), because it has previously been shown to bind to the promoter of Klf15 and stimulate its expression (Asada et al., 2011;Lee et al., 2016;Sasse et al., 2013).
Because Fam64a could be a putative transcriptional repressor (Archangelo et al., 2006(Archangelo et al., , 2013, we tested whether Fam64a inhibits GR-mediated transcriptional activation of Klf15 using luciferase reporter assay in HEK293T/17 cells. Three reporter constructs on the human KLF15 locus were used ( Figure 5C). Construct (i) contained a common promoter sequence upstream of the first exon. Construct (ii) contained three of the four GR binding sites that were previously reported (Asada et al., 2011;Sasse et al., 2013), whilst construct (iii) contained the fourth. At baseline, in the absence of exogenous induction of GR signaling, all three constructs showed a weak tendency toward a repressed activity because of Fam64a overexpression ( Figure 5D). The repression was most strongly and significantly observed in the construct (ii), which contains the majority of the GR binding sites ( Figures 5C and 5D). We observed a similar repression in the construct (ii) following exogenous induction of GR signaling by GR overexpression and dexamethasone treatment ( Figure 5E), although the repressive effect was weak. We corroborated these findings using primary cultures of isolated CMs to show that Fam64a repressed Klf15 mRNA expression in the absence or presence of exogenous GR induction by dexamethasone ( Figure 5F). Dexamethasone-induced activation of Klf15 was completely blocked by Fam64a. These data indicate that Fam64a inhibits Klf15 expression at least in part by GR-mediated transcriptional regulation through action on the previously described GR binding sites.
We next performed rescue experiments in vitro to assess whether forced expression of Klf15 restores the phenotypes induced by Fam64a overexpression by using primary cultures of isolated CMs. Overexpression of Fam64a promoted CM cell cycle progression, as shown by increased positivity for Ki67 and pH3 (Figure 5G). This treatment also inhibited CM differentiation, as shown by reduced expression of genes important for mature differentiated CMs ( Figure 5H), increased expression of immature fetal genes ( Figure S9), a decreased CM cell size ( Figure 5I), and frequent appearance of sarcomere disorganization ( Figure 5J). These data indicate that the phenotypes induced in Fam64a TG mice were recapitulated in this in vitro setting. All of these phenotypes were almost completely restored by concurrent expression of Klf15 ( Figures 5G-5J), suggesting that impaired CM differentiation and enhanced CM proliferation observed . Continued (J-K) Representative immunofluorescence images for sarcomeric a-actinin (red) and DAPI (blue) in longitudinal heart sections (J) and isolated CMs (K) from WT and TG mice at > 25 weeks. In WT mice, highly organized sarcomere structure was observed. In contrast, disorganization of sarcomeres was frequently observed in TG mice. Scale bar = 20 mm (J) and 50 mm (K).
(L) Representative images of freshly isolated CMs from WT and TG mice at > 25 weeks aged stages, obtained by differential interference contrast optics. CM cell size was evaluated as a two-dimensional projected area. n = 75 CMs from 3 WT mice and 113 CMs from 2 TG mice. **p < 0.01 as compared to WT by Student's two-tailed unpaired t-test. Error bar = SEM. Scale bar = 50 mm.
(M) qPCR analysis of immature fetal genes, and a mature gap junction component connexin 43 (Gja1) in WT and TG mice hearts. Data were shown as normalized to WT. n = 5-11 mice per group. *p < 0.05 and **p < 0.01 as compared to WT by Student's two-tailed unpaired t-test. Error bar = SEM. Based on the findings that cardiac dysfunction in Fam64a TG mice was attributable to impaired CM differentiation early during postnatal development, we next tested whether introduction of Fam64a in differentiated adult WT hearts can circumvent this issue and provide benefits in cardiac regeneration. We used the protocol with direct intramyocardial injection of mRNA encoding Fam64a-FLAG or control EGFP immediately after cryoinjury in WT adult hearts (modified from Kaur and Zangi, 2020;Strungs et al., 2013). The mRNA injection strategy has been reported to achieve a rapid and transient expression of the transgene, which peaks at 24 h and then gradually declines at around $10 days. This could avoid the undesired consequences associated with prolonged transgene activation. The expression tends to be spatially confined around the injection/injury site, thereby enabling targeting of the injury site without causing detrimental effects in the remote region. Time-course analysis at the injury border and the remote region at 1 day, 1 week, and 3 weeks after injury revealed expression of the Fam64a-FLAG protein in $50% of cardiomyocytes at the border region at 1 day ( Figure 6A). As expected, a positive signal was not detected at other timepoints or locations. The estimation by qPCR analysis revealed a 765 G 251-fold (mean G SEM, n = 3 mice) increase in the expression of the delivered mRNA in heart samples in comparison to the non-injected controls. Although cardiac contractile function was seriously damaged in both groups immediately after cryoinjury, the mice receiving Fam64a-FLAG mRNA showed progressively improved functional recovery with minimum left ventricular dilation over a follow-up period of 5 weeks in comparison to those receiving control EGFP mRNA ( Figures 6B-6E). Histological evaluation by Masson's trichrome staining revealed less fibrosis in Fam64a-FLAG group at 3 weeks after injury ( Figure 6F). The assessment of heart sections containing a cryoinjured area at 5 weeks after injury demonstrated greater cell cycle activity in the Fam64a-FLAG group than the EGFP group, as indicated by increased positivity for Ki67 and pH3 ( Figure 6G). Meanwhile, Klf15 expression was slightly decreased both at the mRNA ( Figure 6H) and the protein ( Figure 6I) levels. Genes important for mature differentiated CMs, which were strongly repressed in TG mice (Figures 3A-C and S3), did not change or only marginally decreased ( Figures 6J and 6K). Likewise, immature fetal genes, which were strongly increased in TG mice ( Figure 3M), did not change or only marginally increased ( Figure 6L). None of the changes in these genes reached statistically significance. Consequently, disorganization of the sarcomere structures, which was frequently observed in TG mice ( Figures 3J and 3K), was not observed ( Figure 6M). These data demonstrate that Fam64a inhibits CM differentiation during early development, but does not induce dedifferentiation in once differentiated adult CMs, which would contribute to the functional heart recovery upon injury with augmented CM cell cycle.

DISCUSSION
Current views on the function of Fam64a point to its role as a cell cycle promoter in fetal CMs  and in various cancer cells (Yamada et al., 2018;Yao et al., 2019). Previous work has led us to hypothesize the additional role of Fam64a in promoting dedifferentiation or inhibiting differentiation to maintain undifferentiated states in cells (Yao et al., 2019;Zhang et al., 2019).
In the TG mice maintaining CM-specific postnatal expression of Fam64a, we saw impaired CM differentiation during postnatal development, resulting in cardiac dysfunction in later life characterized by increased expression of immature fetal markers and perturbation of the cardiac rhythm, despite an enhancement of CM proliferation. Mechanistic analysis and rescue experiments revealed that these phenotypes were mediated through transcriptional inhibition of Klf15 by Fam64a ( Figure 5). All of the phenotypes induced by Fam64a overexpression were almost completely restored by concurrent expression of Klf15. A previous study found that Klf15-deficient mice showed perturbed CM rhythmic activity and were susceptible to ventricular arrhythmias, similarly to the effects seen in Fam64a TG mice ( Figures 4C-4F), which were considered to reflect suppressed KChIP2 activity (A) qPCR analysis of Klf15 and Kcnip2 (KChIP2) at fetal, neonatal, adult, and aged stages from WT (circle) and TG (triangle) mice hearts. Data were shown as normalized to WT at fetal stage set at 1. In WT mice, Klf15 expression was significantly increased at 6-13W and afterward as compared to fetal stage (E15-E18) (One-way ANOVA with Tukey's post hoc test). Likewise, Kcnip2 expression was significantly increased at P12-P24 and afterward as compared to fetal stage (E15-E18) (One-way ANOVA with Tukey's post hoc test). By contrast in TG mice, the expressions of both genes were significantly attenuated at all stage as compared to WT mice of the same age (*p < 0.05, **p < 0.01, and ***p < 0.001 as compared to WT by Student's two-tailed unpaired t-test.). n = 3-8 mice per group. Error bar = SEM. (B) Immunoprecipitation (IP) against FLAG peptide that was expressed as a C-terminal tag of overexpressing Fam64a protein in TG mice hearts, followed by western blotting (WB) using glucocorticoid receptor (GR) antibody, which detected GR protein in TG, but not in WT mice heart lysates. This indicates that ll OPEN ACCESS iScience Article (Jeyaraj et al., 2012). KChIP2 is a critical subunit for generating the fast transient outward K + current (I to,f ) in the early repolarization phase (Kuo et al., 2001), and it augments subsequent Ca 2+ influx in CMs (Cordeiro et al., 2012;Thomsen et al., 2009). We found severely depressed transcript levels of KChIP2 ( Figure 5A) and impaired Ca 2+ transients ( Figures 3D-3H) in TG mice. Suppression of K + channel genes ( Figure S3) would also account for the aberrant CM electrical activity. These data suggest that postnatal expression of Fam64a inhibited CM differentiation ( Figure 3) through inhibition of Klf15-KChIP2 axis ( Figure 5), thereby disrupting CM rhythmic activity (Figure 4) and contributing to cardiac dysfunction (Figure 2), despite an enhancement of CM proliferation ( Figure 1). Thus, we propose that Fam64a is not merely a cell cycle promoter; rather, it has an additional role in inhibiting CM differentiation through repression of Klf15. Whether this function of Fam64a is active during fetal development, when endogenous Fam64a is abundantly expressed, needs to be tested.
We demonstrated that GR could, at least in part, mediate the inhibitory effect of Fam64a on Klf15 at the transcriptional level ( Figures 5B-5F), although the effect was rather weak. We identified tripartite motif-containing 28 (Trim28) as another interacting partner of Fam64a in CMs ( Figure S8B). Trim28 is a known coactivator of GR (Chang et al., 1998). Moreover, the nucleosome remodeling and deacetylase (NuRD) complex, which is implicated in gene repression (Denslow and Wade, 2007), interacts with both Fam64a (Zhao et al., 2008) and Trim28 (Schultz et al., 2001). Therefore, an important remaining challenge is to clarify the mechanism of how protein complexes comprising Fam64a, GR, and Trim28, coupled with the NuRD complex, cooperatively repress Klf15 transcription.
The molecular link between Fam64a and Klf15 found in this study provides a hint at a mechanism by which Fam64a promotes CM proliferation. Multiple lines of evidence point to cell cycle inhibitory action of Klf15 in a variety of non-CM cell types through the regulation of cyclins, cyclin-dependent kinases, cyclin-dependent kinase inhibitors, and DNA synthesis regulators (Hong et al., 2012;Ray and Pollard, 2012;Yoda et al., 2015). Thus, it is possible that Fam64a promotes cell proliferation by relieving this inhibitory action of Klf15 in CMs ( Figure 5G).
Recently, Fam64a has been reported to interact with Stat3, and to stimulate its transcriptional activity during colitis-associated carcinogenesis . Interestingly, like Fam64a, Stat3 induces CM proliferation during cardiac regeneration (Nakao et al., 2020), and has been identified as a factor to acquire Figure 5. Continued Fam64a interacts with GR in CMs. Western blotting using FLAG antibody correctly detected Fam64a-FLAG fusion protein (*) in TG, but not in WT mice heart lysates, validating the immunoprecipitation procedure. Three to four mice at > 17 weeks were mixed and used for protein extraction in each genotype.
(C) Three reporter constructs on human KLF15 locus were used in luciferase reporter assay. Construct (i) contains common promoter sequence upstream of the first exon. Construct (ii) contains three (marked as a-c) of the four GR binding sites previously reported, whilst construct (iii) contains the fourth (marked as d). Ex = exon. (D) HEK293T/17 cells were transiently transfected with Fam64a expression vector (Fam) or control empty vector (Vec), and one of the three reporter constructs (ⅰ-ⅲ). The luciferase activity of each reporter construct was normalized to that of the control reporter construct, and was expressed as the activity of Vec set at 1. n = 7 independent experiments. *p < 0.05 as compared to Vec by Student's two-tailed unpaired t-test. Error bar = SEM. (E) HEK293T/17 cells were transiently transfected with Fam64a expression vector (Fam), GR expression vector (GR), or control empty vector (Vec), and one of the three reporter constructs (ⅰ-ⅲ). Cells were treated with dexamethasone (Dex) at 1 mM for 24 h. Luciferase activity of each reporter construct was normalized to that of control reporter construct, and was expressed as the activity of Vec set at 1. n = 4 independent experiments. **p < 0.01, ***p < 0.001 between the indicated groups by One-way ANOVA with Tukey's post hoc test. Error bar = SEM. (F) Primary CMs were isolated from fetal hearts and transduced with baculovirus expressing Fam64a (Fam) or control empty vector (Vec) in the absence or the presence of dexamethasone (Dex) treatment at 1 mM for 24 h. Total RNA was extracted, reverse-transcribed, and subjected to qPCR analysis for Klf15. Data were expressed as Klf15 mRNA expression of the vector (Vec) group set at 1. n = 3 independent experiments. In each experiment, 5-10 fetal hearts were pooled and used for the isolation of CMs. *p < 0.05, and **p < 0.01 between the indicated groups by Student's two-tailed unpaired t-test. Error bar = SEM. iScience Article undifferentiated pluripotent state in various stem cells (Chen et al., 2015;Nakao et al., 2020). Trim28, an interacting partner of Fam64a ( Figure S8B), is recruited to Stat3 target genes to mediate epigenetic activation (Jiang et al., 2018). Thus, it will be intriguing to examine the functional crosstalk between Stat3 pathway and Fam64a-Klf15 axis in CMs. Interestingly, Stat3 transcripts were slightly upregulated in Fam64a TG mice ( Figure S10), suggesting an additional layer of regulation of Stat3 activation by Fam64a at the transcriptional level.
In Fam64a TG mice, the impairment of CM differentiation early during development exacerbated cardiac function in later life, despite an enhancement of CM proliferation. In contrast, introduction of Fam64a in differentiated adult WT hearts improved functional recovery upon injury with augmentation of the cell cycle and no apparent dedifferentiation in CMs ( Figure 6). These data indicate that Fam64a inhibits CM differentiation during early development, but does not induce dedifferentiation in once differentiated adult CMs. This will make Fam64a a promising candidate as a cell cycle promoter to attain heart regeneration, because several studies have pointed out that excessive CM dedifferentiation evoked by persistent induction of cell cycle stimulants caused cardiac dysfunction (Gabisonia et al., 2019;Ikeda et al., 2019;Kubin et al., 2011), which in some cases could be overcome by an approach of transient induction (D'Uva et al., 2015;Tian et al., 2015). Thus, it is important to optimize an induction protocol for Fam64a, e.g., the intensity and duration that fine-tunes the balance between CM proliferation and dedifferentiation in response to various types of cardiac injury with varying degree of severity. Simultaneous activation of anaphase-promoting complex/cyclosome (APC/C), which targets Fam64a for degradation during each cell cycle (Zhao et al., 2008), will provide another effective means to control the activity of Fam64a.
In summary, this work adds important insights into our understanding of the role of Fam64a. We propose a previously unknown function of Fam64a in inhibiting CM differentiation through repression of Klf15, in addition to the role as a cell cycle promoter. By taking advantage of the feature of Fam64a that does not induce dedifferentiation in adult CMs, future research should aim to identify an optimized induction protocol for Fam64a in injured adult hearts, which will ultimately contribute to development of regenerative therapies of the human heart.

Limitations of the study
In this study, four limitations should be considered. One was that CMs were generally assumed to require dedifferentiation in order to proliferate; therefore, how CMs activated the cell cycle after cryoinjury, without signs of dedifferentiation such as the changes in Klf15 and related genes, is unclear ( Figure 6). However, the concept of CM differentiation/dedifferentiation includes a diverse range of biological processes, such as changes in cell size, T-tubule/sarcomere organization, Ca 2+ handling, cellular metabolism, and gene iScience Article expression of mature/immature markers. Of these, the specific condition(s) for promoting dedifferentiation that must be activated to proceed to proliferation is currently uncertain. Therefore, dedifferentiation may have occurred but was not detected by our analyses. Alternatively, some as yet unknown mechanism might allow progression to proliferation by bypassing the dedifferentiation step. A second limitation was that the mechanisms of how the CM cell cycle remained activated at 5 weeks after cryoinjury ( Figure 6G) is unclear and needs to be clarified, because the expression of external Fam64a ceased at 1 week after cryoinjury (Figure 6A). A third limitation was that, although enhanced CM proliferation was clearly shown in TG mice at the neonatal and the juvenile stage (Figure 1), the changes in the cell cycle genes were minor, and RNA for the qPCR analysis was extracted from the whole ventricle, which included non-CMs ( Figures 1A-1C). Therefore, this limitation needs to be considered in data interpretation. A fourth limitation was that, although the inhibitory effect of Fam64a on Klf15 was clearly demonstrated (Figure 5), the effect was rather weak, especially for the evaluation of reporter repression ( Figure 5E). Therefore, the biological significance should be carefully interpreted.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We are grateful to Nobuhisa Iwachido (Kawasaki Medical School, Japan) for expert technical assistance in H&E staining. This work was supported by JSPS KAKENHI Grant Number 17H02092, 18K19943, 17H06272, and 20H04521 to K.H., A.H., Y.Ujihara., and S.M. from the Ministry of Education, Culture, Sports, Science, and Technology of Japan, and was also supported by Research Project Grants from Kawasaki Medical School. Parts of the Graphical Abstract were drawn by using free images from Servier Medical Art at http://smart.servier.com/.

Mice
Mice with a C57BL/6N background were housed in a temperature-controlled room under a 12-h light:12-h dark cycle conditions and were fed a standard chow diet and water ad libitum. CM-specific Fam64a TG mice were generated as follows: the murine Fam64a sequence with a C terminal FLAG tag was cloned downstream of the alpha myosin heavy chain promoter ( Figure S1). The alpha myosin heavy chain/puro rex/neo was a gift from Mark Mercola (Addgene plasmid #21230; http://n2t.net/addgene:21230) (Kita- Matsuo et al., 2009). This transgene construct was purified, linearized, and injected into fertilized oocytes from C57BL/6N background mice (Transgenic Inc, Japan). The resulting pups were genotyped by PCR using genomic tail DNA and seven founder lines were established. Among these, two lines expressing a sufficient amount of the transgene both at mRNA and protein levels were selected and used for subsequent experiments ( Figure S1). Wildtype (WT) mice with the same background were used for comparison. This study was performed in strict accordance with the recommendations of the Institutional Animal Care and Use Committee at the Kawasaki Medical School. All of the animals were handled according to approved institutional protocols of the Kawasaki Medical School, and every effort was made to minimize suffering. All experiments were performed in accordance with the relevant guidelines and regulations of the Kawasaki Medical School.

Immunoprecipitation and mass spectrometry
We identified the interacting partners of Fam64a using immunoprecipitation against the FLAG peptide that was expressed as a C-terminal tag of the overexpressed Fam64a protein in TG mice hearts, followed by mass spectrometry analysis (n = 2 biological replicates). Immunoprecipitates from WT mice hearts were used as a negative control. Heart tissues were freshly isolated from WT and TG mice, minced, and homogenized using a Kinematica Polytron homogenizer (Fisher Scientific) in IP lysis buffer (Thermo-Fisher) or cytoplasmic extraction reagent I & II (Thermo-Fisher) in the presence of a protease inhibitor cocktail (Thermo-Fisher). After centrifugation and protein quantification, the lysates were subjected to immunoprecipitation using the EZview Red Anti-FLAG M2 affinity gel system (F2426, Sigma-Aldrich) according to the manufacturer's instructions. Elution of the immunoprecipitates was performed with 33 FLAG peptide (F4799, Sigma-Aldrich). The immunoprecipitation procedure was validated by western blotting using FLAG and Fam64a antibodies, which correctly detected the Fam64a-FLAG fusion protein in TG, but not in WT, mouse heart lysates ( Figure S8A).
LC-MS/MS ANALYSIS-In-solution digestion and nano flow-liquid chromatography tandem mass spectrometry were performed (Oya et al., 2019), with some modifications. In brief, the eluted proteins were digested with 10 mg/mL modified trypsin (Sequencing grade, Promega, USA) at 37 C for 16 h. The digested peptides were desalted with in-house made C18 Stage-tips, dried under a vacuum, and dissolved in 2% acetonitrile containing 0.1% trifluoroacetic acid. The peptide mixtures were then fractionated by C18 reverse-phase chromatography (3 mm, ID 0.0753 150 mm, CERI). The peptides were eluted at a flow rate of 300 nL/min with a linear gradient of 5-35% solvent B over 90 min.
DATABASE SEARCHING-The raw files were searched against the Mus musculus dataset (Uniprot Proteome ID UP000000589 2019.06.11 downloaded, 55,197 sequences; 22,986,518 residues) combined with the FLAG-tagged Fam64a sequence and the common Repository of Adventitious Proteins (cRAP, ftp:// ftp.thegpm.org/fasta/cRAP) using MASCOT version 2.6 (Matrix Science) via Proteome discoverer 2.2 (Thermo-Fisher), with a false discovery rate (FDR) set at 0.01. Carbamidomethylation of cysteine was set as a fixed modification. Oxidation of methionine and acetylation of protein N-termini were set as variable modifications. The number of missed cleavage sites was set as 2.
CRITERIA FOR PROTEIN IDENTIFICATION-Scaffold (version Scaffold_4.10.0, Proteome Software Inc., USA) was used to validate the MS/MS-based peptide and protein identifications. Peptide identifications were accepted if they exceeded specific database search engine thresholds. Protein identifications were accepted if they contained at least two identified peptides. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters.
In two biologically independent experiments, a total of 440 proteins were detected under the threshold setting in Scaffold software, as follows: protein threshold of 1.0% FDR, peptide threshold of 0.1% FDR, and Min # peptides = 5. Proteins detected only in TG samples, but not in WT samples, were considered as candidate interacting partners of Fam64a, and the interaction of those proteins with Fam64a in heart tissues was subsequently tested by immunoprecipitation and western blotting using specific antibodies. All mass spectrometry data have been deposited in the ProteomeXchange Consortium via jPOST, with the dataset identifiers PXD020570 and JPST000921. Heart tissues were collected from mice, cut into small pieces, and immediately immersed in RNAlater Stabilization Reagent (Qiagen, Germany). The stabilized tissues were homogenized with a Kinematica Polytron homogenizer (Fisher Scientific), and total RNA was isolated using the ISOGEN or ISOGEN-II systems (Nippon Gene, Japan). For cultured CMs, harvested cell pellets were processed similarly to heart tissues but without the use of the homogenizer. After assessing RNA yield and quality using a NanoDrop One spectrophotometer (Thermo-Fisher), the RNA samples were reverse-transcribed with PrimeScrip RT Master Mix (TaKaRa Bio, Japan), and quantitative real-time PCR was performed using TaqMan Fast Advanced Master Mix in a StepOnePlus real-time PCR system (Applied Biosystems, USA). Quantification of each mRNA was carried out with Actb or Ubc as reference genes, using the DDC T method .

Luciferase reporter assay
Three reporter constructs spanning the promoter region of human KLF15 locus were used ( Figure 5C): construct (i):À694/+228, construct (ii): +1066/+1965, and construct (iii): +9444/+10,643, where the number indicates the genomic position relative to the transcription start site. The sequence of the construct (i) was derived from the LightSwitch Promoter Reporter GoClone (SwitchGear Genomics, USA). The expression vectors were a Fam64a expression vector, GR expression vector, or control empty vector (pFastBac1-VSVG-CMV-WPRE; Hashimoto et al., 2017). HEK293T/17 cells (ATCC CRL-11268) were maintained in DMEM with 5% FBS under standard conditions at 37 C with 5% CO 2 . Cells were plated onto 96-well plates coated with fibronectin, and transient transfection of the reporter construct and the expression vector was carried out using Lipofectamine 2000 (Thermo-Fisher) on the following day. The amount of plasmid used per well was 3 ng for each expression vector/50 ng for each reporter construct. The control empty vector was used to equalize the total amount of DNA for each transfection. The expression of the Fam64a protein was confirmed by western blotting. Cells were treated with dexamethasone (Dex) at 1 mM for 24 h. Luciferase activity was measured on the next day using the LightSwitch luciferase assay system (SwitchGear Genomics) as per the manufacturer's protocol . The luciferase activity of each reporter construct was normalized to that of the control reporter construct (pLightSwitch_Prom) and was expressed as the activity of the control empty vector set at 1.

Histology
Heart tissues were collected from mice, fixed in 4% paraformaldehyde, embedded in paraffin, and vertically sectioned at a thickness of 3 mm. Hematoxylin-eosin (H&E) staining was performed according to standard procedures. Stained sections were observed with a light microscope (BZ-X710, Keyence, Japan).

Echocardiography
Two-dimensional transthoracic echocardiography was performed to evaluate cardiac function using an Aplio 300 system with a 14-MHz transducer (Toshiba Medical System, Japan) (Ujihara et al., 2016). M-mode tracings were used to measure the left ventricular internal diameter at end diastole (LVDd) and end systole (LVDs). Fractional shortening (FS) was calculated as ([LVDd-LVDs]/LVDd)3100 (%), and was used as an index of cardiac contractile function. All examinations were performed on conscious mice to prevent anaesthesia-related impairment of cardiac function. In these non-sedated mice, an FS <65% was considered indicative of the impaired cardiac function (Ivandic et al., 2012).

Locomotor activity measurement
The locomotor activity of mice was monitored using an infrared motion detector (Actimo-100, Shinfactory, Japan), which consists of a free moving space (30 3 20 cm 2 ) with a side wall equipped with photosensors at 2-cm intervals to scan animal movement (Kurokawa et al., 2011). Activity counts accumulated over a 1-h period were measured for a total of 4 days in a 12-h light:12-h dark cycle (lights on at 8 a.m.). Total activity counts during the daytime (8 a.m.-8 p.m.) and nighttime (8 p.m.-8 a.m.) were considered to reflect the locomotor activity in each phase. During the nighttime, we found that the most mice showed characteristic biphasic patterns of locomotor activity, i.e., the first peak during the time period from 8 p.m. to 2 a.m., and the second peak during the time period from 2 a.m. to 8 a.m. (typical example shown in Figure S6A). Thus, the peak activity counts in each phase were used as a measure of the locomotor activity during nighttime.