RBM4a-SRSF3-MAP4K4 Splicing Cascade Constitutes a Molecular Mechanism for Regulating Brown Adipogenesis

An increase in mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) reportedly attenuates insulin-mediated signaling which participates in the development of brown adipose tissues (BATs). Nevertheless, the effect of MAP4K4 on brown adipogenesis remains largely uncharacterized. In this study, results of a transcriptome analysis (also referred as RNA-sequencing) showed differential expressions of MAP4K4 or SRSF3 transcripts isolated from distinct stages of embryonic BATs. The discriminative splicing profiles of MAP4K4 or SRSF3 were noted as well in brown adipocytes (BAs) with RNA-binding motif protein 4-knockout (RBM4−/−) compared to the wild-type counterparts. Moreover, the relatively high expressions of authentic SRSF3 transcripts encoding the splicing factor functioned as a novel regulator toward MAP4K4 splicing during brown adipogenesis. The presence of alternatively spliced MAP4K4 variants exerted differential effects on the phosphorylation of c-Jun N-terminal protein kinase (JNK) which was correlated with the differentiation or metabolic signature of BAs. Collectively, the RBM4-SRSF3-MAP4K4 splicing cascade constitutes a novel molecular mechanism in manipulating the development of BAs through related signaling pathways.


Introduction
Adipose tissues are known to preserve food-derived lipids as excess energy and function as an endocrine organ for energy homeostasis by secreting related hormones [1]. Dysfunction of the physiological roles of adipose tissues occurs with an expanded mass of white adipocytes (WAs) that subsequently induce human obesity [2]. In contrast, brown adipocytes (BAs) were discovered to execute non-shivering thermogenesis by dissipating stored lipid droplets in small rodents and infants [3]. Several studies revealed the impact of cold exposure and exercise on inducing white-to-brown shift within white adipose tissues (WATs) [4,5]. Therefore, BAs are considered a potential target for combating human obesity and related diseases [6]. Lineage tracing analyses indicated the existence of Myf5-positive progenitors for the development of myocytes and BAs [7]. Regulatory factors that participate in the differentiation of distinct adipocytes were uncovered using transcriptome analyses, In our previous study, results of transcriptome analyses disclosed discriminative splicing profiles of 186 genes from a total of 13,217 splicing events in postnatal RBM4a −/− BATs compared to wild-type (WT) BATs [15]. Among those candidates, the splicing profiles of mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) and SRSF3 transcripts were altered after ablation of endogenous RBM4a in postnatal BATs. As shown in Table 1, MAP4K4 isoforms 3 and 4 are major transcripts expressed by postnatal BATs (Table 1A; fragments per kilobase of transcript per million mapped reads (FPKM) = 15.15 and 14.76, respectively), whereas MAP4K4 isoforms (Iso) 1 and 3 were mainly transcribed in embryonic BATs (Table 1A; FPKM = 42.23 and 32.43, respectively). In contrast, depletion of RBM4a mediated a drastic shift in MAP4K4 Iso 3 to Iso 2 (Table 1A, FPKM = 20.32) in embryonic BATs, and reduced expression of MAP4K4 Iso 4 was noted in postnatal RBM4a −/− BATs (Table 1A, FPKM = 5.64) compared to the WT counterparts. Relatively high expressions of total MAP4K4 transcripts were observed in embryonic and postnatal WT BATs (Table 1A, lower, FPKM = 32.8794 and 71.2371, respectively) compared to RBM4a −/− BATs (Table 1A, lower, FPKM = 17.3019 and 30.2237, respectively). Moreover, results of transcriptome assays showed an increase in the relative level of the non-coding SRSF3 transcript (NR_036613) in postnatal WT BATs compared to that of embryonic WT BATs (Table 1B, PII = 53.68% and 13.91%, respectively), whereas expressions of the coding SRSF3 transcript (NM_013663) was elevated in embryonic and postnatal RBM4a −/− BATs (Table 1B, FPKM = 213.5991 and 54.33154, respectively) compared to those of the WT counterparts (Table 1B, FPKM = 108.2105 and 10.98038, respectively). These results disclosed the reprogrammed expression and splicing profiles of MAP4K4 and SRSF3 throughout brown adipogenesis, which is relevant to the abundance of RBM4a. Table 1. Distinct splicing and expression profiles of the mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) and serine/arginine-rich splicing factor 3 (SRSF3) genes were characterized throughout development of brown adipose tissues (BATs). Results of transcriptome analyses were analyzed using the CLC genomics workbench (CLC, GWB) (n = 4). (A) Differential splicing and expression profiles of the MAP4K4 gene were identified in embryonic (E13.5) and postnatal (P0) BATs dissected from wild-type (WT) and RNA-binding motif protein 4a-depleted (RBM4a −/− ) mice. (B) Discriminative splicing and expression profiles of the SRSF3 gene were identified by performing parallel experiments as described in (A). A percentage isoform index (PII) was applied to describe relative levels of indicated transcripts. The reviewed sequences indicated that the MAP4K4 gene generated four transcripts by alternatively selecting exons 16, 17, 21, and 24 ( Figure 1A, upper). Retention of SRSF3 intron 3 (also referred to as exon 4 ) resulted in the generation of premature termination codon-containing transcripts which were considered to be candidates of the nonsense-mediated decay pathway ( Figure 1A, lower) [16]. As shown in Figure 1B, results of the RT-PCR showed that the majority of MAP4K4 transcripts were Iso 1 and 2 generated in embryonic (E)13.5 and E15.5 BATs ( Figure 1B Figure 1B, lanes 10-12). The reprogramming of MAP4K4 and SRSF3 transcripts was consistently represented using an in vitro cell model. As shown in Figure 1C, splicing profiles of the MAP4K4 and SRSF3 transcripts respectively shifted to MAP4K4 Iso 4 and SRSF3 +ex4 transcripts in the presence of the differentiating cocktail ( Figure 1C, lanes 2 and 6) compared to those of C3H10T1/2 cells cultured in growth medium ( Figure 1C, lanes 1 and 5). In contrast, splicing profiles of MAP4K4 and SRSF3 transcripts in RBM4a-depleted cells showed no response to the presence of the differentiating cocktail ( Figure 1C, lanes 4 and 8) compared to those of non-differentiated cells ( Figure 1C, lanes 3 presence of the differentiating cocktail ( Figure 1C, lanes 2 and 6) compared to those of C3H10T1/2 cells cultured in growth medium ( Figure 1C, lanes 1 and 5). In contrast, splicing profiles of MAP4K4 and SRSF3 transcripts in RBM4a-depleted cells showed no response to the presence of the differentiating cocktail ( Figure 1C, lanes 4 and 8) compared to those of non-differentiated cells ( Figure  1C, lanes 3 and 7). These results suggested a potential link between RBM4a abundance and splicing profiles of the MAP4K4 and SRSF3 transcripts throughout brown adipogenesis. (B) Total RNAs prepared from embryonic (E13.5 and E15.5) and postnatal (P0) brown adipose tissues dissected from wild-type (WT) and RBM4a −/− mice (n = 4) were subjected to RT-PCR assays using specific primer sets as listed in Supplementary Table S1. The bar graph presents relative levels of the MAP4K4 Iso 4 and SRSF3 +ex4′ transcripts. (C) C3H1oT1/2 cells were respectively transfected with empty vector or the RBM4a targeting vector (sh-RBM4a), followed by culturing in growth medium and differentiating medium for 48 h. Total RNAs extracted from non-differentiating (0 day) and differentiating cells (2 day) were subjected to RT-PCR assays using indicated primer sets (n = 4). The bar graph shows relative levels of MAP4K4 Iso 4 and SRSF3 +ex4′ transcripts. Signal densities of the RT-PCR products were quantified using TotalLab Quant Software. Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference.

RBM4a Exerts a Specific Influence on Modulating Splicing Profiles of MAP4K4 and SRSF3 Transcripts
Results of Spearman's correlation test verified respective coefficients of 0.68 and 0.63 by comparing increases in relative levels of MAP4K4 Iso 4 and SRSF3 +ex4′ transcripts and the upregulated RBM4a protein throughout the development of WT BATs (n = 4; p = 0.0207 and 0.0334, respectively). An in vitro cell model was next applied to validate impacts of RBM4a on splicing profiles of endogenous MAP4K4 and SRSF3. Results of RT-PCR assays showed that overexpression of RBM4a led to increases in the relative levels of MAP4K4 Iso 4 and SRSF3 +ex4′ (Figure 2A  (B) Total RNAs prepared from embryonic (E13.5 and E15.5) and postnatal (P0) brown adipose tissues dissected from wild-type (WT) and RBM4a −/− mice (n = 4) were subjected to RT-PCR assays using specific primer sets as listed in Supplementary Table S1. The bar graph presents relative levels of the MAP4K4 Iso 4 and SRSF3 +ex4 transcripts. (C) C3H1oT1/2 cells were respectively transfected with empty vector or the RBM4a targeting vector (sh-RBM4a), followed by culturing in growth medium and differentiating medium for 48 h. Total RNAs extracted from non-differentiating (0 day) and differentiating cells (2 day) were subjected to RT-PCR assays using indicated primer sets (n = 4). The bar graph shows relative levels of MAP4K4 Iso 4 and SRSF3 +ex4 transcripts. Signal densities of the RT-PCR products were quantified using TotalLab Quant Software. Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference.

RBM4a Exerts a Specific Influence on Modulating Splicing Profiles of MAP4K4 and SRSF3 Transcripts
Results of Spearman's correlation test verified respective coefficients of 0.68 and 0.63 by comparing increases in relative levels of MAP4K4 Iso 4 and SRSF3 +ex4 transcripts and the upregulated RBM4a protein throughout the development of WT BATs (n = 4; p = 0.0207 and 0.0334, respectively). An in vitro cell model was next applied to validate impacts of RBM4a on splicing profiles of endogenous MAP4K4 and SRSF3. Results of RT-PCR assays showed that overexpression of RBM4a led to increases in the relative levels of MAP4K4 Iso 4 and SRSF3 +ex4 (Figure 2A  RNAs and cell lysates were extracted from C3H10T1/2 cells respectively transfected with an SRSF3targeting vector, or expression vector encoding WT SRSF3 or the derived mutant, followed by culturing in growth medium for 24 h. Splicing profiles of MAP4K4 and SRSF3 transcripts were analyzed using RT-PCR assays with specific primer sets as listed in Supplementary Table S1. Immunoblot analyses were performed using specific antibodies against FLAG-tagged proteins, GAPDH, and SRSF3. The bar graph presents relative levels of MAP4K4 Iso 4 and SRSF3 +ex4′ transcripts in independent experiments (n = 4). Signal densities of the RT-PCR results were analyzed using TotalLab Quant Software, and quantitative results are shown as the mean ± SD (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference.

Figure 2.
Overexpression of RNA-binding motif protein 4a (RBM4a) specifically reprograms splicing profiles of mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) and serine/arginine-rich splicing factor 3 (SRSF3) transcripts. (A) Total RNAs and cell extracts were prepared from C3H10T1/2 cells respectively transfected with an expression vector encoding various SRSF family members or RBM4a, followed by culturing in growth medium for 24 h. (B) Total RNAs and cell extracts were isolated from C3H10T1/2 cells respectively transfected with an expression vector encoding wild-type (WT) RBM4a or the derived mutants, followed by culturing in growth medium for 24 h. (C) Total RNAs and cell lysates were extracted from C3H10T1/2 cells respectively transfected with an SRSF3-targeting vector, or expression vector encoding WT SRSF3 or the derived mutant, followed by culturing in growth medium for 24 h. Splicing profiles of MAP4K4 and SRSF3 transcripts were analyzed using RT-PCR assays with specific primer sets as listed in Supplementary Table S1. Immunoblot analyses were performed using specific antibodies against FLAG-tagged proteins, GAPDH, and SRSF3. The bar graph presents relative levels of MAP4K4 Iso 4 and SRSF3 +ex4 transcripts in independent experiments (n = 4).

RBM4a Modulates the Alternative Splicing of SRSF3 Transcripts in a Sequence-Specific Manner
The interplay between the heterogeneous nuclear ribonucleoprotein (HnRNP) I (also referred to as PTBP1) and the UUUCU element within SRSF3 exon 4 was documented to impair SRSF3-mediated autoregulation [17]. We therefore wondered whether RBM4a exhibited an antagonistic effect on PTBP1-regulated SRSF3 splicing as demonstrated in previous reports [18]. The reporter containing intact SRSF3 exon 3, intron 3, and exon 4 elements and nucleotide-substituted mutants were constructed and applied to in vivo splicing assays ( Figure 3A). Results of RT-PCR assays showed that an increase in SRSF3 +ex4 transcripts generated from the WT reporter was observed after treatment with the differentiating cocktail compared to undifferentiated cells ( Figure 3B, lanes 1 and 2, PII = 27.54% and 53.8%). Nucleotide substitutions within the UUUCU element of SRSF3 exon 4 led to elevation of SRSF3 +ex4 transcripts in transfected cells cultured in growth and differentiating media ( Figure 3B

RBM4a Modulates the Alternative Splicing of SRSF3 Transcripts in a Sequence-Specific Manner
The interplay between the heterogeneous nuclear ribonucleoprotein (HnRNP) I (also referred to as PTBP1) and the UUUCU element within SRSF3 exon 4′ was documented to impair SRSF3mediated autoregulation [17]. We therefore wondered whether RBM4a exhibited an antagonistic effect on PTBP1-regulated SRSF3 splicing as demonstrated in previous reports [18]. The reporter containing intact SRSF3 exon 3, intron 3, and exon 4 elements and nucleotide-substituted mutants were constructed and applied to in vivo splicing assays ( Figure 3A). Results of RT-PCR assays showed that an increase in SRSF3 +ex4′ transcripts generated from the WT reporter was observed after treatment with the differentiating cocktail compared to undifferentiated cells ( Figure 3B, lanes 1 and 2, PII = 27.54% and 53.8%). Nucleotide substitutions within the UUUCU element of SRSF3 exon 4′ led to elevation of SRSF3 +ex4′ transcripts in transfected cells cultured in growth and differentiating media ( Figure 3B, lanes 3 and 4, PII = 50.48% and 77.32%), whereas nucleotide substitutions within the intronic CU element diminished utilization of SRSF3 exon 4′ in both undifferentiated and differentiated cells ( Figure 3B, lanes 5 and 6, PII = 23.67% and 24.24%) compared to those of the WT reporter. Moreover, overexpression of RBM4a mediated increases in relative levels of SRSF3 +ex4′ transcripts generated from the WT reporter and exonic mutant ( Figure 3C, lanes 2 and 5, PII = 56.8% and 85.27%) compared to empty-vector transfectants ( Figure 3C, lanes 1 and 4, PII = 44.19% and 52.02%). In contrast, reduced yields of SRSF3 +e 4′ transcripts generated from the WT reporter and exonic mutant were noted after depletion of endogenous RBM4a ( Figure 3C, lanes 3 and 6, PII = 30.42% and 43.47%). Nevertheless, splicing profiles of the SRSF3 intronic mutant showed no response to altered levels of RBM4a compared to empty-vector transfectants ( Figure 3C, lanes 7-9). These results suggested that RBM4a mainly modulated utilization of SRSF3 exon 4′ through the intronic CU element which potentially functions as a splicing enhancer. SRSF3 reporter and derived mutants, followed by culturing in growth (0 day) or differentiating (2 day) medium (n = 4). (C) An empty vector, or RBM4a-expressing vector, or RBM4a-targeting vector were respectively co-transfected with the WT SRSF3 minigene or derived mutants into C3H10T1/2 cells (n = 4). Total RNAs and cell extracts were isolated and subjected to RT-PCR and immunoblotting assays using primer sets as listed in Supplementary Table S1 and specific antibodies. The bar graph shows relative levels of SRSF3 +ex4′ transcripts. Signal densities of the RT-PCR results were quantified using TotalLab Quant Software. Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference. SRSF3 reporter and derived mutants, followed by culturing in growth (0 day) or differentiating (2 day) medium (n = 4). (C) An empty vector, or RBM4a-expressing vector, or RBM4a-targeting vector were respectively co-transfected with the WT SRSF3 minigene or derived mutants into C3H10T1/2 cells (n = 4). Total RNAs and cell extracts were isolated and subjected to RT-PCR and immunoblotting assays using primer sets as listed in Supplementary Table S1 and specific antibodies. The bar graph shows relative levels of SRSF3 +ex4 transcripts. Signal densities of the RT-PCR results were quantified using TotalLab Quant Software. Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference.

RBM4a and SRSF3 Exert Antagonistic Effects on Utilization of MAP4K4 Exon 17
In addition to MAP4K4 iso 4, inclusion of MAP4K4 exon 17 was identified in other MAP4K4 variants. By surveying the genomic sequence of the MAP4K4 transcript, a high-affinity CU element (CUCUUU) of SRSF3 was noted within MAP4K4 intron 17 ( Figure 4A) [19]. To validate the influence of RBM4a and SRSF3 involved in utilization of MAP4K4 exon 17, the WT MAP4K4 minigene and derived mutants were constructed and applied to the following reporter assays ( Figure 4A). As shown in Figure 4B, results of RT-PCR analyses showed that treatment with the differentiating cocktail mediated an increase in MAP4K4 −ex17 transcripts generated from the WT MAP4K4 reporter compared to undifferentiated cells ( Figure 4B, lanes 1 and 2,

RBM4a and SRSF3 Exert Antagonistic Effects on Utilization of MAP4K4 Exon 17
In addition to MAP4K4 iso 4, inclusion of MAP4K4 exon 17 was identified in other MAP4K4 variants. By surveying the genomic sequence of the MAP4K4 transcript, a high-affinity CU element (CUCUUU) of SRSF3 was noted within MAP4K4 intron 17 ( Figure 4A) [19]. To validate the influence of RBM4a and SRSF3 involved in utilization of MAP4K4 exon 17, the WT MAP4K4 minigene and derived mutants were constructed and applied to the following reporter assays ( Figure 4A). As shown in Figure 4B, results of RT-PCR analyses showed that treatment with the differentiating cocktail mediated an increase in MAP4K4 −ex17 transcripts generated from the WT MAP4K4 reporter compared to undifferentiated cells ( Figure 4B   The wild-type (WT) MAP4K4 reporter and derived mutants were respectively transfected into C3H10T1/2 cells, followed by maintenance in growth (0 day) or differentiating (2 days) medium (n = 4). (C) The WT MAP4K4 reporter and derived mutants were respectively co-transfected with an expressing vector encoding RBM4a or SRSF3 into C3H10T1/2 cells. Total RNAs and cell extracts were prepared from transfected cells after 24 h and subjected to RT-PCR and immunoblotting assays with primer sets as listed in Supplementary Table S1 and specific antibodies (n = 4). The bar graph shows relative levels of MAP4K4 −ex17 transcripts. Signal densities of the RT-PCR results were quantified using TotalLab Quant Software. Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference.

MAP4K4 Isoforms Differentially Modulate Activity of the Brown Adipogenesis-Related Signaling Pathway
Treatment with rapamycin was documented to reduce insulin sensitivity and mitochondrial activity of BAs, which was prohibited in the presence of a JNK inhibitor (19). Moreover, expressions of thermogenic genes by BAs were induced after depletion of MAP4K4 which served as an activator of the JNK signaling pathway [20,21]. Potential impacts of MAP4K4 isoforms on JNK signaling were therefore investigated using an in vitro cell model. Results of an immunoblot assay showed gradual decreases in phosphorylated (p)-JNK-1 throughout the development of BATs ( Figure 5A, left, lanes 1 and 2) and differentiation of C3H10T1/2 cells ( Figure 5B, lanes 1 and 2), whereas depletion of RBM4a impaired the reduction in p-JNK-1 along with the development of BATs ( Figure 5A, lanes 3 and 4) and cultured cells ( Figure 5B, lanes 3 and 4). Nevertheless, total JNK expression was not altered by the differentiating condition or RBM4a ablation ( Figure 5A,B, T-JNK). Phosphorylation of JNK-1 was induced only in the presence of overexpressed MAP4K4 Iso 1, but not MAP4K4 Iso 4, compared to that of the empty-vector transfectant ( Figure 5C, p-JNK, lanes 1-3). Overexpressed MAP4K4 isoforms exhibited no influence on expression of total JNK ( Figure 5C, T-JNK) or the loading control. Results of IP assays showed that JNK-1 protein was more associated with FLAG-tagged MAP4K4 Iso1 than with overexpressed MAP4K4 Iso4 in transfected cells ( Figure 5D, lanes 5 and 6), which may have subsequently led to the differential phosphorylation of JNK-1 proteins ( Figure 5C). Taken together, RBM4a-mediated splicing constituted a regulatory mechanism for manipulating the impact of MAP4K4 on JNK-1 activation. MAP4K4 reporter and derived mutants were respectively transfected into C3H10T1/2 cells, followed by maintenance in growth (0 day) or differentiating (2 days) medium (n = 4). (C) The WT MAP4K4 reporter and derived mutants were respectively co-transfected with an expressing vector encoding RBM4a or SRSF3 into C3H10T1/2 cells. Total RNAs and cell extracts were prepared from transfected cells after 24 h and subjected to RT-PCR and immunoblotting assays with primer sets as listed in Supplementary  Table S1 and specific antibodies (n = 4). The bar graph shows relative levels of MAP4K4 −ex17 transcripts. Signal densities of the RT-PCR results were quantified using TotalLab Quant Software. Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005), N.D., No Difference.

MAP4K4 Isoforms Differentially Modulate Activity of the Brown Adipogenesis-Related Signaling Pathway
Treatment with rapamycin was documented to reduce insulin sensitivity and mitochondrial activity of BAs, which was prohibited in the presence of a JNK inhibitor (19). Moreover, expressions of thermogenic genes by BAs were induced after depletion of MAP4K4 which served as an activator of the JNK signaling pathway [20,21]. Potential impacts of MAP4K4 isoforms on JNK signaling were therefore investigated using an in vitro cell model. Results of an immunoblot assay showed gradual decreases in phosphorylated (p)-JNK-1 throughout the development of BATs ( Figure 5A, left, lanes 1 and 2) and differentiation of C3H10T1/2 cells ( Figure 5B, lanes 1 and 2), whereas depletion of RBM4a impaired the reduction in p-JNK-1 along with the development of BATs ( Figure 5A, lanes 3 and 4) and cultured cells ( Figure 5B, lanes 3 and 4). Nevertheless, total JNK expression was not altered by the differentiating condition or RBM4a ablation ( Figure 5A,B, T-JNK). Phosphorylation of JNK-1 was induced only in the presence of overexpressed MAP4K4 Iso 1, but not MAP4K4 Iso 4, compared to that of the empty-vector transfectant ( Figure 5C, p-JNK, lanes 1-3). Overexpressed MAP4K4 isoforms exhibited no influence on expression of total JNK ( Figure 5C, T-JNK) or the loading control. Results of IP assays showed that JNK-1 protein was more associated with FLAG-tagged MAP4K4 Iso1 than with overexpressed MAP4K4 Iso4 in transfected cells ( Figure 5D, lanes 5 and 6), which may have subsequently led to the differential phosphorylation of JNK-1 proteins ( Figure 5C). Taken together, RBM4a-mediated splicing constituted a regulatory mechanism for manipulating the impact of MAP4K4 on JNK-1 activation.

MAP4K4 Isoforms and SRSF3 Exert Discriminative Effects on Brown Adipogenesis
An in vitro cell model coupled with overexpression of MAP4K4 isoforms or alteration of SRSF3 abundances was next conducted to validate the influence of MAP4K4 isoforms and SRSF3 on the development of brown adipogenesis. Results of qPCR assays indicated that the presence of overexpressed MAP4K4 Iso 4 mediated an increase in expression of uncoupling protein 1 (UCP1) transcript, that serve as a key factor involved in non-shivering thermogenesis in BAs compared to the empty-vector transfectant ( Figure 6A, UCP1). In contrast, overexpression of MAP4K4 Iso1 resulted in reduced level of UCP1 transcript. An increase in BMP2 transcripts with concomitant decreases in brown adipocytes-specific (PRDM 16 and BMP7) and beige cell-related factors (Cited 1 and Hoxa9) were noted in the MAP4K4 Iso 1-and SRSF3-overexpressing cells ( Figure 6A,B, gray bar) compared to those of the empty-vector transfectants ( Figure 6A,B, white bar). Overexpression of MAP4K4 Iso 4 and depletion of SRSF3 reversely manipulated the expression profiles of adipocyte-related transcription factors ( Figure 6A,B, black bar). Besides the altered expression profiles of a thermogenic factor, overexpression of MAP4K4 Iso 4 mediated elevated basal and maximal OCRs, and ATP production ( Figure 6C, red bar) compared to those of empty-vector transfected C3H10T1/2 cells ( Figure 6C, blue bar). Reversely, overexpression of MAP4K4 Iso1 resulted in reduced basal and maximal OCRs, and ATP production ( Figure 6C, green bar). SRSF3-knockdown cells also exhibited elevated basal and maximal OCRs, and ATP production ( Figure 6D, red bar), whereas the reduced basal and maximal OCRs, and ATP production were monitored in SRSF3-overexpressing cells ( Figure 6D, green bar) compared to empty-vector transfectants ( Figure 6D, blue bar). These results indicated the repressive influence of MAP4K4 iso 1 and SRSF3 on brown adipogenesis, whereas the presence of MAP4K4 iso 4 enhanced the physiological signatures of brown adipocytes.

MAP4K4 Isoforms and SRSF3 Exert Discriminative Effects on Brown Adipogenesis
An in vitro cell model coupled with overexpression of MAP4K4 isoforms or alteration of SRSF3 abundances was next conducted to validate the influence of MAP4K4 isoforms and SRSF3 on the development of brown adipogenesis. Results of qPCR assays indicated that the presence of overexpressed MAP4K4 Iso 4 mediated an increase in expression of uncoupling protein 1 (UCP1) transcript, that serve as a key factor involved in non-shivering thermogenesis in BAs compared to the empty-vector transfectant ( Figure 6A, UCP1). In contrast, overexpression of MAP4K4 Iso1 resulted in reduced level of UCP1 transcript. An increase in BMP2 transcripts with concomitant decreases in brown adipocytes-specific (PRDM 16 and BMP7) and beige cell-related factors (Cited 1 and Hoxa9) were noted in the MAP4K4 Iso 1-and SRSF3-overexpressing cells ( Figure 6A,B, gray bar) compared to those of the empty-vector transfectants ( Figure 6A,B, white bar). Overexpression of MAP4K4 Iso 4 and depletion of SRSF3 reversely manipulated the expression profiles of adipocyte-related transcription factors ( Figure 6A,B, black bar). Besides the altered expression profiles of a thermogenic factor, overexpression of MAP4K4 Iso 4 mediated elevated basal and maximal OCRs, and ATP production ( Figure 6C, red bar) compared to those of empty-vector transfected C3H10T1/2 cells ( Figure 6C, blue bar). Reversely, overexpression of MAP4K4 Iso1 resulted in reduced basal and maximal OCRs, and ATP production ( Figure 6C, green bar). SRSF3-knockdown cells also exhibited elevated basal and maximal OCRs, and ATP production ( Figure 6D, red bar), whereas the reduced basal and maximal OCRs, and ATP production were monitored in SRSF3-overexpressing cells ( Figure 6D, green bar) compared to empty-vector transfectants ( Figure 6D, blue bar). These results indicated the repressive influence of MAP4K4 iso 1 and SRSF3 on brown adipogenesis, whereas the presence of MAP4K4 iso 4 enhanced the physiological signatures of brown adipocytes. Figure 6. Mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) isoforms and serine/arginine-rich splicing factor 3 (SRSF3) exhibit differential impacts on brown adipogenic gene expressions and metabolic signatures. (A,B) C3H10T1/2 cells were respectively transfected with an empty vector, expression vectors encoding MAP4K4 isoforms, or SRSF3-overexpressing or SRSF3targeting vectors (n = 4). Total RNAs were prepared from transfected cells and subjected to qPCR assays with specific primer sets as listed in Supplementary Table S2. The bar graph shows relative levels of adipogenic genes normalized with the levels of Gapdh transcripts. (C,D) C3H10T1/2 cells were respectively transfected with an empty vector, expressing vectors encoding MAP4K4 isoforms, or SRSF3-overexpressing, or SRSF3-targeting vector, followed by culturing in growth medium for 24 h and then were subjected to bioenergetic analyses. The bar graph shows mean values of the basal and maximal oxygen consumption rates, and ATP production which were measured using a Seahorse XF24 Bioanalyzer (n = 4). Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005).

Discussion
AS cascades constitute a common mechanism for determining cell and tissue development. For instance, NMD-coupled autoregulation and cross-regulation of PTBP1 and neuronal PTB (also Figure 6. Mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) isoforms and serine/arginine-rich splicing factor 3 (SRSF3) exhibit differential impacts on brown adipogenic gene expressions and metabolic signatures. (A,B) C3H10T1/2 cells were respectively transfected with an empty vector, expression vectors encoding MAP4K4 isoforms, or SRSF3-overexpressing or SRSF3-targeting vectors (n = 4). Total RNAs were prepared from transfected cells and subjected to qPCR assays with specific primer sets as listed in Supplementary Table S2. The bar graph shows relative levels of adipogenic genes normalized with the levels of Gapdh transcripts. (C,D) C3H10T1/2 cells were respectively transfected with an empty vector, expressing vectors encoding MAP4K4 isoforms, or SRSF3-overexpressing, or SRSF3-targeting vector, followed by culturing in growth medium for 24 h and then were subjected to bioenergetic analyses. The bar graph shows mean values of the basal and maximal oxygen consumption rates, and ATP production which were measured using a Seahorse XF24 Bioanalyzer (n = 4). Quantitative results are shown as the mean ± SD. Statistical significance was determined using Student's unpaired t-test (* p < 0.05; ** p < 0.01; *** p < 0.005).

Discussion
AS cascades constitute a common mechanism for determining cell and tissue development. For instance, NMD-coupled autoregulation and cross-regulation of PTBP1 and neuronal PTB (also referred to as PTBP2) led to reprogramming of splicing networks throughout neuronal development [22]. In this report, a postnatal switch of the SRSF3 and RBM4a protein expression reprogrammed splicing profiles of MAP4K4 transcripts, which lessened the repressive effect of JNK signaling on the development and functioning of BATs.
MAP4K4 contains a serine/threonine kinase domain at the N-terminus and a citron-homology (CNH) domain at the C-terminus which participate in protein interactions [23]. The CNH and kinase domains of the MAP4K4 protein were demonstrated to be essential for activation of downstream signaling, including the JNK pathway [24]. Functional MAP4K4 isoforms were previously identified to contain a distinct intermediate region and a regulatory region at the C-terminus which are encoded by alternatively spliced transcripts [24,25]. Despite this, the influences of distinct MAP4K4 isoforms on carcinogenesis or adipogenesis are as yet uncharacterized. Repressive effects of MAP4K4 on insulin sensitivity, oxidative metabolism, and adipogenesis were revealed using RNA interference (RNAi)-based screening [26]. MAP4K4 was next documented to abolish adipogenesis by downregulating the expression of peroxisome proliferator-activated receptor (PPAR) gamma protein through various signaling pathway, including Wnt, Tumor necrosis factor alpha, and c-JNK pathway [27][28][29]. Although the influence of JNK signaling on adipogenesis was controvertible, treatment with an immunosuppressive agent, such as rapamycin, was recently demonstrated to diminish insulin sensitivity and mitochondrial activity in skeletal muscles and BAs through a MAP4K4-activated JNK pathway [19,27]. Application of a JNK inhibitor inversely lessened the influence of the MAP4K4-mediated pathway on reducing insulin signaling in BATs [19]. In this study, the presence of MAP4K4 Iso4 alone was first reported to enhance the differentiation and metabolism of pre-adipocytes, which may release the repressive effect of MAP4K4 Iso1 throughout brown adipogenesis. These results suggested that the effect of functional protein can be fine-tuned by manipulating the composition of encoded isoforms instead of altering the transcriptional activity. Nevertheless, the influence of MAP4K4 gene on development of WATs or beige cells was worthy of further investigation.
The SRSF family functions as a major group that orchestrates post-transcriptional regulation to coordinate spatial and temporal gene profiles involved in diverse cellular processes, including differentiation and maturation [28,29]. The AS-coupled NMD mechanism constitutes a common mechanism for auto-and cross-regulation of SRSF family members, such as SRSF3, SRSF6, and other splicing regulators [18,30,31]. The binding tendency of SRSF3 toward the CU element (CUCUUC), which frequently functions as an exonic splicing silencer, was identified using cross-linked IP-coupled sequencing and subsequent functional assays [32,33]. The interplay between SRSF3 and a corresponding element within the 3 region of its own exon 4 led to an increase in NMD-targeted SRSF3 transcripts, subsequently maintaining the homeostasis of SRSF3 expression [17]. Upregulation of hnRNP proteins was documented to impair the autoregulation of SRSF3 splicing and subsequently resulted in imbalanced expression of SRSF3 in various cancers [17]. The simultaneous interplay between RBM4a and the CU element within the SRSF3 exon 4 and downstream intron constituted another mechanism for maintaining the homeostasis of SRSF3 during BAT development. Nevertheless, the regulation of SRSF3 splicing throughout BAT development and the repressive effect of SRSF3 on metabolic activity of BAs was first reported in this study. Although the reduced mass of BATs was previously noted in the SRSF3-knockout mice [34], overexpression of SRSF3 was demonstrated to enhance the relative levels of PKM2 transcripts which were predominantly transcribed in WAs [35]. These controversial results suggested that proper expressions of SRSF3 might play a potential role in the development of BAT or other adipose tissues, which is worthy of further investigation.
A gradual increase in RBM4a was previously demonstrated to reprogram splicing profiles of transcription and splicing regulators throughout brown adipogenesis. In this study, a novel splicing network composed of RBM4a, SRSF3, and MAP4K4 was identified in BAs using a transcriptome analysis, and subsequently manipulating the effect of downstream JNK signaling which contributes to insulin resistance and reduced metabolism. Therefore, the impacts and regulatory mechanism of a brown adipogenic splicing network are worthy of further investigation, which could function as a therapeutic target for obesity-related diseases.

Ethics Statement for Animal Research
Animal experiments were approved by the Institutional Animal Care and Use Committee of Taipei Medical University (no. LAC-2016-0351) and were conducted according to relevant guidelines to minimize animal suffering. The creation of RBM4a −/− mice and isolation of interscapular BATs at distinct stages were described in previous studies [14,16].

Transcriptome Analyses
Transcriptome assays of mice BATs were performed as in previous studies [15]. Briefly, total RNAs were purified using a PureLink RNA mini kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. Total RNAs (8 µg) with a high integrity number (RIN > 8.0) were used for library construction with the NEBNext Ultra RNA Library Prep Kit (NEB, Ipswich, MA, USA) according to the manufacturer's instructions. Paired-end reads (150 bp) were amplified on an Illumina Hi-Seq 4000 platform with the prepared libraries. Preliminary reads were trimmed, filtered, and aligned to the mouse reference genome (GRCm37) using the Tophat v2.0.9 program. Aligned reads were applied to transcriptome assemblies using the Cufflink program. Transcriptome assemblies generated from individual samples were merged using the Cuffmerge utility to estimate transcript levels in each condition. Expression levels and the statistical significance of the merged assemblies were examined using a Cuffdiff and CLC genomic workbench analysis.

Cell Culture and In Vitro Differentiation
Mouse C3H10T1/2 fibroblast cells were cultured in Dulbecco's modified Eagle medium (DMEM; Invitrogen) supplemented with 10% fetal bovine serum (FBS; Invitrogen). In vitro differentiation was conducted by shifting C3H10T1/2 cells to induction medium supplemented with 20% FBS, 0.5 mM IBMX, 12.7 µM dexamethasone, and 10 µg/mL insulin. The induction medium was replaced with differentiation medium (DM) supplemented with 10% FBS and 10 µg/mL insulin after 48 h and replenished every 2 days.

Plasmid Construction
Coding regions of human MAP4K4 and SRSF3 were amplified using a polymerase chain reaction (PCR) with a mice fetal complementary (c)DNA library as the template. The PCR product was digested with Hind III/EcoR I and EcoR I/EcoR V restriction enzymes and inserted into the p3XFLAG-CMV14 vector (Sigma, St. Louis, MO, USA). The mouse genomic element containing SRSF3 exon 3, intron 3, and exon 4 was PCR-amplified with genomic DNA prepared from C3H10T1/2 cells as the template. The PCR product was digested with EcoR V and Sac I restriction enzymes, and then the insert was placed into the pUC19 vector (NEB). The MAP4K4 minigene was constructed by inserting mouse MAP4K4 genomic fragments containing exon 15, exon16, exon 18, and the complete introns into the pUC19 vector (NEB). PCR-amplified fragments were digested with EcoR I and Hind III restriction enzymes prior to DNA ligation. Derived mutants of the expressing vectors or minigene reporters harboring substituted nucleotides were constructed using the QuikChange site-directed mutagenesis system (Stratagene, Amsterdam, The Netherlands). The vector-based short hairpin RNA targeting mouse SRSF3 was purchased from the RNAi core facility at Academia Sinica (Taipei, Taiwan).

Transient Transfection, Reverse-Transcription (RT)-PCR, and Quantitative (q)PCR Analyses
Cultured C3H10T1/2 cells at~60% confluence were transfected with the indicated plasmid using Lipofectamine 3000 according to the manufacturer's protocol (Invitrogen). Total RNAs and cell extracts