Distinct expression of SHMT2 isoform 1 and 3 in cisplatin-sensitive and cisplatin-resistant ovarian cancer
To clarify the mechanisms underlying cisplatin resistance in ovarian cancer, quantitative proteomics analysis was performed to globally screen the potential molecular targets regulated by cisplatin resistance. We used protein to perform on 9 cisplatin-sensitive and 9 cisplatin-resistant ovarian cancer tissues were used for quantitative proteomics analysis and SHMT2 was identified to be downregulated in the cisplatin-resistant ovarian cancer tissues (Fig. 1A). Expression of SHMT2 was further analyzed using 24 cisplatin-sensitive and 12 cisplatin-resistant ovarian cancer tissues, and confirmed the downregulation of SHMT2 in cisplatin-resistant ovarian cancer tissues (Fig. 1B). Two SHMT2 isoforms were identified in ovarian cancer tissues, the smaller isoform was upregulated, even though downregulation of total SHMT2 in cisplatin-resistant tissues (Fig. 1B). Real-time RT-PCR demonstrated that the total SHMT2 mRNA expression was significantly decreased in cisplatin-resistant compared to the cisplatin-sensitive ovarian cancer tissues (Fig. 1C). There are multiple SHMT2 variants, the individual variant expression was then studied and found that variant 1 and variant 4 were significantly decreased and increased in the cisplatin-resistant tissues, which encode isoform 1 and isoform 3 of SHMT2 respectively (Fig. 1C). With regards to other variants, there were no significant difference in the cisplatin-sensitive and cisplatin-resistant ovarian cancer tissues (Fig. 1C). The online splicing data from TCGA database (tsvdb.com/index.html) also confirmed expression of multiple SHMT2 variants in ovarian cancer tissues (Fig. 1D). To further determine the potential importance of SHMT2 in clinical settings, we analyzed clinical outcomes from GEPIA OV samples. Compared with normal tissues, the expression level of SHMT2 was significantly increased (Fig. 1E). In addition, the expression of SHMT2 among Stage II, Stage III and Stage III group were significantly different (Fig. 1F). Kaplan-Meier curve showed that neither OS (Fig. 1G) nor PFS (Fig. 1H) were significantly associated with the mRNA expression level of SHMT2. To better understand the relevance of SHMT2 expression in OV patients, KM plotter was used to evaluate prognostic value based on Affymetrix microarrays. It is worth noting that the expression of SHMT2 was detected by different Affymetrix microarrays, the effect on prognosis was different (Fig. 1I-J). This might because different probes recognize different SHMT2 variants, and the different effects on prognosis suggest that different SHMT2 variants might have different biological behaviors, which may be related to cisplatin resistance.
SHMT2 isoform 3 is expressed in cisplatin-resistant ovarian cancer under metabolic stress
To further explore the role of SHMT2 isoforms in the process of cisplatin resistance, SHMT2 expression was studied using paired cisplatin-resistant and parental SKOV3 and A2780 cells. SHMT2 expression did not show significant difference between the cisplatin-resistant SKOV3 (SKOV3/DDP) and A2780 (A2780/DDP) cells compared with their respective parental cells (Fig. 2A). In addition, only single band representing SHMT2 isoform 1 was observed in ovarian cancer cell lines (Fig. 2A). Single-cell RNA sequencing analysis showed that SHMT2 was mainly expressed in epithelial and mesenchymal cells, in addition, SHMT2 also expressed in endothelial cells (Fig. 2B-C). Next, cancer-associated fibroblast (CAF) and ovarian cancer (OC) cells were extracted from fresh ovarian cancer tissues, Western blot demonstrated that CAF expressed SHMT2 isoform 1, while some OC cells expressed SHMT2 isoform 3 (Fig. 2D-E), suggesting that the expression of SHMT2 isoform 3 might be mainly derived from ovarian tumor epithelial cells. SHMT2 isoform 3 could be detected in some primary OC cells at the beginning, but with continuous cell culture, SHMT2 isoform 3 could not be detected. However, SHMT2 isoform 3 could be detected again under the conditions of hypoglycemia and hypoxia (Fig. 2F). Real-time RT-PCR showed that variant 1 decreased, while variant 4 decreased in OC#2 and OC#5 after 2–3 passages, culture under hypoglycemic and hypoxic condition recovered the expression of variant 4 (Fig. 2G). On the other hand, such fluctuation of expression of SHMT2 variants was not observed in OC#3, which did not express SHMT2 variant 4 originally (Fig. 2G). In the cisplatin-resistant SKOV3 (SKOV3/DDP) and A2780 (A2780/DDP) cells, SHMT2 isoform 3 could be detected in cisplatin-resistant SKOV3/DDP and A2780/DDP cells under hypoglycemic and hypoxic culture, but could not be detected in their respective parental cells (Fig. 2H). RT-PCR showed that hypoglycemic and hypoxic culture increased total and variant 1 SHMT2 expression in SKOV3 cells, while such induction was not observed in A2780 cells (Fig. 2I-J). SHMT2 variant 1 was consistently decreased (Fig. 2J), while SHMT2 variant 4 was consistently increased (Fig. 2K-L) in cisplatin-resistant cells cultured under hypoglycemic and hypoxic condition. In general, OC can express SHMT2 variant 4 only when metabolic stress occurs, that is, under a condition of low-glucose and hypoxic, thereby expressing the SHMT2 isoform 3.
Different effects of SHMT2 knockdown in cisplatin responsiveness in cisplatin-resistant and sensitive ovarian cancer cells
In order to explore the role of SHMT2 in cisplatin resistance, SHMT2 was knocked down using CRISP/Cas9 system in SKOV3/DDP and A2780/DDP and their respective parental cells (Fig. 3A). Knockdown of SHMT2 increased the survival of both SKOV3 and A2780 cells exposed to cisplatin treatment. However, in SKOV3/DDP and A2780/DDP cells knockdown of SHMT2 did not show a significant difference in the viability of the cells after cisplatin exposure (Fig. 3B). Colony formation assay got a similar result. Knockdown of SHMT2 increased the colony number of cisplatin sensitive cells, while there was no significant change in the resistant cells under cisplatin treatment (Fig. 3C-D).
To further understand the role of SHMT2 isoforms in cisplatin resistance, SHMT2 knockdown cells were restored with SHMT2 isoform 1 and 3, respectively (Fig. 3E). Restoration of SHMT2 isoform 1, but not isoform 3, significantly decreased the cell viability (Fig. 3F) and colony formation (Fig. 3G-H) of SKOV3 and A2780 cells under cisplatin exposure. SHMT2 isoform 3 increased colony formation of SKOV3 cells, but not A2780 cells (Fig. 3G-H). Neither isoform 1 nor isoform 3 demonstrated any effects on viability and colony formation of SKOV3/DDP and A2780/DDP cells (Fig. 3F-H).
Opposite roles of SHMT2 isoforms in maintenance of stem cell like features of ovarian cancer cells
As mentioned above, SHMT2 isoform 3 only expressed under metabolic stress situation, in order to further understand the roles of SHMT2 isoforms, a low-glucose and hypoxic cell culture environment were constructed for the following experiments. SHMT2 isoform 3, but not isoform 1, significantly promoted survival of both cisplatin-sensitive and cisplatin-resistant cells under low-glucose and hypoxic conditions (Fig. 4A-B). SHMT2 isoform 1 decreased, while isoform 3 increased stem cell like features of ovarian cancer cells, as identified by spheroid formation ability (Fig. 4C-D), expression of stemness markers CD44 and CD133 (Fig. 4E), as well as tumor formation rate and stem cell frequency using the tumor-limiting dilution assay (Fig. 4F).
Selective utilization of alternative promoters of SHMT2 in cisplatin-resistant ovarian cancer cells under metabolic stress
SHMT2 gene contains three promoters, and different variants will be transcribed due to selective utilization of different promoters (Fig. 5A-B). The reporters containing − 841/+72, -841/+448, -841/+724 fragments showed highest and similar activities, while no obvious activity of reporter containing + 82/+448 or + 520/+724 fragments was observed in both cisplatin-sensitive and cisplatin-resistant ovarian cancer cells under normal condition (Fig. 5C), indicating that ovarian cancer cells preferentially utilized promoter 1, while promoter 2 and promoter 3 were rarely activated under normal culture condition. The activity of reporter containing − 841/+72 fragment significantly decreased, while the activity of reporter containing + 82/+448 fragment significantly increased in SKOV3/DDP and A2780/DDP under hypoglycemic and hypoxic condition (Fig. 5C). The enrichment of RNA Pol II pSer2 and Pol II pSer5 of SKOV3/DDP and A2780/DDP to the promoter 1 region was significantly decreased under the low-glucose and hypoxic condition, while there was no significant difference in SKOV3 and A2780 cells (Fig. 5D).
Selective utilization of SHMT2 promoter 2 by HIF1α and TFE3 complex
To understand the regulation of SHMT2α by transcription factors, the specific transcription factor binding sites around promoter 2 which were associated with metabolic stress were surveyed and XBP1, HIF1α and TFE3 were screened out. Knockdown of TFE3 and HIFα in low-glucose and hypoxic condition decreased the expression of SHMT2 isoform 3, while XBP1 knockdown demonstrated no obvious effect (Fig. 6A). ChIP confirmed significant enrichment of HIF1α and TFE3 to the promoter 2 region in cisplatin-resistant cells under low-glucose and hypoxic culture condition (Fig. 6B). Knockdown of TFE3 and HIFα decreased the activity of promoter 2, and knockdown of XBP1 had no significant difference to the activity of promoter 2 (Fig. 6C). To further study whether there was a regulation between TFE3 and HIFα, ChIP showed that TFE3 knockdown inhibited the recruitment of HIF1α in both SKOV3/DDP and A2780/DDP cells (Fig. 6D). HIF1α knockdown decreased recruitment of TFE3 in A2780/DDP, while had no effect in SKOV3/DDP (Fig. 6D). Molecular docking model predicted direct interaction of HIF1α and TFE3 (Fig. 6E). DuoLink confirmed interaction of TFE3 and HIF1α in cisplatin-resistant ovarian cancer cells under the low-glucose and hypoxic condition (Fig. 6E).