PRRX1‐induced epithelial‐to‐mesenchymal transition in salivary adenoid cystic carcinoma activates the metabolic reprogramming of free fatty acids to promote invasion and metastasis

Abstract Objectives Increasing evidences demonstrate a close correlation between epithelial‐to‐mesenchymal transition (EMT) induction and cancer lipid metabolism. However, the molecular mechanisms have not been clarified. Materials and methods In our study, the relative expression level of PRRX1 was detected, its relationship with free fatty acid (FFA) and PPARG2 was analysed in 85 SACC tissues and 15 salivary glands from the benign salivary tumours. We also compared the FFAs composition and levels in these SACC cells. PPARG2 was detected in PRRX1‐induced FFAs treatment as well as Src and MMP‐9 were detected in FFAs treatment–induced invasion and migration of SACC cells, and ChIP test was performed to identify the target interactions. Results Our data showed that overexpression of PRRX1 induced EMT and facilitated the invasion and migration of SACC cells, and PRRX1 expression was closely associated with high FFAs level and poor prognosis of SACC patients. Furthermore, PRRX1 silence led to the increase of PPARG2 and the reduction of FFAs level and the migration and invasion of SACC cells. And inhibition of PPARG2 rescued FFAs level and migration and invasion capabilities of SACC cells. Free fatty acids treatment induced an increase of Stat5‐DNA binding activity via Src‐ and MMP‐9‐dependent pathway. Conclusions Collectively, our findings showed that the PRRX1/PPARG2/FFAs signalling in SACC was important for accelerating tumour metastasis through the induction of EMT and the metabolic reprogramming of FFAs.


| INTRODUC TI ON
Salivary adenoid cystic carcinoma (SACC) is one of the most common highly malignant salivary gland tumours and accounts for 10% of salivary gland tumours and 21%-24% of adenocarcinoma and overall <1% of all cancers in the United States or <3% in China. [1][2][3] Five-year survival rate for patients with SACC is >70% but drops to 50% after 10 years, and 20% after 20 years when invasion into adjacent tissues and haematogenous spread to distant organs (lung, bone and liver). Approximately 40% of SACC patients have distant metastases, and metastasis is one of the most common poor prognostic factors in SACC patients. 4 Epithelial-mesenchymal transition (EMT) in cancer cells has been considered to be not a simple process to acquire invasive and metastatic abilities, 5-8 but a complicated and comprehensive reprogramming, involved in metabolism, epigenetics and differentiation. 9 Metabolic reprogramming of cells has long been appreciated to contribute to oncogenesis. 10,11 Alteration in free fatty acids (FFAs) metabolism in cancer cells is one of the most important parts of the lipid metabolism. 12 However, the molecular mechanism underlying lipid metabolic changes during EMT and metastatic progression of cancer, including alterations in regulatory gene expression, still remain undefined.
Free fatty acids, either saturated or unsaturated, consist of the terminal carboxyl group and the hydrocarbon chain mostly occurring in even numbers of carbons. 13 Epidemiologic study exhibited that there was negative association between breast cancer and colon cancer patients and cis-mono-unsaturated fatty acids and positively with trans fatty acids. 14 An altered free fatty acid pattern played a detrimental role in the progression of colorectal and prostate cancer patients. 15,16 Elongation of fatty acids from C16 to C18 had been reported to promote both hepatic lipid accumulation and inflammation. 17 G-proteincoupled receptor 120 (GPR120) and GPR40 were identified as FFA receptors (FFARs) for long-and medium-chain fatty acids. GPR120 negatively and GPR40 positively regulated cellular functions during the malignant progression of lung cancer. 18 These indicated that changed FFAs could affect the biological behaviours and malignant characteristics of cancer cells. However, the epidemiological evidence of the correlation between increased FFAs and poor SACC prognosis has been unknown. PRRX1, the paired-related homeobox transcription factor, is an EMT inducer conferring the migratory and invasive properties of cancer cells. 19,20 In pancreatic ductal adenocarcinoma, PRRX1b promoted tumour invasion, dedifferentiation and EMT. And PRRX1a stimulated metastatic outgrowth in the liver, tumour differentiation and mesenchymal-epithelial transition (MET). 21 On the other side, the global gene expression assay in abdominal subcutaneous adipose tissue showed that homeobox transcription factors (such as PRRX1) and extracellular matrix structural proteins were upregulated after fat loss, implicating PRRX1 as an important regulator of adipose tissue function. 22 Integrative computational analysis of phylogenetic conservation detected that PRRX1 contributed to the release of FFAs and insulin resistance. 23 The data indicated that PRRX1 was an EMT inducer of cancer cells, and PRRX1 can mediate lipid metabolism of adipose tissue; however, to our knowledge, whether PRRX1 could regulate lipid metabolism to involve in the invasion and metastasis of cancer cells has yet not been reported.
Importantly, in breast cancer cells undergoing EMT, an increased expression of fatty acid synthase (FASN) resulted in saturated fatty acids (FAs) accumulation and relocation through activating the EMT inducer VEGF/ VEGFR2 signalling. 24 The elevated free fatty acid uptake via CD36 was associated with EMT induction in hepatocellular carcinoma. 25 These results suggest that EMT may change FFAs to regulate lipid metabolism of cancer cells during cancer progression. We therefore hypothesize that PRRX1 is important for tumour invasion and metastasis through both the induction of EMT and the metabolic reprogramming of FFAs.

| Immunohistochemistry
Fresh tissue samples were fixed in 10% formaldehyde and embedded in paraffin. Sections were cut and stained using a conventional immunohistochemistry procedure. Sections were incubated with PRRX1 antibody (1:80 dilutions; Novus Biologicals) or PPARG2 antibody (1:50 dilutions; ProteinTech) for 2 hours, followed by incubation with secondary antibodies (DAKO) for 30 minutes.

| Lenti virus transfection and confirmation of transfection
The PRRX1 overexpressed lentiviral vector with a luciferase reporter gene was synthesized by Guangzhou Cyagen Biosciences Inc It was constructed by ligating the human PRRX1 sequence (654 bp) into the BamHI/Asc II sites of the pLV.ExBi.P/Puro-EF1α-PRRX1-IRES-luc2 (8934 bp). The lentiviral vector was packaged using pCD/NL-BH*DDD packaging plasmid mix (Addgene) and transiently cotransfected into 293T cells to generate recombinant virus particles. After 48 hours of infection, lentivirus in the supernatant was transduced into SACC cells, using 5 μg/mL of polybrene (Sigma-Aldrich, Germany) at the optimal MOI (multiplicity of infection) of each cell. Stable clones were maintained on 5 μg/mL of puromycin (Sigma-Aldrich). Fluorescence intensity of D-luciferin was observed by fluorescence microscope to indicate the lentivirus transfection efficiency.

| Real-time RT-PCR
Total RNA was extracted from cells using the Trizol (Invitrogen).
The PCR conditions used were initial denaturation at 95°C for

| Immunofluorescence
Cells were cultured at a density of 3 × 10 4 cells per chamber. Upon reaching 70% confluency, culture media was removed, washed and   In SACC tissues, the FFA concentration of the normal salivary gland tissues was designated as the relative baseline. Compared with the baseline, the FFAs concentration was divided into the low FFAs group and the high FFAs group.

| Wound healing assay
Cells cultured in serum-free media for 24 hours in wound healing assays. Cells were plated in 6-well plates at 2.0 × 10 5 cells/ well.
The individual wells were wounded by scratching with a pipette tip and incubated with medium containing no FBS to 24 hours.
Wound closures were photographed to measure the remaining cell-free area in triplicates wells. The percentage of the cell-free area was calculated and showed as mean ± SD of three independent experiments.
After incubated for 24 hours, migrated cells on the lower surface of the membrane were stained with 0.1% crystal violet. Five fields per filter were counted.

| Chromatin immunoprecipitation (ChIP) assays
ChIP assays were performed using a ChIP Assay Kit (Abcam) according to the manufacturer's instructions. Briefly, the minimum number of SACC-LM or SACC-83 cells was 1 × 10 6 cells for every ChIP. These cells were fixed and lysed, and the extracted DNA was sheared by the sonicator to an optimal DNA fragment size of 200-1000 bp.
Chromatin was then precipitated with nonspecific IgG antibodies (Sigma), ChIP-grade PRRX1 antibodies (Proteintech) or ChIP-grade rabbit anti-histone H3 (Sigma). Then, DNA was extracted and purified, and PCR was performed with primers for a PPARG2 promoter fragment.

| Xenograft tumour model
The mice were randomly distributed into four groups (n = 5): in- via the tail vein. The mice were monitored daily for body weight, behaviour and water/food consumption. Tumour growth was monitored by measuring the tumour size in two orthogonal dimensions with vernier calipers every 4 days and the tumour volume was calculated.
The incidence and volume of metastases were estimated by serial imaging of mice for bioluminescence using the IVIS 100 imaging system coupled to a data-acquisition personal computer equipped with Living Image software (Xenogen). The photon emission level (representative of luciferase activity) was used to assess the relative tumour burden in the mice. After tumour cells injection for 4 weeks, the mice were anaesthetized and sacrificed. The tumours were surgically excised, weighed, fixed in 10% formalin and embedded in paraffin.

| Statistical analysis
All statistical analyses were performed using the SPSS package (version 17.0). A value of P < .05 was considered statistically significant.

| PRRX1 is positively associated with FFA accumulation in salivary adenoid cystic carcinoma patients
To evaluate the clinical outcomes of PRRX1 in SACC, we exam-  Table 1). Fifty-three patients with PRRX1-positive expression had a significantly worse prognosis than PRRX1-negative patients (P < .05, Figure 1B). And in Oncomine database, the expression of PRRX1 was also upregulated in 16 cases of salivary adenoid cystic carcinoma and 15 cases of oral squamous cell carcinoma ( Figure S1A-C). This indicated that PRRX1 may be one of the important markers in SACC patients.
Recently, FFA metabolic reprogramming, which alters paradigm in energy metabolism as well as serves as mediators in signal transduction, plays an important role in cancer progression. 16,27 Here, we applied GC/MS to examine FFAs distribution in 85 SACC patients and analysed their association with the prognosis of patients to uncover the role of FFAs in SACC. The data showed that there was a significant correlation of tumour FFA level with both disease-free survival (P = .022) and overall survival (P = .012; Figure 1C, Figure   S1D). Compared with the normal salivary tissue and SACC cases with negative PRRX1, the composition of FFAs of SACC with positive PRRX1 was more complex, and the number of the carbon chain from C16 to C18 extended products added (P < .01, Figure 1E, Table   S1). Moreover, the FFA quantitative analysis displayed that the FFA level in PRRX1-positive group was significantly higher than in PRRX1 negative, and the FFA level in SACC cases was obviously higher than normal salivary tissue (P < .05, Figure 1F). This suggested that there may be a close relationship between PRRX1 and FFAs distribution in SACC patients. In addition, PPARG2 could control lipid metabolism of adipocytes including adipose cell differentiation, lipid storage, FFA transport and β-oxidation. 28 Therefore, we detected PPARG2 expression in SACC samples by immunohistochemical staining. The results showed that PPARG2 expression was negatively the expression of PRRX1 in SACC tissue ( Figure 1A), and PPARG2 overexpression implied the good prognosis of SACC patients (P = .031; Figure 1D).

| PRRX1 induces EMT to promote migration and invasion of salivary adenoid cystic carcinoma cells
PRRX1 has been shown to be an EMT inducer in some kinds of cancers, but the role of PRRX1 in SACC remains unknown. 19,20 To examine the potential of PRRX1 to induce EMT in salivary adenoid cystic carcinoma, we applied the overexpressed lentiviral vector to transfect PRRX1 to salivary adenoid cystic carcinoma  and paracancerous tissue (n = 15), respectively. There was statistically different FFAs composition a PRRX1-positive SACC, PRRX1-negative SACC and paracancerous tissue (*P < .05). Compared with paracancerous tissues and SACC with negative PRRX1, the composition of FFAs of PRRX1-positive SACC was more complex, and the carbon chain from C16 to C18 extend product increased (*P < .05). F, The levels of FFAs in PRRX1-positive SACC, PRRX1-negative SACC, and paracancerous tissue, respectively. The FFAs quantitative analysis displayed that the FFAs levels of SACC tumour tissues with PRRX1 negative were significantly higher than tissues adjacent to the benign salivary tumours (P = .032), and the FFAs level of PRRX1-positive group was significantly higher than PRRX1-negative group (P = .019). Error bars represent the mean ± SD of triplicate experiments. One-way ANOVA statistical analysis was performed to determine significance

| PRRX1 induces the accumulation of FFAs by mesenchymal cancer cells
We next ascertained the relation between PRRX1-induced EMT and  TA B L E 1 Clinical-pathologic characteristic of 85 patients with SACC and association between PRRX1 expression and these variables compared with the centre of the tumour ( Figure S3D,E). These data indicated that PRRX1-induced mesenchymal cancer cells secreted more FFAs at the invasive front due to increased levels of PRRX1.

| Downregulation of PPARG2 is required for upregulation of PRRX1 in response to FFAs treatment
Previous studies indicated that PPARG2 was involved in the regula- Further, we examined whether there was the direct association between PRRX1 and PPARG2 expression. We found that PPARG2 dramatically reduced in PRRX1 overexpressed SACC cells, and PPARG2 increased in PRRX1-silenced SACC cells ( Figure 4D,E, Figure   S5C). This indicated that PPARG2 expression was significantly inverse

| FFAs treatment induces an increase of Stat5-DNA binding activity via Src and MMP-9dependent pathway
It was reported that Src and MMP-9 played important roles in ara- The data showed that the positive rate of PRRX1 at the invasive front was significantly higher than that at the centre of SACC (*P < .05). Error bars represent the mean ± SD of triplicate experiments. Unpaired t test was performed to determine significance and invasion of SACC cells were attenuated in Stat 5 siRNA groups by using Transwell and scratch-wound assays ( Figure 5D,E, *P < .05, **P < .01). This indicated that FFAs treatment might induce an increase of Src-Stat5 binding activity via Src and MMP-9-dependent pathway to promote the invasion and migration of SACC cells.

| PRRX1 and PPARG2 were responsible for lung metastasis of SACC-83 cells in vivo
We next examined whether the PRRX1/PPARG2 pathway observed in cell culture could modulate the metastasis of salivary adenoid cystic carcinoma in animal models. As shown in Figure 6A, there was no significant difference of tumour volume between PRRX1 overexpressed group and the vector in the xenograft nude mice model when 4 weeks after subcutaneously injecting SACC-83 cells (P = .91, Figure 6A). This Compared with the control group without GW9662, the composition of free fatty acids of the Rosiglitazone stimulating groups was more complex, and the long carbon chain extend product from C16 to C18 increased, and C19 and C20 emerged. C, The FFA level in the GW9662 stimulation of SACC cells by FFA quantification colorimetric/fluorometric kit (**P < .01). The FFA concentration of GW9662 stimulating group increased by 0.453 mmol/L and 1.107 mmol/L, respectively, in vector and PRRX1 groups in SACC-LM cells, 0.411 mmol/L and 1.144 mmol/L, respectively, in vector and PRRX1 groups in SACC-LM cells. D, The mRNA level of PPARG2 and PRRX1 was detected by RT-PCR. The mRNA level of PPARG2 reduced to 52.17% and 49.63% in PRRX1 overexpressed SACC-LM and SACC-83 cells. The mRNA level of PPARG2 increased 3.17 and 2.95 times in PRRX1 siRNA groups of SACC-LM and SACC-83 cells, respectively. PPARG2 expression was significantly downregulated in PRRX1 overexpressed SACC-LM and SACC-83 cells, while PPARG2 expression was significantly upregulated by PRRX1 siRNA transfection (**P < .01). E, PPARG2 protein expression was detected by Western blot. The protein level of PRRX1 which aimed at PPARG2 was consistent with mRNA level. F, ChIP test showed that the combination capacity of PRRX1 and PPARG2 promoter in PRRX1 overexpressed group was significantly increased (**P < .01). Error bars represent the mean ± SD of triplicate experiments. Unpaired t test was performed to determine significance overexpressed group, while cancer cells were not found in the liver tissues. And there was no cancer metastatic cell in the lung and liver of the group of SACC-83 via the tail vein ( Figure 6D). These data suggested that PRRX1 contributed to the metastatic phenotype of SACC. were significantly increased. The percentage of C14 reduced from 52.35% to 1.32%, while the percentage of C16 and C18 increased from 47.65% to 98.68% ( Figure 6F, Table S8). Importantly, the ability of SACC-83 cells with PRRX1-overexpressed to metastasize to the lung in nude mice was significantly repressed by upregulation of PPARG2 with administration of Rosiglitazone. These data suggested that PRRX1 and PPARG2 were responsible for the metastasis formation of SACC in mice.

| D ISCUSS I ON
Over the past decade, much research into carcinoma metastasis has focused on cancer cells themselves or their association with tumour immunity and inflammation microenvironment, without taking into much account the unique but complex tumour metabolic microenvironment. 32 Evidence demonstrates a close correlation between EMT induction and lipid metabolic reprogramming in human cancer, which supports the energy requirements of  Peroxisome proliferator-activated receptor (PPAR) has PPAR-α, PPAR-β and PPAR-γ, and PPARG2 is a subtype of PPAR-γ. Previous studies have demonstrated that PPARG2, high expression in large intestine, adipose tissue and hematopoietic cells, is known to activate adipocyte genes. 29 Recently, PPARγ expression has been described in a variety of human malignancies, including breast cancer, prostate cancer, glioblastoma, non-small cell lung carcinoma, ovarian cancer and pancreatic carcinoma. 30 However, the current knowledge about the correlation between PPARG2 and metabolic syndromes in SACC is still missing. We found that in SACC cells, PPARG2 agonist led to the lower concentration of FFAs, and the lower migration and invasion abilities. In contrast, PPARG2 antagonist generated higher FFA levels, and the higher cell

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.