Differential expression of HSF4 in CRC was identified by GEPIA [21] with COAD and READ cohort in TCGA data. R2 analysis platform (AUMC, CEMM) was utilized to evaluate HSF4 expression in CRC datasets, including GSE14333 (Sieber, n = 290), GSE4554 (Watanabe, n = 84), GSE17538 (Smith, n = 232), GSE2109 (EXPO, n = 315) of CRC, and GSE8671 (Mara, n =32) of adjacent colon.

HCT116, LoVo, SW480, DLD1, HT29 and SW620 (Human CRC cells) were obtained from NCACC (Shanghai, China). FHC cells was kindly donated by Dr. Liang Peng (Guangzhou Medical University). All cell lines were identified by NCACC using the CLAIdentiFiler™Plus PCR Kit (Applied Biosystems, Foster, CA, USA) based on short tandem repeat analysis. In addition, all cell lines have been tested by the MycoProbe® Kit (R&D Systems, Minneapolis, MN, USA) and determined to be free of mycoplasma contamination. CRC and FHC cell lines were cultured in RPMI-1640 medium (HyClone, South Logan, UT, USA; 10% FBS+1% penicillin-streptomycin solution), and grew in a 37 °C humidified atmosphere containing 5% CO${}_2$.

Antibodies used in this study included anti-STAT3 (#9139S/Cell signaling, Danvers, MA, USA), anti-HSF4 (#18797-1-AP/Proteintech, Chicago, IL, USA; #PA5-68416/Invitrogen, Carlsbad, CA, USA), anti-p-STAT3 (#9145S/Cell signaling, Danvers, MA, USA), anti-MACC1 (ab226803/Abcam, Cambridge, MA, USA), anti-$\beta$-actin (#20536-1-AP/Proteintech, Chicago, IL, USA). All antibodies were diluted with reference to the instructions.

Three paired CRC tissues and matched adjacent tissue (2–5 cm from cancer tissue) were collected from CRC patients aged from 56 to 71. Immunohistochemical (IHC) staining as previously described [16]. Cytoplasm of tumor cell was probed with anti-HSF4 (#PA5-68416, Invitrogen, Carlsbad, CA, USA) antibody and was usually positive for HSF4. This study strictly adhered to the guidelines of the Declaration of Helsinki, and was approved by the ethics committee of the First People’s Hospital of Yunnan Province (Protocol No. KHLL2021-KY109).

RNA from FHC and six CRC cells was extracted using TRIzol® reagent (Invitrogen, Carlsbad, CA, USA). RNA from each cell was taken as 1 µL to determine the RNA concentration using an ultra-micro spectrophotometer (Blue ray, Chengdu, Sichuan, China). Referring to the instructions of Hairpinit ™ miRNAs qPCR kit (Genepharma, Suzhou, Jiangsu, China), 3 µg of RNA was taken to perform reverse transcription and amplification of miR-330-5p. The reaction program for reverse transcription was performed at 25 °C for 30 min, 42 °C for 30 min and 85 °C for 5 min. The reaction program for PCR amplification was pre-denaturation at 95 °C for 3 min (1 cycle), amplification at 95 °C for 12 s and 62 °C for 40 min (40 cycles). RNU6B as an endogenous control. The sequence of primers are listed in Supplementary Table 1.

hsa-miR-330-5p mimics, HSF4-targeted shRNA and their negative scramble controls were synthesized by Genepharma (Shanghai, China). The targeting sequences of HSF4-targeted shRNAs are shown in Supplementary Table 1. OmicsLink™ ORF cDNA clone of HSF4 (#Z4249) was purchased from FulenGen (Guangzhou, China). CRC cells were transfected with HSF4 cDNA clone or miR-330-5p mimics using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to manual.

Cell viability was accessed using the Cell Counting Kit-8 (CCK-8) method (MCE, Monmouth Junction, NJ, USA) following manual. In colony formation assay, Lovo and HT29 cells were seed into 6-well plates with 300 per well. Then, cells were cultured at 37 °C for 10 d. Lovo and HT29 cells were subsequently subjected to crystal violet staining for 5 min. Colonies consisting of more than 50 cells per colony were observed and counted. In the wound healing assay, a sterile pipette tip was applied to create a scratch on the cell layer. Photomicrographs were captured at 0 h and 48 h after scratching, using a microscope. As previously described [22], Transwell with 8 µm filter (Corning, Corning, NY, USA) was applied cell invasion assay.

The miRNA with target binding to the CDS region of HSF4 was predicted by the DIANA-microT-CDS algorithm and was finalized as miR-330-5p. A fragment (485 bp) containing three potential bindings of miR-330-5p and HSF4 was amplified. Subsequently, the fragment was subcloned downstream of the luciferase gene sequence of the GV272 luciferase vector. GV272 luciferase vector was transfected into 293T cells and assayed for luciferase activity.

Treated Lovo and HT29 cells were harvested and lysed using RAPI (Cell signaling, Danvers, MA, USA). Protein concentration of the lysates was determined using BCA protein assay kit (Beyotime, Shanghai China). Cell lysates with 30 µg was loaded in SDS-PAGE for electrophoresis. Subsequently, proteins of each group were transferred to polyvinylidene fluoride (PVDF) membranes. Color development was performed on the PVDF membrane using chemiluminescent substrate (Invitrogen, Carlsbad, CA, USA). $\beta$-actin served as an internal protein control.

In this study, GraphPad Prism 8.0.1 (GraphPad Software, Santiago, CA, USA) was used as the statistical analysis software. All experiments were repeated at least three times and the corresponding data were expressed as mean $\pm$ SD. To compare differences in function assays between two groups or among multiple groups, T tests and one-way analysis of variance were conducted. p 0.05 was considered statistically significant.

The design process of this research is depicted in Fig. 1. In this study, we first analyzed HSF4 expression in the COAD and READ cohorts from the TCGA database, respectively, by using the online GEPIA tool. The results displayed that HSF4 was commonly overexpressed in tumor tissues compared to adjacent tissues (Fig. 2a). Moreover, HSF4 was also at a lower level in primary colon cancer than adjacent colon tissues [N colon (Mara, n = 32)] in an integrative analysis of multiple Gene Expression Omnibus (GEO) database (Fig. 2b). This study probed HSF4 protein in human CRC using IHC. As illustrated in Fig. 2c, HSF4 protein expression was enhanced significantly in CRC tissues, but not in adjacent tissues. To evaluate the predictive potential of HSF4 expression in CRC, we performed COX regressions. HSF4 expression (HR = 1.572/p 0.001), tumor stage (HR = 2.061/p 0.001), T (HR = 2.444/p 0.001), N (HR = 1.975/p 0.001), M staging (HR = 4.19/p 0.001), as well as age (HR = 1.028/p = 0.002) were related to overall survival (OS) of CRC patients in the TCGA cohort (Table 1). In the multivariate analysis, HSF4 expression remained to be associated with poor survival of CRC (HR = 1.297/p = 0.046, Table 1). OS was lower in high HSF4-expressing patients than in low HSF4-expressing subgroups (p = 0.036, Fig. 2d).

Fig. 1.

Flowchart of this study regarding the analysis and experiments. HSF4: Heat Shock Transcription Factor 4; CRC: Colorectal Cancer; CDS: Coding Sequences.

Fig. 2.

HSF4 is overexpressed in CRC and associated with poor prognosis. (a) HSF4 expression is significantly increased in tumors both in colon adenocarcinoma (COAD) and rectum adenocarcinoma (READ) cohorts. (b) HSF4 expression is up-regulated in CRC cohorts than the cohort of normal colonic tissues [N colon (Mara, n = 32)] in an integrative analysis of multiple Gene Expression Omnibus (GEO) database. T and N stand for COAD/READ and adjacent tissues, respectively. (c) Immunohistochemistry (IHC) analysis on 3 pairs of CRC samples reveals that HSF4 is strongly detectable in CRC tissues but weakly in adjacent colonic tissues. (d) Overexpressed HSF4 exhibits significant association with short survival term of patients by analyzing survival information of CRC patients in The Cancer Genome Atlas (TCGA) cohort (p = 0.036). * stands for p 0.05.

Further, logistic regression analysis was performed in this study to calculate the relationship between HSF4 expression and clinical characters of CRC patients. As shown in Table 2, overexpression of HSF4 was significantly correlated to tumor stage [OR = 1.498 for Ⅳ vs. (Ⅰ+Ⅱ+Ⅲ), p = 0.023], T classification [OR = 1.613 for T4 vs. (T1+T2+T3), p = 0.011], N classification [OR = 1.535 for N0 vs. (N1+N2), p = 0.001], M classification [OR = 1.532 for M1 vs. M0, p = 0.021].

Since clinical specimens demonstrated that HSF4 expression was closely associated with tumor development and survival, we performed in vitro cell assays to validate the effect of HSF4 on modulating cell behaviors of CRC. An HSF4-specific shRNA plasmid was constructed and introduced into HT29 and LoVo cells, respectively. The expression of HSF4 protein was reduced in HT29 and LoVo cells after 48 h of transfection (Fig. 3a). Functionally, it was shown that decrease of HSF4 mitigated cell growth (Fig. 3b), and inhibited cell colony formation (Fig. 3c). Similarly, the migratory and invasive capacities of HT29 and LoVo cells were impaired (Fig. 3d,e). This suggests that HSF4 is a facilitator for malignant biological behavior in CRC.

Fig. 3.

Knockdown of HSF4 induces inhibition of cell growth, migration and invasion of CRC cells. (a) A specific short hairpin RNA (shRNA) targeting HSF4 is validated to effectively downregulate HSF4 expression in HT29 and LoVo cells. Down-regulation of HSF4 contribute to restriction of cell growth (b) and colony formation (c). Downregulation of HSF4 significantly postpones cell invasion (d) and migration (e) of HT29 and LoVo cells.

MACC1 was firstly reported as a promotive gene involved in CRC metastasis, and subsequently has been verified to regulate multi-aspects of tumor progression across cancer entities [19]. We associated HSF4 with MACC1 by analyzing TCGA dataset of CRC cohort, as demonstrated by a positive correlation between HSF4 and MACC1 in CRC samples (Fig. 4a). HSF4 had a significant efficiency to predict prognosis in combination with MACC1. As displayed in Fig. 4b, low expression of HSF4 and MACC1 exhibited better OS than high expression of HSF4 and MACC1, while it is no difference of survival between HSF4 ${}^{\text{high}}$/MACC2 ${}^{\text{low}}$ and HSF4 ${}^{\text{low}}$/MACC2 ${}^{\text{high}}$ cohorts (Fig. 4c). This study evaluated the potential regulatory role between HSF4 and MACC1. In immunoblotting assay, the expression of MACC1 protein was down-regulated in HT29 and LoVo cells after 48 h of HSF4 shRNA transfection (Fig. 4d). This suggests that the promotional effect of HSF4 on CRC growth and metastasis is achieved by promoting MACC1 expression.

Fig. 4.

HSF4 positively correlates with MACC1 expression in CRC. (a) HSF4 and MACC1 exhibited positive relationship in CRC samples is demonstrated by analyzing TCGA dataset of CRC cohort (R = 0.19, p = 1.4 $\times$ 10${}^{-5}$). Concurrent expression of HSF4 and MACC1 has a significant efficiency to predict prognosis of CRC patients (p = 0.016) (b), however there is no difference of survival between HSF4${}^{\text{high}}$/MACC2${}^{\text{low}}$ and HSF4${}^{\text{low}}$/MACC2${}^{\text{high}}$ cohorts (p = 0.66) (c) in survival analysis of TCGA CRC dataset. (d) shRNA-induced HSF4 expression leads to a decrease of MACC1 expression level by immunoblotting.

miR-330-5p is expressed in different patterns in multiple cancer entities by using the OncomiR tool based on TCGA data. This implies the regulative role of miR-330-5p is dependent on tumor environment. Notably, miR-330-5p was declined in the COAD cohort than normal colon tissues (Fig. 5a). Moreover, qPCR assay showed that miR-330-5p was generally expressed at a decreased level commonly across CRC cells compared to normal colon FHC cells (Fig. 5b). Therefore, we exogenously regulated miR-330-5p in HT29 and LoVo cells to explain miR-330-5p effect on malignant phenotype of CRC. The results indicated that overexpression of miR-330-5p repressed cell growth during an observation of 72 hours and colony formation (Fig. 5c,d). In addition, it also reduced migration and invasion ability of CRC cells (Fig. 5e,f).

Fig. 5.

miR-330-5p is a tumor suppressive miRNA in CRC. (a) The expression pattern of miR-330-5p is bifacial in a variety of tumor types. Notably, miR-330-5p expression is decreased in COAD (colon cancer) by analyzing data extracted from TCGA using OncomiR tool. (b) qPCR assay shows that miR-330-5p is commonly decreased in six CRC cell lines in comparison with normal colonic FHC cells. (c) Restoration of miR-330-5p in HT29 and LoVo cells inhibits cell growth rate by a CCK-8 assay. (d) Restoration of miR-330-5p represses colony formation of HT29 and LoVo cells. Restoration of miR-330-5p apparently postpones cell migration and invasion of HT29 and LoVo cells by wound healing (e) and transwell assays (f). COAD: Colon adenocarcinoma; THCA: Thyroid carcinoma; BLCA: Bladder Urothelial Carcinoma; BRCA: Breast invasive carcinoma; ESCA: Esophageal carcinoma; STAD: Stomach adenocarcinoma; FDR: False discovery rate.

It has been recently emerged that miRNAs can target CDS of mRNA in addition to targeting 3${}^{\prime }$UTR, to regulate translation of mRNA in mammalian cells [6, 7]. Interestingly, we predicted three putative miRNA-recognition elements (MREs) of miR-330-5p within CDS region of HSF4 by using mircoT-CDS algorithm, while there was none of putative binding sites of miR-330-5p in 3${}^{\prime }$UTR of HSF4 mRNA. In accordance to the inhibitive cellular effects of miR-330-5p on CRC, we assumed that HSF4 may be negatively regulated by miR-330-5p. The three predicted MREs of miR-330-5p within CDS of HSF4 is as shown in Fig. 6a. By immunoblotting, overexpression of miR-330-5p markedly inhibited HSF4 expression in LoVo and HT29 cells (Fig. 6b). Dual-luciferase reporter assay was performed to investigate if miR-330-5p targets CDS of HSF4 mRNA. Assay results expectedly showed that miR-330-5p attenuated the luciferase activity of wild-type CDS sequence of HSF4 harboring all the three putative MREs but not the mutant one (Fig. 6c).

Fig. 6.

miR-330-5p inhibits HSF4 expression by recognizing its coding sequences (CDS) region. (a) By online prediction using microT-CDS software, three MREs are found within CDS region of HSF4 mRNA. (b) Restoration of miR-330-5p significantly inhibits HSF4 expression by immunoblotting in both HT29 and LoVo cells. (c) miR-330-5p is observed to weaken luciferase activity of wild-type CDS of HSF4 harboring the three predicted MREs, whereas the mutant CDS of HSF4 remedies this inhibition of luciferase activity.

To determine the suppressive effect of miR-330-5p on CRC via targeting HSF4, we performed function rescue experiment by introducing HSF4-expressing vector into LoVo and HT29 cells under ectopic miR-330-5p expression. In mechanism, as previously described, MACC1 was closely involved in augmenting HGF/c-MET signaling. Considering MACC1 expression was associated with HSF4 in CRC and downregulated along with HSF4 knockdown, we anticipated miR-330-5p may indirectly influence downstream cascades of HGF/c-MET signaling pathway. In immunoblotting assay, restoration of HSF4 expression in miR-330-5p expressing cells was validated (Fig. 7a). Moreover, we expectedly found overexpression of miR-330-5p apparently declined MACC1 expression, and more importantly, impaired the activity of phosphorylation of STAT3 which represents a crucial branch downstream HGF/c-MET pathway (Fig. 7a). In contrast, restoration of HSF4 expression in the presence of miR-330-5p partially abolished MACC1 inhibition and p-STAT3 inactivation (Fig. 7a). As expected, restoration of HSF4 expression partially mitigated miR-330-5p-induced cell growth inhibition in CRC cells (Fig. 7b). Similarly, HSF4 alleviated miR-330-5p-induced repression of colony formation, migration, and invasion (Fig. 7c–e) of CRC cells. These findings of rescue assays suggest miR-330-5p plays inhibitory role in regulating cell behaviors of CRC by downregulating HSF4-mediated the activity of the MACC1/STAT3 pathway.

Fig. 7.

miR-330-5p attenuates activity of MACC1/STAT3 signaling via suppressing HSF4 expression. HSF4 expression in the context of miR-330-5p expression is restored and validated via immunoblotting. In a series of function assays, restoration of HSF4 expression abolishes miR-330-5p-induced inhibition of cell growth (a), colony formation (b), migration (c) and invasion ability (d) of HT29 and LoVo cells. (e) An immunoblotting assay shows that, upon ectopic miR-330-5p expression, expression of MACC1 is inhibited as well as phosphorylation activity of STAT3. Reversely, restoration of HSF4 expression can partly alleviates the inhibition of MACC1/STAT3 signaling pathway led by miR-330-5p.