Antagonism between HTRA3 and TGFβ1 Contributes to Metastasis in Non–Small Cell Lung Cancer

Non–small cell lung cancer (NSCLC), which comprises the majority of lung cancers, remains one of the most common causes of cancer deaths worldwide (1). Chemotherapy, target therapy, and emerging immunotherapy are the main treatments for advanced stage NSCLCs, yet the 5-year relative survival remains grim mostly due to local and distant metastasis (2). Metastasis is a multistep process that requires cancer cells with versatile self-reprogramming capabilities to switch from the epithelial to mesenchymal state (3, 4). Epithelial–mesenchymal transition (EMT) is of critical importance in the early events of lung cancer cell metastatic dissemination, and TGFβ1 is involved (5). TGFβ1 has dichotomous functions in tumor progression. In tumor initiation, TGFβ1 acts as a suppressor by inducing cell differentiation and apoptosis to prevent uncontrollable proliferation and transformation. In late-stage cancer, the TGFβ1/TGFβ receptor type II/TGFβ receptor type I complex induces SMAD-dependent and -independent pathways, which drive EMT remodeling, leading to tumor metastasis (6, 7). Elevated levels of TGFβ1 are usually detected in NSCLC tissue. Thus, the complex and often contradictory functions of TGFβ1 necessitate both a better understanding of the special downstream effectors and a search for specific inhibitors of different TGFβ-related pathways for NSCLC treatment.

High temperature requirement A3 (HTRA3) is a member of the HtrA family, and it was first characterized as an upregulated factor during mouse uterine placentation (8). Two transcriptional variants of HTRA3 mRNA (long and short) have been identified corresponding to two HTRA3 protein isoforms (49 and 39 kDa, respectively) produced through alternative splicing (9). The long isoform (HTRA3-L) has four distinct domains: the insulin-like growth factor binding (IGFB) domain, Kazal-type protease inhibitor domain, trypsin-like serine protease domain, and postsynaptic density protein 95-Discs large-Zona occludens 1 (PDZ) domain. The 39-kDa short isoform (HTRA3-S) lacks the PDZ domain at the C-terminus and instead has a unique sequence of seven amino acids, which is encoded by a separate exon (10). Downregulation of HTRA3 is observed in several cancers (11, 12). The underlying mechanisms for its reduction are not fully understood, but a lung cancer study proposed that it may be related to cigarette smoke-induced methylation within the first exon of HTRA3 (13). In addition, HTRA3 proteolytic activity is indispensable for lung cancer cell death caused by chemotherapeutic drugs (14). Furthermore, HTRA3 is negatively correlated with lymph node metastasis in invasive mammary tumors (15). Inhibition of HTRA3 triggers HTR-8 cell invasion, and loss of HTRA3 stimulates invasion by interstitial trophoblasts (16). Given these studies, HTRA3 could be a tumor suppressor that contributes to invasive cell characteristics. To date, the specific roles and molecular events of HTRA3 in invasion and metastasis are not well identified.

A previous mouse study demonstrated that HTRA3 is expressed in a distinct set of embryonic tissues in which development is largely regulated by the TGFβ family proteins and that is has inhibitory activity on TGFβ signaling (17). ΔN HTRA3 in mouse cleaves small, leucine-rich proteoglycans, such as decorin and biglycan, which are known to bind several specific TGFβ proteins and regulate their activities (18, 19). However, the correlation between HTRA3 and TGFβs in human cancer and their mechanistic interactions have not been further addressed.

In this study, we focus on the interaction between HTRA3 and TGFβ1 in NSCLC, hoping to shed light on the action of HTRA3 in controlling the invasion-metastasis cascade of lung cancer.

Human lung adenocarcinoma HCC827, HCC4006, A549, and H358 cell lines and mouse breast cancer 4T1 cell line were obtained from the ATCC and cultured in high glucose DMEM or RPMI1640 supplemented with 10% FBS. Cell lines were independently validated by short tandem repeat (STR) DNA fingerprinting at Chinese Academy of Sciences, and tests for mycoplasma infection were negative.

Human recombinant TGFβ1 was purchased from PeproTech. The antibodies against HTRA3, PD-L1 (human), PD-L1 (mouse) were obtained from Abcam. Anti-TβRII, anti-SMAD2, anti-phospho-SMAD2 (Ser 465/467), anti-SMAD3, anti-phospho-SMAD3 (Ser423/425), anti-ERK, anti-phospho-ERK (T202/Y204), anti-AKT, anti-phospho-AKT (S473), anti-snail, anti-slug, anti-c-Jun, anti-vimentin, and anti-Ki67 antibodies were purchased from Cell Signaling Technology. Anti-E-cadherin antibody was from BD Biosciences, and anti-TβRI and anti-GAPDH antibodies were obtained from Santa Cruz Biotechnology. The TGFβ1 signaling inhibitor SB431542, the proteasome inhibitor MG-132, the cysteine proteases inhibitor MG-101 (ALLN), and the serine and cysteine proteases inhibitor LeupeptinHemisulfate were purchased from Selleckchem.

Archived tissues for tissue microarray (TMA) construction were obtained from a cohort of 150 patients who received curative resection of NSCLC between 2009 and 2017 in Zhongshan Hospital affiliated to Fudan University. TMAs were constructed by Shanghai Outdo Biotech Co, Ltd. The postsurgical patient surveillance was similar to those specified in previous reports (20). Disease-free survival (DFS) was measured from the date of surgery until tumor relapse or death, and overall survival (OS) was defined as the interval between surgery and either death or the last observation taken. The ethical committee of Zhongshan Hospital approved the current research and written informed consent was obtained for each patient.

The human full-length cDNA of long HTRA3 isoform (HTRA3-L) was purchased from Obio Technology, and subcloned into the lentiviral expression plasmid pLenti-CMV-EGFP-PGK-Puro. The HTRA3-L expression vector or empty vector was transfected into HEK 293T cells according to the instructions to produce recombinant lentivirus. Then, we generated A549, H358, and 4T1 cell lines stably overexpressing HTRA3-L and corresponding vector control cell line as detailed previously (21). Short (PDZ-deleted) HTRA3 isoform (HTRA3-S) was also subcloned and cells stably overexpressing HTRA3-S were generated as well.

The Matrigel invasion assay was carried out according to previously described (21). Five ng/mL TGFβ1 was added to both the upper and the lower chambers. For wound healing assays, cells were seeded into 12-well plates and when cell confluence had reached about 90%, a 200 μL pipette tip was used to make wounds across the cell monolayer. Wounded cells were treated with TGFβ1 for 24 hours and wound healing was observed and photographed within the inverted phase-contrast microscope at different time points. Duplicate wells for each condition were examined and each experiment was repeated three times.

Western blotting was described previously (21). GAPDH was used as the protein loading control and all experiments were performed at least three independent times.

Cells were cultured on glass slides and treated as indicated. Cells were quickly rinsed in PBS, followed by fixing with 4% paraformaldehycle and permeabilizing in 0.1% TritonX-100. Nonspecific sites were blocked by incubation with 3% BSA. Then the samples were probed with appropriate primary antibody (anti-E-cadherin, 1:100; anti-Vimentin, 1:200; anti-p-Smad2, 1:200; anti-p-Smad3, 1:200), followed by anti-mouse Alexa Fluor 488 or anti-rabbit Alexa Fluor 594 secondary antibodies. The coverslips were mounted in mounting medium with DAPI. The fluorescence images were obtained with a Leica confocal microscope (Mannheim, Germany).

IHC for HTRA3, TGFβ1, and PD-L1 were performed in TMAs. The expressions were classified as either high or low with a cut-off value of 6 (score of 6–12 as high expression, score of 0–5 as low expression; ref. 22). Besides, PD-L1 and Ki-67 expressions were performed with monoclonal rabbit anti-mouse PD-L1 antibody (dilution, 1:200) or anti-mouse Ki-67 antibody (dilution, 1:200). The protocol was as previously described (20).

Four-week-old female BALB/c mice were purchased from Shanghai SLAC Laboratory Animal Co., Ltd., and were allowed to acclimate for 1 week under pathogen-free conditions. All experiments were performed according to the institutional ethical guidelines on animal care and approved by the Animal Ethics Committee in Ruijin Hospital. Mice were randomly divided into three groups: 4T1-HTRA3-L group, 4T1-HTRA3-S group, and control group. Then a suspension of 4.0 × 10⁵ cells in 0.1 mL of PBS was inoculated into the mammary gland of female mice in these three groups accordingly. After 4 to 5 weeks, the mice were scanned by Siemens Inveon micro-PET/CT to examine primary and metastatic tumors.

Prior to micro-PET/CT imaging, the mice were fasted overnight, with water being available ad libitum. On the day of imaging, the mice were weighted and then prewarmed to a body temperature of 37°C before an approximate 100 to 140 μCi of ¹⁸F-fludeoxyglucose (FDG) was administered by tail vein. The body temperature was maintained at 37°C throughout the uptake period. Imaging examination was performed 50 minutes after ¹⁸F-FDG injection. Each mouse was imaged 10 minutes for CT scanning and another 5 minutes for PET scanning. During the micro-PET/CT imaging, the mice were placed in an imaging chamber and kept under isoflurane anesthesia. The imaging was analyzed by Inveon Research Workplace and the region of interest was obtained. Standardized uptake values (SUV) of region of interest was calculated and compared among different groups. After micro-PET/CT analysis, all mice were sacrificed by carbon dioxide asphyxiation and the breast tumors and lungs were removed. In situ tumor volumes were measured by the following formula: ab²/2 (where a and b refer to the largest and smallest dimensions). To examine the metastasis, 50 sequential sections (5 μm) were cut from the lungs and every 10th section was stained with hematoxylin and eosin according to standard protocols.

To evaluate anti-PD-1 treatment, BALB/c mice were randomly inoculated with 4T1-HTRA3-L or 4T1-vector cells as described above. PD-1mAb (BioXcell) was administered via intraperitoneal injection at 10 mg/kg twice a week. Breast tumor volumes and metastatic nodules in lungs were evaluated. PD-L1 and Ki-67 expressions were examined in tumors in situ.

Total RNA was isolated with the MiniBEST Universal RNA Extraction Kit (Takara), and reverse transcribed to cDNA using a PrimeScript RT Master Mix (Takara) following the supplier's instructions. The quantitative real-time PCR was done with sequence-specific primer pairs using an ABI ViiA 7 Sequence Detector System(Applied Biosystems), and SYBR Premix Ex Taq (Takara). The following primer pairs were listed in Supplementary Table S1. All reactions were done in triplicate. Relative expression levels and SDs were calculated using the comparative quantification.

A total of 1 × 10⁶ lung cancer cells were harvested and stained for 30 minutes at 4°C with primary antibody to anti-human-PD-L1. Samples were analyzed and sorted on a BD FACS Aria (BD Biosciences).

siRNA sequences specifically targeting human TβRI, TβRII, c-Jun, Sp1, FosB, SMAD2, and SMAD3 were purchased from Shanghai GenePharm. Cells were transfected with either target-specific siRNAs or a scrambled control siRNA using Lipofectamine RNAi MAX reagent (Life Technologies) according to the manufacturer's instructions.

HCC827 cells were seeded in 24-well plates at a density of 1.0 × 10⁵ cells/well. After an overnight incubation, cells were transiently cotransfected with either reporter construct pGL3-HTRA3-pro-Wt or pGL3-basic (1.5 μg/well) and c-Jun plasmid (0–3.0 μg/well). The Renilla luciferase reporter plasmid was included as an internal control. After 16 hours of transfection, cells were harvested and luciferase activity was measured with the dual luciferase reporter assay system (Promega) according to the manufacturer's recommendations. Firefly luciferase values were normalized to Renilla luciferase activity and are presented as the mean ± SD of three independent experiments.

The chromatin immunoprecipitation (ChIP) assay was performed using SimpleCHIP Plus Enzymatic Chromatin IP Kit (Cell Signaling Technology) according to the manufacture's protocol. Antibodies for c-Jun, positive control H3, negative control IgG were used. Real-time qPCR reaction along with DNA agarose gel was performed to quantify target DNA fragment. Seven sets of primers that cover HTRA3 promoters from about −3 kb to 0 were synthesized, and the relative average ChIP intensities of each primer were analyzed.

The Human Phospho-Kinase Array Kit (#ARY003B) was purchased from R&D Systems. Array screening was performed following the manufacture's protocol.

Statistical analysis was performed with SPSS 19.0 (IBM) and GraphPad Prism (GraphPad Software Inc.). Data are presented as mean ± SD of three independent experiments. Quantitative data between groups were compared using Student t test. Categorical data were analyzed by the Chi-square test. Correlation analysis was performed among HTRA3, TGFβ1, and PD-L1 expressions in TMAs. DFS and OS were calculated by the Kaplan–Meier method and differences were analyzed by the log-rank test. All tests were two-tailed and P ≤ 0.05 was considered statistically significance.

To determine the potential correlation between HTRA3 and TGFβ1 in lung cancer, IHC staining was used to examine the two proteins simultaneously in TMAs with 150 pairs of NSCLC tumor and adjacent normal tissues. The clinical and pathologic characteristics are shown in Table 1. According to N and M stages, we categorized tumor specimens into nonmetastatic and metastatic groups. Statistically, HTRA3 was frequently downregulated at the protein level in NSCLC tissues and tended to have decreased expression in metastatic subtypes (P < 0.001, Fig. 1A). The levels of TGFβ1 in tumors were significantly higher than in matched normal tissue, but with no significant difference regardless of whether the lung cancer was metastatic or not (P = 0.722, Fig. 1B). In 78 tumor cases without nodal or distant metastasis, a total of 66.7% of the cases harbored elevated HTRA3 expression; however, the level of HTRA3 was not statistically correlated with that of TGFβ1 (r = 0.135, P = 0.239; Fig. 1C–E). In 72 metastatic cases, we observed a strong negative correlation between HTRA3 and TGFβ1 expression, as over 50% of HTRA3-low samples had a high TGFβ1 level (r = −0.400, P = 0.002; Fig. 1F–H).

We next investigated the prognostic value of HTRA3 and TGFβ1. Consistent with its role as a tumor suppressor, higher HTRA3 expression was correlated with longer DFS (P < 0.001) and OS (P < 0.001; Fig. 1I; Supplementary Fig. S1A). Notably, in high-HTRA3 cases, patients' survival was not significantly different in samples with varied TGFβ1 levels (P = 0.930 and 0.858, respectively). In low-HTRA3 cases, we noticed that TGFβ1 overexpression was significantly associated with decreasing DFS and OS (P = 0.045 and 0.018, respectively). We separately examined the expression of the two proteins and relevant clinical outcomes in both adenocarcinoma and squamous cell carcinoma and found consistent results (Supplementary Fig. S1B–S1R). A relatively stronger correlation was detected between HTRA3 and TGFβ1 in metastatic adenocarcinomas (r = −0.444, P = 0.011).

Because smoking status is significantly correlated with some driver mutations in lung cancer, we also accessed the association between smoking history and HTRA3 expression. Seventy-four percent (56 of 76) of tumors from past and current smokers had low HTRA3 expression. Likewise, 62% (46 of 74) of tumors from nonsmokers had relatively high levels of HTRA3 (P < 0.001; Fig. 1J). We further found that in smokers with low HTRA3 expression, 44.7% had high levels of TGFβ1 whereas only 29% had low TGFβ1 levels. In nonsmoking patients with high HTRA3 expression, there was not a remarkable difference in TGFβ1 expression levels (28.4% vs. 33.8%, Fig. 1K).

Overall, we concluded that HTRA3 was negatively correlated with TGFβ1 in a metastatic context; high HTRA3 levels were associated with better prognosis in NSCLC independent of TGFβ1 expression levels. In addition, we detected a significant difference in smoking status between the samples with low- and high-HTRA3 expression.

To explore the role of TGFβ1 in HTRA3 expression, we treated two NSCLC cell lines with exogenous TGFβ1. In both HCC827 and HCC4006 cells, we detected a decrease in the long variant of HTRA3 (HTRA3-L) protein along with expected changes in EMT markers; these changes were reversed by SB431542, a selective inhibitor of TGFβ1 signaling. The short variant of the HTRA3 (HTRA3-S) protein did not change upon TGFβ1 treatment (Fig. 2A; Supplementary Fig. S2A). This observation was also confirmed at the mRNA level (Fig. 2B; Supplementary Fig. S2B). Knockdown of TβRI or TβRII, the membrane receptors of the TGFβ1 ligand, also downregulated HTRA3-L (Fig. 2C). Two cell lines were exposed to various concentrations of TGFβ1, and the expression of HTRA3-L was inhibited at all concentrations from as low as 1.25 ng/mL. Meanwhile, we found a marked reduction in HTRA3-L protein levels at 48 and 72 hours exposures to TGFβ1. However, no significant changes were observed in HTRA3-S at any dose or time after treatment (Supplementary Fig. S2C–S2J). To gain further insight, we used a series of protein degradation inhibitors—MG132, leupeptin and ALLN—together with TGFβ1 to treat HCC827 cells. HTRA3-L downregulation in this condition was similar to that with TGFβ1 only, suggesting posttranslational degradation may not be involved in TGFβ1-mediated downregulation of HTRA3-L (Fig. 2D; Supplementary Fig. S2K–S2M). These results revealed that reduced HTRA3 expression was induced by TGFβ1 and accompanied by EMT induction.

To further confirm the mechanism by which TGFβ1 regulates HTRA3-L expression, we identified transcriptional factors predicted to modulate HTRA3-L using a publicly available database (TRANSFAC, JASPAR, and GeneCards). We found several binding sequences for transcription factors associated with TGFβ1 as previous studies indicated, including c-Jun, Sp1, Runx3, and FosB (23–26). We performed preliminary testing to evaluate which may be involved in regulating HTRA3-L expression (Supplementary Fig. S3A–S3D). As immunoblotting showed, TGFβ1 treatment increased c-Jun levels, whereas c-Jun knockdown abolished TGFβ1-induced HTRA3-L reduction. Luciferase reporter assays ascertained the effects of c-Jun on HTRA3-L transcription that exogenous c-Jun induction significantly suppressed transcriptional activity of HTRA3-L in a dose-dependent manner (Fig. 2E). Furthermore, TGFβ1 treatment induced a nearly three-fold suppression of luciferase expression driven by the HTRA3 promoter, also demonstrating that TGFβ1 decreased HTRA3 expression through transcriptional interactions (Fig. 2F).

We then performed ChIP assay to validate the direct binding of c-Jun to the HTRA3-L promoter. We surveyed the whole promoter region of HTRA3-L from −3,000 to 0 bp upstream of the transcription start site using seven sets (approximately 500 bp covered per set) of primers containing a total of 17 pairs. According to qPCR analysis, the occupancy of c-Jun was significantly enriched in areas covered by three distinct pairs of primers from −2,200 to −1,900 bp; these primers were named a, b, and c (Fig. 2G and H; Supplementary Fig. S3E and S3F). The results of qPCR revealed that c-Jun recruitment was significantly increased upon TGFβ1 treatment and decreased upon c-Jun knockdown (Fig. 2I; Supplementary Fig. S3G–S3I). As a whole, we concluded that TGFβ1 enhanced expression of c-Jun, which may directly bind to HTRA3-L and inhibit its transcription and thus expression (Fig. 2J).

To contrarily explore whether HTRA3 functions in TGFβ1-related biological and EMT processes, we used A549 and H358 cells to generate cell lines stably overexpressing HTRA3-L, HTRA3-S, and their corresponding vector controls (Supplementary Fig. S4A and S4B). We also generated stably transformed 4T1 cell lines for a spontaneous cancer model (Supplementary Fig. S4C). The 4T1 cell line secretes excessive TGFβ1 and exhibits basal levels of SMAD2/3 phosphorylation, suggesting a cell autonomous mechanism of TGFβ1/SMAD signaling (27). TGFβ1 treatment significantly increased the migratory and invasive potentials of vector controls in A549, H358, and 4T1 cells. However, overexpression of HTRA3-L resulted in potent migratory and invasive inhibition as assessed by wound-healing assay and Matrigel invasion analysis (Fig. 3A and B; Supplementary Fig. S4D and S4E). Elevated HTRA3-S in three cell lines did not affect TGFβ1-induced migration and invasion. We then used fludeoxyglucose-PET (FDG-PET) and CT scans to evaluate metastatic growth in living mice. Micro-PET/CT scans showed successful tumor implantation for all inoculated mice. Compared with vector groups, 4T1-HTRA3-L mice exhibited less extensive tumor spreading and relatively low levels of SUVs (Fig. 3C and D). The average SUVs between vector and HTRA3-S mice were not statistically difference (P = 0.706). In line with the PET/CT analysis, HTRA3-L overexpression significantly reduced nodule formation in the lungs (Fig. 3E and F; Supplementary Fig. S4F and S4G). Importantly, there was no significant alteration in primary tumor size among vector- and HTRA3-overexpression mice. Together, our data demonstrated that high levels of HTRA3 are a barrier to the TGFβ1-induced invasion-metastasis cascade.

To determine the molecular mechanism of HTRA3-L in suppressing TGFβ1-induced cell motility, we further analyzed the effect of HTRA3-L on TGFβ1-induced EMT in A549 and H358 cells. TGFβ1 exposure for a period of 24 hours induced an EMT in both cell lines carrying vector controls, as evidenced by the elongation of cell bodies, reduced E-cadherin level, and increased vimentin level. However, the morphology in HTRA3-L–overexpressing cells remained nearly unchanged after TGFβ1 treatment, and they had relatively small changes in EMT markers (Fig. 4A and B; Supplementary Fig. S4H and S4I). A series of signaling pathways have been associated with EMT induction, including SMAD2/SMAD3 signaling, the MEK/ERK axis, and the PI3K/AKT pathway, all of which have well-validated roles in EMT modulation (28). Thus, we next examined the status of these pathways upon HTRA3-L influence. As immunoblotting and immunofluorescence staining showed, HTRA3-L overexpression attenuated phosphorylation of SMAD2/3; the phosphorylation of MEK, ERK, and AKT were barely affected in cells treated with TGFβ1, largely excluding SMAD-independent signaling (Fig. 4C–F; Supplementary Fig. S4J–S4L). We also found that HTRA3-L decreased the TGFβ1-induced expression of Snail and Slug, which are transcriptional factors that bind to E-box elements in the promoter of E-cadherin and lead to transcriptional repression and induction of the mesenchymal phenotype. Thus, our results indicated that HTRA3-L functions as a brake on SMAD2 and SMAD3 activation in response to TGFβ1 (Fig. 4G). This brake prevents cancer cells from undergoing the mesenchymal transition and attenuates the invasion-metastasis process.

Previous studies revealed that TGFβ1 and related mesenchymalization is responsible for PD-L1–dependent immunosuppression and resistance to the anti-PD-1 treatment (29). Here, we sought to determine whether HTRA3 functions in this immune process. First, we found that extrinsic addition of TGFβ1 increased the expression of PD-L1, which was abrogated by HTRA3-L overexpression (Fig. 5A; Supplementary Fig. S5A). Correspondingly, we transiently knocked down SMAD2 and SMAD3, and the addition of TGFβ1 had no effect on PD-L1 expression; these suggest that SMAD2/3 is required for PD-L1 upregulation (Fig. 5B; Supplementary Fig. S5B). Of note, we detected PD-L1 expression in HTRA3-L–overexpressing cells and found no difference compared with the vector control (Fig. 5C; Supplementary Fig. S5C). 4T1-HTRA3-L and its corresponding vector cells were transplanted into BALB/c mice to establish immunocompetent models; these mice were then exposed to the PD-1 mAb. Compared with the 4T1-vector group, HTRA3-L-bearing mice showed a greater reduction in primary tumor size and number of lung metastatic nodules upon PD-1 mAb treatment (Fig. 5D; Supplementary Fig. S5D). Ki-67 staining of primary tumors by IHC demonstrated that proliferation was more severely inhibited by PD-1 mAb in HTRA3-L mice (Fig. 5E). Besides, elevated HTRA3-L downregulated the expression of PD-L1, which was consistent with the in vitro findings (Supplementary Fig. S5E). Mice in the HTRA3-L overexpressing group had better survival with anti-PD-1 treatment than those in the vector group (P = 0.05, Fig. 5F). We also detected PD-L1 expression in the same TMAs utilized for the analysis of HTRA3, and a negative correlation was found between PD-L1 and HTRA3 levels (P = 0.018; Fig. 5G and H; Supplementary Fig. S5F). Altogether, our results demonstrated that high level of HTRA3-L eliminated TGFβ1-induced PD-L1 upregulation and rendered cancer cells more sensitive to anti-PD-1 treatment, indicating a role of HTRA3 in TGFβ1-related tumor immunosuppression. Because HTRA3-L attenuated TGFβ1-induced SMAD2/3 activation, we hypothesized it may also inhibit TGFβ1-induced PD-L1 upregulation upon SMAD2/3 involvement.

EMT is reported to confer resistance to target therapies as well. Thus, we additionally tried to address whether HTRA3-L affected the efficacy of EGFR-TKIs (Gefitinib and Osimertinib). In the A549-vector cell line, the IC₅₀ for the two drugs was elevated upon exposure to TGFβ1. In the A549-HTRA3-L cells, the increase in IC₅₀ induced by TGFβ1 was much lower. However, the above analysis did not detect significant differences, possibly because A549 cells are not sensitive to EGFR-TKIs (Supplementary Fig. S5G).

The human HTRA3 protease is involved in placentation, mitochondrial homeostasis, stimulation of apoptosis, and tumor suppression (30). Molecular mechanisms underlying HTRA3 functions are poorly understood, and knowledge concerning its cellular targets is very limited. In this study, we uncovered that decreased levels of HTRA3 in metastatic lung cancer is related to activated TGFβ1, which facilitates c-Jun binding to the HTRA3-L promoter and inhibiting its transcription. However, HTRA3-L overexpression negatively regulates TGFβ1-induced EMT and cell metastasis by suppressing SMAD2/3 activation. This is the first evidence of HTRA3′s important role in controlling TGFβ1 signaling in NSCLC.

Our results indicated the varied correlation of HTRA3 and TGFβ1 in metastasis and nonmetastasis contexts. In the metastatic subgroup, enhanced TGFβ1 is correlated with low HTRA3 levels. In the nonmetastatic subgroup, increased HTRA3 is not negatively associated with TGFβ1 expression. Notably, there are no significant differences in the level of TGFβ1 among tissues with or without metastasis. As such, a serial of factors may determine TGFβ's tumor metastasis promoter functions, including extracellular matrix (ECM) and select transcription factors and protein families (31, 32). Some previous studies reported that HTRA3 exhibits substrate specificity towards certain ECM proteoglycans such as decorin, biglycan, and fibronectin (17). ΔN-HTRA3 cleaves decorin and biglycan, whereas TGFβ has a mutual functional interaction with these molecules (33, 34). Thus, we hypothesize that HTRA3 may modulate ECM factors in the tumor microenvironment to halt TGFβ's oncogenic effects and metastasis in early-stage lung cancer. In late-stage carcinomas, such modulation is weakened, and TGFβ1 regains its carcinogenicity and promotes metastasis (Supplementary Fig. S6). However, the underlying mechanisms of HTRA3 in regulating the dichotomous functions of TGFβ1 are still unclear and require future studies.

HTRA3 attenuated TGFβ1-mediated EMT and oncogenic effects through the suppression of phosphorylation of SMAD2/3 in our study. In an attempt to explore the kinase acting downstream of HTRA3 upon TGFβ1 exposure, we performed a phospho-kinase antibody array screen to identify kinases that are upregulated or downregulated compared with corresponding control. We found that phosphorylated STAT3 (Y705) scored as the highest downregulation and HSP60 as the upregulation (Supplementary Fig. S7). STAT3 is a member of the STAT protein family. It is phophorylated by receptor-associated JAK, forms homo- or hetero-dimers, and translocates to the cell nucleus where it acts as transcription activators (35). Specially, STAT3 becomes activated in response to such ligand as interferons, EGF, and inflammatory factors (IL5 and IL6), indicating HTRA3 may also serve a role in inflammatory reaction (36). Meanwhile, accumulating evidence suggests that STAT3 and SMAD engage in crosstalk in various biological contexts, including tumorigenesis, EMT, and T-cell differentiation (37). HSP60 belongs to heat shock proteins. The function of HSP60 is quite similar to part of HTRA3′s—facilitating the correct folding of imported proteins and preventing misfoldings (38). Some studies in cancer reveal a loss of HSP60 indicates a poor prognosis and the risk of developing tumor infiltration, whereas its correlation with HTRA3 remains further elucidated (39, 40).

Our results also indicate a potential association between HTRA3 and PD-L1 expression. HTRA3-L inhibits TGFβ1-induced upregulation of PD-L1 and even sensitizes cells to PD-1 treatment. Recent advances have revealed a mechanistic link between EMT and immune evasion through the elevation of multiple immune checkpoints (41–43). A novel mechanism of tumor intrinsic regulation of PD-L1 has been uncovered in lung cancer linking EMT to cytotoxic T-cell dysfunction and metastasis (44). A study focusing on circulating tumor cells (CTC) found that CTCs with mesenchymal transitions are associated with anti-PD-1 resistance (29). We found SMAD2/3 was required for PD-L1 upregulation. Accordingly, recent two reports identified that TGFβ1-driven PD-1/PD-L1 expression is associated with binding of SMAD2/3 to its promoter regions. In response to TGFβ1, SMAD2/3 is activated and binds to CR-C, one of the conserved upstream regulatory regions hypersensitive to DNase I that control PD-1 expression in response to CD8 T cells. Meanwhile, the chromatin organizer Satb1 is proved to regulate in this process (45, 46). Thus, we deduced that HTRA3′s ability in sensitizing PD-1 treatment and involvement in TGFβ1-related tumor immunosuppression may be associated with its brake to SMAD2/3 activation and subsequent EMT.

HTRA3 is unique among the HtrAs because it has two isoforms. It is not yet known whether there is a functional difference between the HTRA3-L and HTRA3-S isoforms. HTRA3-L harbors a unique PDZ domain, which is necessary for HTRA3 trimer formation and important regulatory functions (10). The PDZ domain is not required for proteolysis as the in vitro proteolytic activity of the two variants is similar (47). However, a previous study in lung cancer found that HTRA3-L is more efficient than HTRA3-S in promoting cancer cell apoptosis (14). Moreover, the PDZ domain is necessary for recruitment of TβRI and PDZ-containing proteins, such as scribble (Scrib), discs-large (Dlg), and lethal giant larvae (Lgl), which regulate metastasis (48, 49). Our results suggest that the long variant is more closely correlated with TGFβ1 and metastasis. Further research is warranted to elucidate the specific significance of each isoform in the clinical context in NSCLC.

In summary, we demonstrate that HTRA3 is a novel mediator of TGFβ1 signaling pathway and reversely reveal roles of TGFβ1 in regulation HTRA3 in the early and late periods of lung cancer's invasion-metastasis cascade. The findings here have significant implications regarding our understanding the links between HTRA3 and TGFβ1. HTRA3 suppresses TGFβ1-induced EMT and sensitizes PD-1/PD-L1 immune therapy, suggesting that HTRA3 could be an effective anticancer target in the battle against NSCLC.

No potential conflicts of interest were disclosed.

Conception and design: M. Feng, J. Qu

Development of methodology: J. Zhao, J. Zhang, J. Qu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Zhao, M. Feng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Zhao, M. Feng, D. Liu, H. Liu, J. Zhang

Writing, review, and/or revision of the manuscript: J. Zhao, M. Shi, J. Qu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Zhao, M. Shi, J. Zhang

Study supervision: J. Zhang, J. Qu

The authors would like to thank Dr. Ming Lu from Shanghai Jiaotong University School of Medicine for reviewing and revising the manuscript. The authors are also grateful to Dr. Meiyu Geng and Jing Ai from Shanghai Institute of Materia Medica, Chinese Academy of Sciences, as well as other members of Dr. Meiyu Geng's laboratory, for helpful suggestions and discussions. This study was funded by the Shanghai Key Discipline for Respiratory Diseases (No. 2017ZZ02014; J. Qu was directly supported by the grant) and the National Natural Science Foundation of China (No. 81802258; J. Zhao was directly supported by the grant).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.