The study population included 56 LARC patients (39 male and 17 female), with a median age of 57.6 years (41.3-72.7 years), who underwent total mesorectal excision (TME) and refused to accept nCRT in our hospital from July 2015 to July 2016. Postoperative pathology showed that among the 56 patients, 37 had pT3 and 19 had pT4 tumors; in addition to having pT3 and pT4 tumors, 12 patients had pN0, 27 had pN1, and 17 had pN2 diseases. There were 43 patients with R0 resection and 13 with R1 resection. Among them, 29 patients completed more than 50Gy postoperative radiotherapy, and 36 completed more than 6 cycles of chemotherapy with fluorouracil based chemotherapy. As of March 2021, 36 (64.28%) patients have had local recurrence and/or distant metastasis and 31 (55.36%) patients have died.

After the surgical specimen of enrolled patients was isolated, a fresh section of tumor tissue, approximately 10×10×5mm³ in size, was immediately cut from the central area of the tumor in a sterile environment ( Figure 1A ). It was then divided into two parts after being washed three times with normal saline. One part was divided into two parts and stored at -150°C refrigerator before being used for miRNA microarray analysis. The other part was cut into 2×2×2mm³ segments of tissues and implanted into the dorsal side of the forelimb or hindlimb of BALB/c-nude mice to construct the first generation PDX model. Two-four nude mice were implanted for each specimen, and 1-4 points were implanted for each nude mouse ( Figure 1B ). After the xenograft tumor grew to approximately 10mm in diameter, it was then removed to build a second-generation PDX model on the forelimb shoulder-back of nude mice ( Figure 1C ). Irradiation reactivity experiments were also conducted after the xenograft tumor grew to approximately 10mm in diameter. A total of 259 male SPF grade BALB/c-nude mice, aged 4-6 weeks and weighing 18-20g, were used in this study, all of which were provided by the Institute of Zoology, Chinese Academy of Medical Sciences with all characteristics having been confirmed. Conformity certificate number: SCXK (Beijing) 2015-0013. Animal experiments involved in this study were approved by the Animal Ethics Committee of the Fourth Hospital of Hebei Medical University. Each patient signed an informed consent approved by the Ethics Committee of the Fourth Hospital of Hebei Medical University. In order to clearly illustrate the experimental process, a flow chart of the study design was shown in Figure 2 . The raw data of key steps in this study can be found in the online link https://pan.baidu.com/s/1-SI1TFD15xCEvdrLLqPfTQ with the extraction code snls.

In order to obtain the irradiation response status of PDX models, a single irradiation of 16Gy was administered after the diameter of the implanted tumor reached a diameter 10mm. The scheme is as follows: the limbs of awake nude mice were affixed to foam boards using adhesive tape with tumor surface was covered by a tissue compensation membrane ( Figure 1D ) for radiotherapy with 6MV X-rays. Tumor thickness was defined by the vertical distance ranging from top to bottom. The distance from the irradiation source to the tumor center is 100cm, the dose rate is 500 MU/min, and the size of the irradiation field is 2×2cm. The nude mice were sacrificed 10 days after irradiation, and tissues were excised for HE staining. According to the rectal cancer regression grade (RCRG) standard (23), two experienced pathologists evaluated the irradiation reactivity of the PDX model. Under this scoring criterion, tumor regression was classified into three levels: RCRG 1: Sterilization or only microscopic foci of adenocarcinoma remaining, with marked fibrosis; RCRG 2: Marked fibrosis with macroscopic disease present; RCRG 3: Little or no fibrosis, with abundant macroscopic disease. In this study, we defined RCRG 1 as irradiation sensitive tissue and RCRG 3 as irradiation resistant tissue.

Human rectal cancer cell lines, including HRC-99, HR-8348, SW1463, SW837, and RCM-1, were purchased from Shanghai Fudan University Cell Bank. All cell lines were maintained in our laboratory and cultured in a RMPI 1640 medium (Invitrogen, USA) supplemented with 10% of fetal bovine serum (FBS, Invitrogen, USA), using a humidified 5% CO₂ incubator at 37°C. Cells were collected during the logarithmic growth phase for subsequent experiments. According to the pre-set irradiation dose gradient, cells were seeded in six-well plates, and the number of cells in a single well of each six-well plate were equal. The irradiation dose corresponding to the number of inoculated cells in each well were as follows: 0Gy 0.15×10³ cells, 2Gy 0.3×10³ cells, 4Gy 0.6×10³ cells, 6Gy 1.2×10³ cells, and 8Gy 2.4×10³ cells. Cells were incubated for an additional 14 days after irradiation. After incubation, cells were washed with PBS, fixed with paraformaldehyde, and stained with crystal violet solution (0.2%). The survival fraction (SF) was calculated using the numbers of colonies divided by the numbers of cells seeded multiplied by the plating efficiency (PE). Three independent experiments were performed.

According to the result of RCRG, three pairs of pre-stored tumor tissues corresponding to RCRG 1 and RCRG 3 were taken from -150°C refrigeration. Total RNA was isolated from frozen tumors, five rectal cancer cell lines, and transfected HR-8348 cells, purified using Trizol™ Plus RNA purification Kit (Invitrogen, CA, USA) according to the manufacturer’s recommended protocol, and quantified by UV absorbance at 260 and 280 nm. Denaturing agarose gel electrophoresis was used to evaluate the quality of the samples. Subsequently, miRNAs microarrays were performed by Shanghai Bohao Biotechnology Corporation Co. (http://www.shbiochip.bioon.com.cn , Shanghai, China). In this study, 2549 human miRNA precursor loci were annotated using Agilent Human miRNA chip V21.0 database according to the standard protocol (https://www.agilent.com/cs/library/usermanuals/public/G4170-90011.pdf). The detection rate and quality control status of each sample was detailed in Supplementary Table 1 .

After microarray analysis, quantitative reverse transcription polymerase chain reaction (qRT-PCR) was used for validation of the screening results in tissues and cells, as well as for detection of target genes that may be regulated by miRNA-96-5p. Complementary DNA (cDNA) was synthesized from the total RNA using the GoScript Reverse Transcription System (Promega, Wisconsin, USA) according to the manufacturer’s instruction. TB Green Fast PCR Mix (Takara-Bio-USA) was used as the amplification reagent. The melting curve analysis was performed to confirm the specificity of PCR products. U6 is a highly conserved small nuclear RNA (snRNA), which is relatively stable in different tissues and cells of the same organism, and is one of the most commonly used miRNA housekeeper genes (24, 25). Therefore, in this study, we used U6 as the internal reference of miRNAs. A study on GAPDH mRNA expression in a panel of 72 human tissues by Barber et al. (26) found that there were great differences in the expression of GAPDH between-tissue, but the expression variability of GAPDH gene was generally small within-tissue. Because the tissues and cells involved in this study were all from rectal cancer, we used GAPDH as the internal reference of the target genes expressions of miRNA-96-5p in this study. The total volume of the reaction system was 25μL, including 100 ng/μL cDNA 2μL, 10 μmol/L primer 2μL, Green system 12μL, ddH2O 9μL. Fold change (FC)= ΔCt(RCRG3)/ΔCt(RCRG1) and ΔCt= (Ct miRNA-Ct U6)RCRG3/(Ct miRNA-Ct U6)RCRG1 were used to calculate the expression level multiples of differential miRNAs in irradiation-resistant and irradiation-sensitive cancer tissues. The relative expression levels of miRNAs and mRNAs in cells were calculated by using 2 ^{- ΔΔCT} method. The primers of miRNAs were purchased from RiboBio Biotechnology Co., Ltd. (Guangzhou, China), and the sequences were patented by the company and specific base sequences could not be provided. The mRNAs primers and reaction conditions were listed in Supplementary Table 2 .

We analyzed the association of differentially expressed mRNAs with biological processes (BP) and molecular functions (MF) in the Gene Ontology (GO) database (http://geneontology.org). In addition, the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of differentially expressed Genes were carried out to comprehensively study the gene and expression information in order to identify the differentially enriched pathways. The enrichment analysis was performed using Fisher’s exact test in cluster Profiler from R/bioconductor (https://www.bioconductor.org). The standard of selection was the number of genes that fall on a GO term/or pathway ≥ 2, with a P-value < 0.05.

LV10N-hsa-miRNA-96-5p-inhibitor (5’-AGCAAAAATGTGCTAGTGCCAAA-3’), LV10N-hsa-miRNA-96-5p-NC (5’-TTCTCCGAACGTGTCACGT-3’), LV-SWP-GFP-GPC3 (Forward: 5’-CATCGGTACCATGGCCGGGACCGTGCG-3’, Reverse: 5’-TCGACTCGAGCACCAGGAAGAAGAAGCACACCACCG-3’) and LV-SWP-GFP were purchased from GenePharma Co., Ltd. (Shanghai, China). HR-8348 cells in the logarithmic growth phase were seeded into 24-well plates (0.5×10⁵ cells/well), 0.5mL complete culture medium (10% FBS+RPMI-1640) was added to each well, and incubated at 37°C with 5% CO2 for 24h. Then 10% FBS+DMEM culture medium containing Polybrene (5μg/mL) and the corresponding lentivirus (LV10N-hsa-miRNA-96-5p-inhibitor/-NC, pcDNA3.1-GPC3 vector and pcDNA3.1 empty vector) were used to replace the original medium. After 12-24h, the medium was removed and 0.5mL/well complete medium (10% FBS+RPMI-1640) was added. After 72h, the infection status of the cells was observed under a fluorescence microscope. 0.6µg/ml Purinomycin (Sigma-Aldrich, St. Louis, MO) was added into each well of successfully transfected cells lines. The expression levels of miRNA-96-5p and GPC3 were determined by qRT-PCR analysis.

In order to detect the irradiation reactivity of HR-8348 before and after transfection, we performed a clone formation experiment. The experimental procedures were the same as the screening of irradiation resistant cell lines.

The proliferation of HR-8348 cells, LV10N-hsa-miR-96-5p-inhibitor HR-8348 (HR-8348-IN) cells, and LV10N-hsa-miR-96-5p-NC HR-8348 (HR-8348-NC) cells were examined by MTS assay (Sigma-Aldrich, USA) at the indicated time points, accordingly to the descriptions of the MTS assays of Cory (27) and McCauley et al. (28). Cells were seeded at a density of 1.0×10³ cells/well in 96-well culture plates, and cultured for 0/24/48/72/96h. The absorbance values were determined on a microtiter plate reader (Expire Technology, Perkin Elmer) at 492 nm. Three independent experiments were done in triplicate.

HR-8348 cells in different conditions were seeded into 6-well plates at a density of 5.0×10^5/well and cultured conventionally. When the cells grew to full fusion, a sterile pipette tip was used to lightly scratch the cells at the centre of the 6-well plate. The wounded monolayers were washed with PBS to remove cell debris, and the cells were cultured in an incubator. Closure of the wound was observed under an inverted microscope at 0, 12 and 24h after scratching, and the distance between the two edges was measured. Ten fields of view were randomly selected, and images were acquired at the indicated timepoints. Image−Pro Plus version 5.0 software (Media Cybernetics, Inc., USA) was used to analyze all images.

Cells invasion assays were performed in a 24-well transwell chamber (Corning, USA), containing the 8-μm pore size polycarbonate membrane filter and was precoated with 20μl of Matrigel (Corning, USA) for invasion. Briefly, 1.0×10⁵ cells in different clones were seeded in the upper chambers and incubated in 200 μl RPMI medium (without FBS), while 600 μl medium with 10% FBS was placed in the lower chambers. The plates were incubated for 24h in an incubator. Subsequently, the invaded cells in the lower chamber were fixed with 4% paraformaldehyde for 10 min, stained with 0.1% crystal violet (Beyotime, China) for 5 min and lightly washed with PBS twice. Eventually, the number of invaded cells in five random fields of view were counted and photographed with a fluorescence microscope (Olympus, Japan) at 200×magnification.

Biological prediction softwares (TargetScan (http://www.targetscan.org), miRDB (http://www.mirdb.org), miRTarbase (http://mirtarbase.mbc.nctu.edu.tw), Tarbase (http://www.microrna.gr/tarbase)) were used to predict the target genes that might be regulated by miRNA-96-5p. Genes that can be predicted by the above four software programs were considered as the preliminary screening results, then the screening results were searched in Pubmed library (https://www.ncbi.nlm.nih.gov/pubmed) to further screen the possible target genes that may be related to biological tumor behavior. The target genes were verified by qRT-PCR and Western blot analysis, and their expression profiles in rectal cancer were obtained by Gene Expression Profiling Interactive Analysis (GEPIA) website (http://gepia.cancer-pku.cn/detail.php).

Total proteins from cells were extracted using RIPA lysis buffer containing the protease inhibitor PMSF. Cytoplasmic proteins were extracted using a specialized cytoplasmic protein extraction kit (Sangon Biotech Co., Ltd., Shanghai). Simply, the mixture of cell lysis products and pre-cooled Buffer (1μL DTT, 10μL PMSF, 1μL protease inhibitor, and 5μL phosphatase inhibitor) was centrifuged at 4100 rpm for 15 min at 4°C, and its supernatant was further centrifuged at 18000 RPM for 60min at 4°C again. The final supernatant was cytoplasmic proteins. Western blot was performed routinely, the primary antibodies were as follows: anti-CAV1 (Abcam, ab32577, 1:200), anti-DDIT3 (Abcam, ab11419, 1:1000), anti-PDCD4 (Abcam, ab51495, 1:1000), anti-MBD4 (Abcam, ab227625, 1:200), anti-DAB2 (Abcam, ab33441, 1:500), anti-FOXO3 (Abcam, ab109629, 1:1000), anti-GPC3 (Abcam, ab174851,1:2000), anti-β-catenin (Cell Signal TECHNOLOGY, 8480T, 1:1000), anti-GSK 3β (Cell Signal TECHNOLOGY, 12456T, 1:1000), anti-p-GSK 3β (Cell Signal TECHNOLOGY, 5558T, 1:1000), anti-CD44 (Cell Signal TECHNOLOGY, 37529T, 1:1000), anti-c-Myc (Cell Signal TECHNOLOGY, 5065T, 1:1000) and anti-GAPDH (Bioworld, AP0063, 1:5000). Goat anti-rabbit IgG or goat anti-mouse IgG (1:12,000 dilution; Sigma, USA) were used as the secondary antibody. Bands were visualized using an ECL Western blot detection kit (Amersham, USA). The ECL-based detection was performed with Chemiluminescence Reagent according to the manufacturer’s instructions. The level of GAPDH was used as a loading control.

The target gene analysis of miRNA-96-5p was performed using the biological prediction site microRNA.org (http://www.microrna.org/microrna/home.do), and dual-luciferase reporter gene assays were used to determine whether GPC3 was the direct target gene of miRNA-96-5p. The luciferase reporter vectors (pmirGlo-GPC3-3’UTR-WT and pmirGlo-GPC3-3’UTR-MUT) were synthesized by GenePharma Corp. The pRL-TK vector expressing Renilla was used as a reference to control for differences in cell number and transfection efficiency. MiRNA-96-5p mimics and miRNA-NC were co-transfected with luciferase reporter vectors into HR-8348 cells. Then, dual-luciferase reporter assays were performed according to the manufacturer’s instructions (GenePharma Corp. China).

All statistical analyses were carried out using SPSS for Windows version 17.0 (SPSS). Student’s t-test, oneway analysis of variance (ANOVA) and Spearman correlation analysis were used to analyze all of the data. All cell culture experiments were performed in triplicate. Data are presented as mean ± standard deviation (SD). Differences were considered statistically significant for P <0.05.

In our study, a total of 439 sites were implanted in primary transplantation, with a success rate of 58.78% (258/439), and 56 sites were implanted in secondary transplantation, with a success rate of 76.79% (43/56). Finally, PDX models of 29 patients were used in the irradiation experiment. The diameter of the implanted tumors was 10-15mm with a median of 12mm.

The volume of xenograft tumor was calculated using the formula for a spheroid “volume = length × height × width × π/6” prior to and 10 days after irradiation. As shown in Table 1 , there was no significant difference between the volumes of implanted tumors before and after radiation in vivo. After the measurement of gross volume, the nude mice were sacrificed and the tumor tissues were stripped. All specimens were found to have different degrees of liquefaction and degeneration. Among them, 3 specimens were mostly replaced by liquefied and necrotic tissues ( Figure 1E ), and the remaining specimens were tumors. After cleaning the liquefied necrotic tissue on tumor surface, the volume of the isolated tumor tissue was measured ( Figure 1F ) and compared with the volume of the tumor before and after irradiation, and there was a large difference between them (P<0.001) ( Table 1 ). The response of the implanted tumor to irradiation was evaluated according to the RCRG standard. It was found that 17.24% (5/29) of specimens were RCRG 1, 72.41% (21/29) specimens were RCRG 2, and 10.34% (3/29) specimens were RCRG 3 ( Figures 1G–I ).

The colony-forming ability and survival fraction of rectal cancer cell lines were shown in Table 2 and Figures 1J, K . HR-8348 cells presented with the highest resistance to irradiation than other cell lines (P<0.05). The trend of radio-resistant in these five cell lines was HR-8348> HRC-99> RCM-1> SW837> SW1463.

The miRNAs microarrays indicated that a total of 14 miRNAs were differentially expressed with a fold change value of ≥1.5, 4 up-regulated and 10 down-regulated ( Figure 3A ). Furthermore, using qRT-PCR, we confirmed that the relative expression levels of 4 up-regulated and 4 down-regulated miRNAs were consistent with the microarray data ( Figure 3B ).

miRNAs positively related to radio-resistance were tested in each cell lines using the qRT-PCR method. Interestingly, among these four miRNAs, the expression level of miRNA-96-5p was not only positively correlated with the radio-resistance, but also consistent with the trend of the irradiation resistance of these five cell lines ( Figure 4 ). Compared to the most radio-sensitive SW1463 cells, the radio-resistant HR-8348 cells expressed approximately 2.3 fold of miRNA-96-5p ( Table 3 ). The Spearman correlation analysis verified a positive correlation with irradiation resistance and expression of miRNA-96-5p in rectal cancer cells (r_s=0.938, P =0.000).

Next, we examined the potential role of miRNA-96-5p by suppressing miRNA-96-5p expression in HR-8348 cells. The expression of miRNA-96-5p was successfully down-regulated in its inhibitor transfected HR-8348 (HR-8348-IN) cells (5.75± 0.45 vs. 0.0036 ± 0.00095) ( Figure 5A ). Through the clone formation experiment under irradiation, it was found that the irradiation resistance of HR-8348 cells was decreased after inhibiting the expression of miRNA-96-5p ( Figure 5B ). In addition, we also conducted a series of cell function experiments to explore the role of miRNA-96-5p. MTS assays showed that the down-regulation of miRNA-96-5p significantly reduced the proliferation rate of HR-8348 cells ( Figure 5C ). Through wound healing assay and invasion assays, we found that the down-regulation of miRNA-96-5p significantly reduced the migration and invasion ability of HR-8348 cells ( Figures 5D–G ), suggesting that miRNA-96-5p has a positive effect on the aggressiveness of rectal cancer cells.

A total of 3419 genes that might be regulated by miRNA-96-5p were predicted using four different bioinformatics software programs. CAV1, DAB2, DDIT3, EDEM1, FoxO3, GPC3, MBD4, PDCD4, SLC25A25 and ZEB110 genes were all cross-predicted in these prediction tools. We further focused on seven genes including CAV1, DAB2, DDIT3, FOXO3, GPC3, MBD4, and PDCD4, being that they have been reported to play important roles in tumorigenesis. The expression profiles of the above seven genes obtained from the GEPIA website showed that GPC3 had the lowest expression in rectal cancer ( Figure 6 ). Moreover, the target gene validation results showed that only GPC3 was up-regulated in HR-8348-IN cells ( Figures 7A–G ). These results suggest that GPC3 may be the target gene regulated by miRNA-96-5p in HR-8348 cells. To determine whether GPC3 gene is directly regulated by miRNA-96-5p, we examined the effect of miRNA-96-5p on the activity of luciferase driven by GPC3 3’-UTR. The results showed that luciferase activity was significantly inhibited in the GPC3-WT group but has no effect in the GPC3-MUT group ( Figure 7H ). These findings implied that GPC3 might be a direct target gene of miRNA-96-5p.

HR-8348 cell lines with GPC3 stable over-expression (HR-8348-GPC3) were obtained by lentivirus transfection, and the transfection efficiency was 1.36 ± 0.38 vs. 207.01 ± 46.19 ( Figure 7I ). Furthermore, we found that HR-8348-GPC3 had significantly higher radiation sensitivity than HR-8348 cells ( Figure 7J ).

From the results of an enriched signal transduction pathways analysis in rectal cancer tissues with different irradiation resistance abilities, we found that Wnt was one of the most significant signal transduction pathways ( Figure 8A ). Therefore, the expression of key genes and downstream genes of canonical Wnt signal pathway were detected by Western blot analysis. In our results, the relative expression levels of β-catenin, a key gene in Wnt signaling pathway, in HR-8348 cells before and after transfection (HR-8348, HR-8348-NC and HR-8348-IN) were 0.804 ± 0.035, 0.767 ± 0.253 and 0.781 ± 0.0185, respectively, without statistical difference (P>0.05). Since it is known from previous reports (29) that the expression and accumulation level of β-catenin in the cytoplasm is an important factor affecting the activity of the Wnt/β-catenin signal transduction pathway, we further checked it in the cytoplasm of the three groups of cells, and found that the levels were 0.631 ± 0.05, 0.665 ± 0.038, and 0.321 ± 0.0108, respectively, and the ratios of β-catenin expression in the cytoplasm to total cell were 0.81 ± 0.0442, 0.823 ± 0.0321 and 0.392 ± 0.0088, respectively, (P<0.05) ( Figure 8B ). These results suggested that β-catenin expression in the cytoplasm of HR-8348 cells was inhibited with the down-regulated expression of miRNA-95-5p.

To investigate the mechanisms of the down-regulated β-catenin in the cytoplasm of HR-8348-IN cells, we detected the expression of GSK 3β and its phosphorylation, key regulatory factors of β-catenin. The relative expression levels of GSK 3β in the three cell lines were 0.979 ± 0.007, 1.06 ± 0.014, and 0.847 ± 0.044, respectively (P>0.05). However, the expression level of p-GSK 3β in HR-8348-IN cells was significantly lower than that in the other two groups (P<0.05), and the expression levels in each group were 0.872 ± 0.056, 0.698 ± 0.073, and 0.143 ± 0.003, respectively. The ratios of p-GSK 3β to GSK 3β in three groups were 0.82 ± 0.0314, 0.668 ± 0.0216 and 0.173 ± 0.0194, respectively (P<0.05) ( Figure 8C ). These results suggested that the phosphorylation level of GSK 3β was decreased in HR-8348 cells under the down-regulation of miRNA-96-5p expression.

Based on the above results, we speculated that the down-regulation of miRNA-96-5p in HR-8348 cells might inhibit the activity of the Wnt/β-catenin signaling pathway. To verify this hypothesis, we examined the expression of CD44 and c-Myc, the downstream genes of Wnt/β-catenin pathway. Our results showed that the expressions of these two genes were significantly decreased in HR-8348-IN cells, the expression levels of CD44 in the three cell lines were 1.016 ± 0.053, 1.045 ± 0.163, 0.475 ± 0.069 and c-Myc were 0.940 ± 0.069, 1.004 ± 0.180, 0.343 ± 0.052, respectively ( Figure 8D ).

To determine whether GPC3 was involved in the change of Wnt/β-catenin activity, the expression of β-catenin and GSK 3β in HR-8348 before and after GPC3 up-regulated were examined. We found that the ratios of β-catenin expression in cytoplasm to total cells were 0.818 ± 0.0275, 0.827 ± 0.0153, and 0.159 ± 0.0176 (P<0.05) ( Figure 8E ), and p-GSK 3β to GSK 3β was 0.802 ± 0.0301, 0.79 ± 0.017, and 0.093 ± 0.0121, respectively (P<0.05) ( Figure 8F ). These results indicated that the activity of the Wnt/β-catenin signal transduction pathway in HR-8348 cells was inhibited under the up-regulation of GPC3 expression.