B4GALT1 promotes immune escape by regulating the expression of PD-L1 at multiple levels in lung adenocarcinoma

The four MPLC patients and their clinical data were collected from Jiangsu Provincial People’s Hospital. Tumour tissues were obtained for research purposes from the surgical specimens on the day of surgery, and the tumour tissues were identified by the pathologist as MIA or IAC. These tumour samples were sent to Oncocares Inc. (Suzhou, China) for RNA sequencing. We applied the R “umap” package to perform the uniform manifold approximation and projection (UMAP) analysis. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) analyses were conducted with the “enrichR” (R package). The abundance levels of 22 immune cells were determined using CIBERSORT (https://​cibersort.​stanford.​edu).

The Cancer Genome Atlas (TCGA) RAN-sequencing data and patient information were downloaded from the UCSC Xena dataset (http://​xena.​ucsc.​edu/​welcome-to-ucsc-xena/​), and TCGA protein expression data were obtained from The Cancer Proteome Atlas dataset (http://​tcpaportal.​org/​tcpa/​index.​html). The GSE166720 and GSE31210 datasets were obtained from the Gene Expression Omnibus (GEO) database (http://​www.​ncbi.​nlm.​nih.​gov/​geo/​). The Cancer Cell Line Encyclopedia (CCLE) mRNA data were obtained from http://​www.​broadinstitute.​org/​ccle. Survival analysis was performed using the “survival” and “survminer” packages in R. The cytotoxic T-lymphocyte (CTL) level was calculated based on the average expression levels of CD8A, granzyme A (GZMA), CD8B, granzyme B (GZMB), and perforin 1. The Tumour Immune Dysfunction and Exclusion (TIDE) score was calculated using online software (http://​tide.​dfci.​harvard.​edu). The detailed method is described in the TIDE algorithm [21].

We purchased two LUAD cell lines (A549 and PC9) from the Institute of Biochemistry and Cell Biology of the Chinese Academy of Sciences (Shanghai, China). The medium consisted of DMEM or RPMI 1640 containing 10% foetal bovine serum (FBS) and 1% antibiotics (100 U/ml penicillin and 100 mg/ml streptomycin) and was used to maintain the PC9 or A549 cells, respectively. Lewis cells (LLCs) were purchased from Zhong Qiao Xin Zhou Biotechnology Co., Ltd (Shanghai, China) and maintained in DMEM.

Total RNA was extracted using TRIzol reagent (Invitrogen, MA, USA) and reverse-transcribed with a PrimeScript RT Reagent Kit. The resulting complementary DNA (cDNA) was used for qRT‒PCR with SYBR Green. The results were analysed with the delta-delta CT method, and values were normalized against GAPDH. The qRT‒PCR primers are shown in Supplementary Table S1.

Cell lysates were prepared in radioimmunoprecipitation assay buffer, and proteins were resolved by SDS‒PAGE using sodium dodecyl sulfate–polyacrylamide gels followed by protein transfer onto nitrocellulose membranes (Sigma‒Aldrich, St. Louis, MO, USA). The gels were visualized by autoradiography. A GAPDH antibody served as the internal control. Other antibodies used for the western blot (WB) assay were as follows: GAPDH (Abclonal, AC001); B4GALT1 (Abcam, ab121326; SantaCruz, sc-515551); Cyclin D1 (Abclonal, A19038); CDK4 (Abclonal, A11136); PD-L1 (Cell Signaling Technology, 15,165; Abcam, ab213524; Abcam, ab205921; SantaCruz, sc-293425); TAZ (Cell Signaling Technology, 8418); DDDDK-Tag (Abclonal, AE063); and HA-tag (Abclonal, AE008).

B4GALT1 cDNA was cloned into the expression vector pcDNA3.1. The X-tremeGEN™ HP DNA transfection reagent (Roche, Basel, Switzerland) was used for plasmid transfection, and the Lipo2000 reagent (Invitrogen, Shanghai, China) was used for small interfering RNA (siRNA) transfection, both performed according to the manufacturer’s instructions. In general, A549 or PC9 cells were seeded on coverslips in six-well plates and transfected with plasmid or siRNA on the second day. Cells were harvested 48 h after transfection for qRT‒PCR and western blot assays. The sequences of the siRNAs are summarized in Supplementary Table S1.

Cell viability was measured with an MTT kit (Roche Applied Science, Indianapolis, IN, USA). The cells were seeded in 96-well plates at 2*10³/well posttransfection. Then, every 24 h, the cells were incubated in an incubator with 20 μl MTT for 4 h, and the crystals were dissolved at room temperature for 10 min with 150 μl dimethyl sulfoxide. The optical density was measured at 490 nm using a spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

Five thousand cells were seeded in each well of a 6-well plate for the colony formation experiment, and growth medium was added as usual and replaced 1 week later. Methanol was applied for 15 min after the colonies had grown for 2 weeks, followed by 0.1% crystal violet (Sigma) staining for 30 min. Then, the formed clones were counted to reflect the colony-forming ability of the clones.

EdU labelling and staining were performed with an EdU cell proliferation detection kit (RiboBio, Guangzhou, China). After adding cells at 5 × 10³ cells/well to 96-well plates, 50 mΜ EdU labelling medium was added to 96-well plates 48 h following transfection and incubated for 2 h in an incubator at 37 °C with 5% CO₂. The cells were then treated with 4% paraformaldehyde and 0.5% Triton X-100 for anti-Edu working solution staining. The nuclei were labelled with diamidino-2-phenylindole. Fluorescence microscopy was performed to calculate the percentage of EdU-positive cells.

Invasion and migration assays were carried out using Corning’s Transwell system (24 wells, 8 mm pore size, New York, NY, USA). For migration assays, 5 × 10⁴ cells posttransfection were seeded into the upper chambers of the plates with 350 μl of serum-free medium, and 700 μl of medium containing 10% FBS was added to the lower chambers. In Matrigel invasion assays, Matrigel (Sigma‒Aldrich)-precoated Transwell membranes were used. After incubation for 16 h, the upper surface cells were removed, and the cells that penetrated the membrane to the lower surface were stained with methanol and 0.1% crystal violet. Photographs were taken with an inverted microscope (Olympus, Tokyo, Japan).

Cells were seeded onto a 6-well plate and cultured until they reached 90–100% confluence. Using a small pipette tip, confluent cells were scratched and washed twice with phosphate-buffered saline (PBS). Images of the same positions of each well at 0 or 16 h were taken with a microscope (Olympus, Tokyo, Japan). Wound closure was determined as a percentage of wound confluence using ImageJ software.

After 48 h of transfection, trypsinized cells were collected and stained with fluorescein isothiocyanate (FITC)-Annexin V and propidium iodide using a FITC Annexin V Apoptosis Detection Kit (BD Biosciences, Franklin Lakes, NJ, USA). Live, dead, early apoptotic cells, and apoptotic cells were detected using a flow cytometer (FACScan; BD Biosciences) equipped with Cell Quest software (BD Biosciences). The proportion of apoptotic cells was calculated as the sum of the ratio of early apoptotic cells to late apoptotic cells. Cell cycle analysis was performed using the CycleTEST PLUS DNA Reagent Kit (BD Biosciences), and the results were analysed with FACScan. The percentages of cells in the G0-G1, S, and G2-M phases were determined and compared.

Using Ficoll separation, we isolated peripheral blood mononuclear cells (PBMCs) from healthy donors’ peripheral blood. We also isolated splenocytes from the spleens of C57BL/6 mice. Positive selection of CD8 + T cells from PBMCs and splenocytes was performed using a CD8 MicroBeads Kit (Miltenyi Biotec 130–045-201, 130–116-478; Cologne, Germany). The isolated CD8 + T cells were cultured in RPMI 1640 medium supplemented with 10% foetal bovine serum, 1% antibiotics, 1X MEM nonessential amino acid (Gibco 11,140,050; Carlsbad, CA, USA), 1 mM sodium pyruvate (Gibco 11,360,070), 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (Gibco 15,630,080), 20 ng/ml IL2 (Peprotech 200–02; Cranbury, NJ, USA), and CD3/CD28 T-cell activator (Gibco 11161D) for 2 weeks. After transfection with siRNA or plasmid for 24 h, tumour cells and CD8 + T cells were cocultured at a 3:1 ratio (T cells: tumour cells) for 48 h. After removing T cells and cancer cell debris, crystal violet-stained living cancer cells were measured at 570 nm.

Five-week-old male BALB/c nude mice and C57BL/6 mice were purchased from Jiangsu Gempharmatech, Jiangsu, China. For subcutaneous xenograft experiments in BALB/c nude mice, 1 × 10⁷ A549 cells (si-B4galt1 or si-control) were resuspended in 100 μl PBS and inoculated into the left and right flanks of mice. For subcutaneous xenograft experiments in C57BL/6 mice, 2 × 10⁶ LLC cells (si-B4galt1 or si-control) suspended in 100 μl PBS were injected into the flanks. Once tumours reached approximately 100 mm³, mice were treated with mouse programmed cell death protein-1 (PD-1) mAb (BioXcell, BE0146; New Haven, CT, USA) or IgG isotype control (BioXcell, BE0089) by intraperitoneal injection (100 mg per mouse in 100 ml D-PBS buffer). Tumour sizes were measured every 2–3 days. Tumour weight was recorded, and volumes were estimated according to 1/2 × (length × width²).

After C57BL/6 mice were sacrificed, fresh tumour tissue was excised and digested with collagenase IV (Gibco 17,104,019) and DNase I (Roche 10,104,159,001) for 2 h and filtered through 70-μm cell strainers to obtain a single-cell suspension. Cells were washed and then resuspended in 40% Percoll and centrifuged against 70% Percoll to sort lymphocytes. The cells were stimulated for 4 h with a leukocyte activation cocktail (BD Pharmingen, 550,583; San Diego, CA, USA) for intracellular cytokine staining. After blocking nonspecific antibody binding with CD16/CD32 antibody (553,141), surface staining was performed with the following antibodies: anti-CD45 (553,080), anti-CD3e (553,064), and anti-CD8α (551,162) mAbs from BD Pharmingen. After fixation and permeabilization with a Fixation/Permeabilization kit (BD Biosciences, 554,714), intracellular GZMB was stained with an anti-GZMB antibody (Biolegend, 372,204; San Diego, CA, USA). Stained cells were analysed by BD FACS Canto II. The data were analysed with FlowJo software.

ells were then treated with 50 μg/ml cycloheximide (Selleck, Houston, TX, USA), and samples were collected at 0, 4, 8, and 12 h. Then, the samples were lysed to harvest proteins for subsequent western blot analysis.

Tissue sections were obtained from the Pathology Department of Jiangsu Provincial People’s Hospital and were made from formalin-fixed paraffin-embedded samples from four MPLC patients. Sections were first stained with primary antibodies, followed by biotin-conjugated secondary antibodies and avidin–biotin-peroxidase complexes. Diaminobenzidine was used to visualize the target proteins. The images were analysed using 3D HISTECH QuantCenter 2.1 software. The H-score was used to quantify the immunohistochemical results. H-scores were used to quantify the immunohistochemical results, which were calculated according to the following formula: H-Score = ∑(pi × i) = (percentage of weak intensity × 1) + (percentage of moderate intensity × 2) + (percentage of strong intensity × 3) [22].

For multiple immunofluorescence, paraffin sections were dewaxed in water and then placed in a box filled with ethylenediaminetetraacetic acid (EDTA) antigen repair buffer (pH 8.0) in a microwave oven for antigen repair. After blocking with 3% bovine serum albumin (BSA) at room temperature for 30 min, the first primary antibody was added and incubated at 4 °C overnight, after which the slides were placed in PBS (pH 7.4) and washed by shaking on a decolorization shaker, followed by incubation with the corresponding secondary antibody at room temperature for 50 min. DAPI staining solution was applied to counterstain the nuclei with incubation for 10 min in the dark at room temperature. Finally, the slides were sealed with an anti-fluorescence quench sealer and placed under a fluorescence microscope for imaging. The antibodies employed for immunohistochemical and immunofluorescence were as follows: B4GALT1 (Abcam, ab32137); PD-L1 (Servicebio, GB11339A; Proteintech, 2B11D11); GZMB (Servicebio, GB14092); CD3 (Servicebio, GB11014); and CD8 (Servicebio, GB12068).

Following the manufacturer’s instructions, we performed coimmunoprecipitation experiments using the Pierce Classic Magnetic IP/Co-IP Kit (Thermo Fisher Scientific, San Jose, CA). Briefly, A549 cells were transfected with B4GALT1-HA (or empty vector) and PD-L1-Flag/TAZ-Flag (or empty vector) for 24 h, and then cell lysates were incubated with DDDDK-Tag mAb (Abclonal, AE063) or anti-HA-Tag mAb (Abclonal, AE008) at room temperature for 2 h. Then, Protein A/G magnetic beads were bound to the antigen/antibody complexes at 4 °C overnight, followed by two washes with lysis/wash buffer and one wash with purified water. Finally, the beads were resuspended in 100 μl of Lane Marker Sample Buffer and boiled for 10 min for subsequent western blotting.

GST and GST–B4GALT1 proteins were purchased from OriGene (WX0383AA, WX02BF90), and we performed GST pull-down assays using the Pierce™ GST Protein Interaction Pull-Down Kit (Thermo Fisher Scientific, San Jose, CA). Briefly, we immobilized 150 µg GST and GST-B4GALT1 fusion proteins in 50 µL glutathione agarose and incubated them together for 60 min at 4 °C with gentle rocking. Then, A549 cell lysates were added to the immobilized GST-B4GALT1 and GST. The two fusion proteins were incubated overnight at 4 °C with gentle rocking. The bound proteins were eluted with 10 mM glutathione elution buffer and examined by SDS‒PAGE and western blotting.

Following seeding in a 24-well plate, cells were cotransfected with the CD274 promoter-reporter gene vector, the Renilla luciferase vector, and either B4GALT1 pcDNA or TAZ siRNA. After 48 h of transfection, cell lysates were prepared using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA). Renilla luciferase activity was used as the internal control for normalizing the luciferase activity.

Eight LUAD samples obtained from four MPLC patients were collected, and their radiomic features and clinicopathologic characteristics are summarized in Fig. 1A. In each patient, the two primary tumours were of different pathological types, MIA or IAC. Then, we performed RNA sequencing to explore key molecular events driving the progression of preinvasive LUAD to invasive LUAD in these samples (Fig. 1B). Subsequently, unsupervised clustering of RNA-sequencing data with UMAP identified two clusters that corresponded exactly to the MIA and IAC lesions of these four patients (Fig. 1C), suggesting distinct transcriptional characteristics for MIA and IAC. GO and KEGG analyses revealed that genes highly expressed in IAC were significantly enriched in glucose-metabolic pathways (Fig. 1D). Furthermore, N-glycan biosynthesis, with enrichment of beta-1,4-galactosyltransferase activity signalling pathways, was also found, indicating that glycosylation played a key role in the development of early-stage lung cancer. As one type of glycosylation, the N-glycosylation pattern can regulate protein stability, and beta-1,4-galactosyltransferase is a key enzyme involved in N-glycan biosynthesis. Here, we focused on B4GALT1, a gene encoding β-1,4-galactosyltransferase, which was the most significantly upregulated gene in IAC samples. Therefore, we speculated that B4GALT1 may play a key role in the evolution of MIA to IAC for early-stage LUAD. Further analysis of TCGA-LUAD data showed that B4GALT1 mRNA expressions were indeed increased with stage in early-stage LUAD but not in advanced LUAD (Fig. 1E). Moreover, the GSE166720 dataset, which classified 53 early-stage lung adenocarcinomas into indolent (AIS and MIA) and invasive tumours (IAC) according to pathological grade, showed that B4GALT1 mRNA expressions were higher in invasive tumours than in indolent tumours (Fig. 1F). Additionally, high B4GALT1 expression predicted worse DFS and OS in TCGA-LUAD (Fig. 1G-H). High B4GALT1 also predicted poor OS in the GSE31210 dataset with stage I and stage II LUAD patients (Fig. 1I). Together, the results suggested that B4GALT1 might play a crucial role in the tumorigenesis of early-stage LUAD.

To explore the functional role of B4GALT1 in the tumorigenesis of LUAD, we used a specific siRNA and an overexpression plasmid to manipulate the expression of B4GALT1 in A549 and PC9 cells (Figure S1A-C). MTT tests, colony formation assays, and EdU experiments were performed to identify the role of B4GALT1 in cell proliferation. MTT tests showed that B4GALT1 knockdown could significantly decrease cell viability. In contrast, B4GALT1 overexpression increased cell viability (Fig. 2A). Similarly, EdU assays revealed that B4GALT1 had a significant impact on LUAD cell proliferation (Fig. 2B). Colony formation assays demonstrated that B4GALT1 expression also had an impact on clonogenic ability (Fig. 2C). Moreover, B4GALT1 knockdown significantly induced LUAD cell apoptosis (Fig. 2D). Then, we determined whether B4GALT1 affected LUAD cell proliferation by altering the cell cycle. Flow cytometry results showed that B4GALT1 knockdown was accompanied by cell cycle arrest at the G1/G0 stage (Fig. 2E). We then examined the expression of G1/S phase transition-related proteins. As determined by western blotting, B4GALT1 knockdown or overexpression led to decreases or increases in cyclin D1 and CDK4 protein levels in LUAD cell lines (Figure S1D).

Next, we tested the potential impact of B4GALT1 on cell migration and invasion. In wound-healing assays, B4GALT1 expression significantly impaired wound closure (Fig. 3A). Transwell assays also demonstrated that B4GALT1 knockdown impeded tumour migration. Conversely, B4GALT1 overexpression significantly promoted cell migration (Fig. 3B). Furthermore, B4GALT1 knockdown restrained LUAD cells from invading through matrices, whereas B4GALT1 overexpression facilitated the invasion of LUAD cells (Fig. 3C).

We subcutaneously injected A549 cells into BALB/c nude mice to determine whether B4GALT1 influences LUAD tumour growth in vivo. The tumour volume in the B4GALT1 knockdown group decreased significantly compared with that in the control group. At the experimental endpoint, the average weight of tumours in the B4GALT1 knockdown group was significantly lower than that in the control group (Fig. 3D). Moreover, Ki67 staining showed a decrease in the proportion of Ki67-positive cells among tumours formed by B4GALT1 knockdown (Fig. 3E).

Through GO analysis of highly expressed genes in IAC, we also detected an enrichment of genes with functions associated with the T-cell immune response (Fig. 1D). Previous studies have reported increased immunosuppression during the progression of LUAD from MIA to IAC [10]. Therefore, we applied the CIBERSORT algorithm to estimate the relative abundance of immune cells among MIA and IAC tumours. We found that IAC had less CD8 + T-cell infiltration than MIA (Fig. 4A). Subsequent immunofluorescence staining analysis also showed fewer CD8 + T cells in IAC tumours than in MIA tumours (Fig. 4B). Immunohistochemical staining for GZMB, the key CD8 + T-cell effector molecule, revealed that IAC tumors had lower GZMB level, indicating the attenuated activity of CD8 + T cells in IAC tumours (Fig. 4C). To determine whether B4GALT1 induced this alteration in the immune status, immunohistochemical staining for BGALT1 was carried out on these samples (Fig. 4D), and a clear inverse correlation between B4GALT1 level and the number of CD8 + T cells was found, as well as for GZMB (Fig. 4E). We further validated these correlations in the TCGA-LUAD dataset. B4GALT1 mRNA levels negatively correlated with CD8 + T-cell infiltration (Figure S2A), and tumours with higher B4GALT1 mRNA levels had higher tumour immune dysfunction and exclusion (TIDE) scores than those with lower B4GALT1 expression, suggesting the immunosuppressive role of B4GALT1 in LUAD tumorigenesis (Figure S2B). In addition, the correlation between the cytotoxic T-lymphocyte (CTL) level and OS for LUAD patients was assessed under the condition of high or low B4GALT1 mRNA levels in two independent GEO datasets. As shown in Figure S2C, high expression of B4GALT1 was inversely related to a CTL-mediated survival benefit, implying that in tumour microenvironments characterized by high B4GALT1 expression, CTLs may become dysfunctional.

To determine whether B4GALT1 was an important regulator of dysfunctional T cells, we explored data from publicly available clustered regularly interspersed short palindromic repeats (CRISPR) screens in the context of murine cancer cells cocultured with murine primary T cells that specifically target cancer cell antigens. Based on nine such screens in five independent studies [23‐26], we found that gRNAs targeting B4GALT1 were consistently negatively selected (Fig. 4F). These results indicated that B4GALT1 could contribute to CD8 + T-cell dysfunction and may make LUAD cells more susceptible to CD8 + T-cell-mediated cytotoxicity when inactivated. To further verify this conclusion, T-cell-mediated tumour cell-killing assays were performed to test the impact of B4GALT1 expression on CD8 + T-cell activity, which were isolated from the peripheral blood of healthy donors and the spleens of C57BL/6 mice (Fig. 4G). As expected, B4GALT1 inhibition significantly enhanced LUAD cell death mediated by CD8 + T cells, and B4GALT1 upregulation rendered LUAD cells more resistant to CD8 + T cells in both the A549 and LLC cell lines (Fig. 4H).

Increasing evidence indicates that immune suppression is linked to the dysregulation of immune checkpoints [27, 28]. Therefore, we compared the expression levels of well-known immune checkpoint molecules between tumours with high and low B4GALT1 expression (Fig. 5A). Notably, CD274 (PD-L1) exhibited the most significantly higher expression in B4GALT1-high tumours than in B4GALT1-low tumours. CCLE and TCGA data further substantiated that high B4GALT1 expressers had higher expression levels of PD-L1 mRNA and protein (Figure S3A-B). Furthermore, there was a highly significant positive correlation between B4GALT1 and PD-L1 level among our MPLC samples (Figure S3C). In agreement with the results above, decreased or increased mRNA and protein levels of PD-L1 were elicited by B4GALT1 knockdown or overexpression in A549 and LLC cells (Fig. 5B-C, Figure S3D-F). Therefore, it is reasonable to hypothesize that B4GALT1 mediates CD8 + T-cell dysfunction by affecting PD-L1 expression in LUAD. This speculation was confirmed by the restoration of CD8 + T-cell killing ability by CD274 knockdown in B4GALT1-overexpressing LUAD cells (Fig. 5D, Figure S3G).

Next, we asked how B4GALT1 affects the level of PD-L1. B4GALT1 is a key gene involved in N-glycan biosynthesis and participates in the formation of N-glycosylation. Therefore, we speculated that N-linked glycosylation may be involved in the B4GALT1-mediated regulation of PD-L1 via a mechanism involving posttranslational modification. First, after treatment with cycloheximide (CHX), we found that B4GALT1 depletion reduced the half-life of the PD-L1 protein in both A549 and LLC cells, whereas stable B4GALT1 expression significantly increased the protein stability of PD-L1 (Fig. 5E, Figure S3H). Besides, co-IP verified the interaction between B4GALT1 and PD-L1 (Fig. 5F). To confirm the interaction biochemically, the GST fusion protein containing B4GALT1 was expressed and purified to perform GST pulldown assays. The result showed that GST-B4GALT1 could pull down PD-L1 in A549 cell lysates (Fig. 5G). Then, we sought to investigate whether B4GALT1 regulates the PD-L1 protein through N-linked glycosylation. A549 and LLC cells were treated with tunicamycin (an N-linked glycosylation inhibitor), and PD-L1 expression was significantly downregulated compared with that in nontreated cells, consistent with B4GALT1 knockdown. Moreover, tunicamycin rescued the PD-L1 expression induced by B4GALT1 overexpression (Fig. 5H, Figure S3I). To further identify the N-linked glycosylation site of PD-L1, we first employed the NetNGlyc 1.0 server tool to predict the putative N-linked glycosylation motifs of PD-L1. Then, B4GALT1-overexpressing cells were transfected with FLAG-tagged plasmids carrying either a full-length coding sequence of PD-L1 or a coding sequence containing a mutated N-linked glycosylation site at N35, N192, or N200 for CHX analysis (Fig. 5I). The results showed that mutating the N-linked glycosylation sites at N192 and N200 significantly reduced PD-L1 protein stability even in the presence of B4GALT1 overexpression, suggesting that these residues are the N-linked glycosylation sites of PD-L1 modified by B4GALT1. Taken together, the above results suggested that B4GALT1 interacts with PD-L1 by mediating its glycosylation, thus promoting immune escape in LUAD.

Our data above demonstrated that B4GALT1 could enhance PD-L1 protein stability. In addition, we noticed changes in CD274 mRNA after B4GALT1 knockdown or overexpression (Fig. 5B, Figure S3E). These results indicated that B4GALT1 may also have a transcriptional effect on CD274. To test this hypothesis, luciferase reporter assays were performed and revealed that CD274 promoter activity was enhanced by B4GALT1 overexpression (Fig. 6A, Figure S4A). In addition, interestingly, we found that highly expressed genes in IAC tumours were enriched in the Hippo signalling pathway (Fig. 1D). In prior studies, WW domain-containing transcription regulator 1, (TAZ), an important transcription factor of the Hippo signalling pathway, was found to increase CD274 promoter activity to induce PD-L1 upregulation [29, 30]. Here, we also validated the regulation of CD274 by TAZ in LUAD. TAZ knockdown led to substantial decreases in CD274 mRNA and protein expression levels in A549 and LLC cells (Fig. 6B-C, Figure S4B-C). Further reporter gene assays revealed that TAZ knockdown decreased the activity of the full-length CD274 promoter but not the promoter construct with a deleted region (Fig. 6D, Figure S4D). Importantly, B4GALT1 overexpression did not affect the activity of the CD274 promoter construct with the deleted region, indicating that B4GALT1 affects CD274 promoter activity through TAZ (Fig. 6E). Furthermore, TAZ knockdown significantly mitigated the B4GALT1-induced increases in full-length CD274 promoter activity and CD274 mRNA levels (Fig. 6F-G, Figure S4E-F), suggesting that B4GALT1 mediates the transcriptional regulation of CD274 through TAZ. Moreover, TAZ knockdown reversed B4GALT1-induced PD-L1 protein level (Fig. 6H, Figure S4G).

We further sought to explore the mechanism by which B4GALT1 impacts TAZ level. Through the analysis of TCGA and GEO datasets, we observed that B4GALT1 expression was significantly related to TAZ level at the protein level instead of the mRNA level (Figure S5A-B), consistent with the qRT‒PCR and WB results (Fig. 6I-J, Figure S5C-D). Thus, we suspected that TAZ protein may be regulated by B4GALT1 at the posttranslational level by acquiring N-glycan modifications. Through CHX analysis, we found that B4GALT1 could increase TAZ protein stability (Fig. 6K, Figure S5E). Moreover, co-IP (Fig. 6L) and GST pull-down assays (Fig. 6M) suggested that B4GALT1 directly interacted with TAZ. Furthermore, TAZ protein levels were significantly decreased after tunicamycin treatment in B4GALT1-overexpressing cells (Fig. 6N, Figure S5F), indicating that the N-linked glycosylation of TAZ is regulated by B4GALT1. Then, we applied the NetNGlyc 1.0 server tool to predict the potential N-linked glycosylation sites of TAZ and mutated the N184 and N256 sites (Fig. 6O). The results of CHX analysis indicated that B4GALT1 overexpression could not promote TAZ protein stability when the N-linked glycosylation site at N256 was mutated, implying that N256 is the N-linked glycosylation site of TAZ modified by B4GALT1.

These results demonstrated the translational modification of PD-L1 which was mediated by B4GALT1/TAZ.

To evaluate the immune regulatory effect of B4GALT1 in vivo, we utilized the LLC mouse LUAD cell line to construct subcutaneous xenografts in immunocompetent C57BL/6 mice, which were then treated with the PD-1 mAb or IgG isotype CTRL (IgG2a) (Fig. 7A). We found that B4galt1 knockdown limited tumour growth compared to the control group, consistent with PD-1 mAb treatment. Moreover, the B4galt1 inhibition and PD-1 blockade combination therapy group showed the most significant reduction in tumour burden compared with the other groups, and neither body weight loss nor other common toxic effects were observed (Fig. 7B-F). Tumour samples were harvested for further analysis at the end of the treatment, and IHC staining and flow cytometry analysis showed that B4galt1 deficiency significantly decreased PD-L1 level and increased CD8 + T-cell and CD8 + GZMB + cell densities (Fig. 7G-H). In particular, cotreatment with PD-1 mAb and B4galt1 inhibition led to the most significant enrichment of activated CD8 + T cells in tumour regions (Fig. 7H). These results confirmed that B4galt1 inhibition enhanced CD8 + T-cell infiltration and function by downregulating PD-L1 expression, thus improving the efficacy of anti-PD-1 therapy. Collectively, these results indicated that B4GALT1 contributed to tumour immune escape by upregulating PD-L1 expression and inhibiting CD8 + T-cell infiltration.