Re:한약발효균 'Lactobacillus acidophilus' 종양용해바이러스, 항암효과 2025 네이처!!

[문형철]

향후 5년 모니터링 과제!!

1. nature  
2. nature communications  
3. articles  
4. article

Lactobacillus acidophilus potentiates oncolytic virotherapy through modulating gut microbiota homeostasis in hepatocellular carcinoma

Download PDF

• Article
• Open access
• Published: 07 April 2025

Lactobacillus acidophilus potentiates oncolytic virotherapy through modulating gut microbiota homeostasis in hepatocellular carcinoma

• Jiayu Zhang, 
• Jinneng Yang, 
• Jinyan Luo, 
• Weili Wu, 
• Haidan Luo, 
• Wenxia Wei, 
• Haimei Lyu, 
• Yuzhi Wang, 
• Hairong Yi, 
• Yijing Zhang, 
• Zongmin Fan, 
• Haiwen Lyu, 
• Vishnu Priya Kanakaveti, 
• Baifu Qin, 
• Ping Yuan, 
• Runxiang Yang, 
• Haipeng Zhang, 
• Tao Zuo, 
• Dean W. Felsher, 
• Mong-Hong Lee & 
• Kai Li 

Nature Communications volume 16, Article number: 3315 (2025) Cite this article

•  
•  
•  

Abstract

Oncolytic viruses (OVs) hold promise for cancer treatment. However, the antitumor efficacy is limited. Microbiota plays a pivotal role in cancer treatment and its impact on oncolytic virotherapy is unknown. Here, we show that VSVΔ51 has higher antitumor efficacy for hepatocellular carcinoma in the absence of microbiota in female mouse models. VSVΔ51 infection causes microbiota dysbiosis, increasing most of the gut bacteria abundance, while decreasing the commensal Lactobacillus. VSVΔ51 reduced intestinal expression‎ of SLC20A1 that binds to Lactobacillus acidophilus (L. acidophilus) CdpA cell wall protein through IL6-JAK-STAT3 signaling, thereby attenuating attachment and colonization of L. acidophilus. L. acidophilus supplementation confers sensitivity to VSVΔ51 through restoring gut barrier integrity and microbiota homeostasis destroyed by VSVΔ51. In this work, we show that targeting microbiota homostasis holds substantial potential in improving therapeutic outcomes of oncolytic virotherapy.

초록

항암 바이러스(OVs)는

암 치료에 유망한 치료법이다.

항암바이러스 : 암세포를 선택적으로 감염시켜 제거하는 바이러스

그러나 항종양 효능은 제한적이다.

미생물 군집은

암 치료에서 중추적인 역할을 하지만,

종양 용해 바이러스 치료에 미치는 영향은 알려지지 않았다.

Oncolytic viruses (OVs) = VSVΔ51

oncolytic virotherapy

종양용해 바이러스 치료에 대한 반응의 이질성은 임상적 효과에 장벽을 형성하는데, 이는 종양 내 바이러스 확산 억제를 통해 치료 저항성이 발생할 수 있으며 이는 치료 실패로 이어질 수 있기 때문이다. 여기서 우리는 크렙스 사이클 유래 대사산물인 이타코네이트의 화학 유도체인 4-옥틸 이타코네이트(4-OI)가 인간 및 생쥐 내성 암세포주, 3차원(3D) 환자 유래 대장 종양 및 장기형 뇌종양 슬라이스를 포함한 다양한 모델에서 VSVΔ51을 이용한 종양 용해 바이러스 치료를 강화함을 보여줍니다. 또한, 4-OI와 VSVΔ51의 병용 투여는 내성 쥐 대장 종양 모델에서 치료 결과를 개선합니다. 기전적으로, 우리는 4-OI가 NRF2/KEAP1 축과 독립적으로 MAVS 및 IKKβ의 시스테인 잔기 변형을 통해 암세포 내 항바이러스 면역을 억제함을 발견했습니다. 우리는 대사산물 유래 약물과 종양용해성 바이러스 제제의 병용이 제1형 인터페론(IFN) 및 NF-κB 매개 항바이러스 반응에 직접 간섭함으로써 항암 치료 결과를 크게 개선할 수 있다고 제안합니다.

본 연구에서는

VSVΔ51이 여성 마우스 모델에서 미생물 군집이 없는 상태에서

간세포암종에 대해 더 높은 항종양 효능을 보인다는 것을 보여준다.

VSVΔ51 감염은

미생물 군집 불균형을 유발하여

대부분의 장내 세균의 풍부도를 증가시키는 반면,

공생균인 Lactobacillus는 감소시킨다.

VSVΔ51은

IL6-JAK-STAT3 신호전달을 통해

Lactobacillus acidophilus (L. acidophilus)의 세포벽 단백질 CdpA에 결합하는

SLC20A1의 장내 발현을 감소시켜,

L. acidophilus의 부착 및 정착을 약화시킵니다.

L. acidophilus 보충은

VSVΔ51에 의해 파괴된 장 장벽 무결성과 미생물군집 항상성을 회복시켜

VSVΔ51에 대한 감수성을 부여합니다.

본 연구에서는 미생물군집 항상성을 표적으로 삼는 것이

종양용해 바이러스 치료법의 치료 결과를 개선하는 데

상당한 잠재력을 지닌다는 점을 보여줍니다.

Similar content being viewed by others

Bifidobacteria metabolize lactulose to optimize gut metabolites and prevent systemic infection in patients with liver disease

Article 16 October 2023

Synergistic antimicrobial and probiotic activity of lactic acid bacteria isolated from Tarkhineh against Candida albicans

Article Open access01 July 2025

Oral supplementation with selected Lactobacillus acidophilus triggers IL-17-dependent innate defense response, activation of innate lymphoid cells type 3 and improves colitis

Article Open access20 October 2022

Introduction

Hepatocellular carcinoma (HCC), accounting for ~75% of all liver cancers1, is the third leading cause of cancer related deaths worldwide2. The development of systemic therapy for HCC has achieved great progress during the last decades, including multi-targeted tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs)3. Despite these unprecedented and encouraging results, only 20–30% of patients respond to treatment3. Thus, there is an urgent need to identify new therapeutics that inhibit tumorigenesis and enhances the responses in HCC.

Oncolytic virus (OVs) are viruses that preferentially replicate in and kill tumor cells while leaving non-neoplastic cells intact4. Oncolytic virotherapy is a potentially attractive strategy for the treatment of HCC5,6. Many OVs are used for clinical trials in liver cancer patients7,8. For example, a phase II trial showed that JX-594 high-dose group increased patients’ median total survival time by 7.4 months compared with the low-dose group, suggesting that JX-594 showed potential antitumor effect and good tolerance in HCC patients9. Vesicular stomatitis virus (VSV), a negative-strand RNA virus, also shows prominent antitumor efficacy in hepatocellular carcinoma10,11,12,13. However, clinical trials have shown that only a subset of patients respond well to OVs9. Thus, enhancing oncolytic efficacy is urgently needed. Most of the studies have mainly focused on the mechanism of OVs in the tumor microenvironment (TME)14,15,16,17,18 but the interaction between OVs treatment and other organs or tissues is rarely explored.

서론

간세포암(HCC)은

전체 간암의 약 75%를 차지하며1,

전 세계적으로 암 관련 사망 원인 중 세 번째로 높은 질환이다2.

지난 수십 년간

다중 표적 티로신 키나제 억제제(TKIs) 및 면역 체크포인트 억제제(ICIs)를 포함한

간세포암 전신 치료법 개발은 큰 진전을 이루었습니다3.

이러한 전례 없는 고무적인 결과에도 불구하고,

치료에 반응하는 환자는 20~30%에 불과합니다3.

따라서

간세포암에서 종양 형성을 억제하고 반응을 향상시키는

새로운 치료법을 규명할 시급한 필요성이 있습니다.

종양용해바이러스(OVs)는

종양세포 내에서 선택적으로 증식하여 이를 사멸시키면서도

비종양세포는 손상시키지 않는 바이러스이다4.

Oncolytic virus (OVs)

종양용해바이러스 치료법은

간세포암 치료를 위한 잠재적으로 매력적인 전략이다5,6.

간암 환자를 대상으로

한 임상 시험에서 다양한 OVs가 사용되고 있다7,8.

예를 들어,

2상 임상시험에서 JX-594 고용량군은

저용량군 대비 환자의 전체 생존 기간 중앙값을 7.4개월 연장시켜,

JX-594가 간세포암 환자에서 잠재적 항종양 효과와 우수한 내약성을 보임을 시사하였다9.

음성 가닥 RNA 바이러스인 수포성 구내염 바이러스(VSV)

역시 간세포암종에서 두드러진 항종양 효능을 보인다10,11,12,13.

그러나

임상 시험 결과,

일부 환자군만이 온콜리틱 바이러스에 잘 반응하는 것으로 나타났다9.

따라서

온콜리틱 효능을 강화하는 것이 시급히 필요하다.

대부분의 연구는 주로 종양 미세환경(TME) 내

OV 작용 기전에 초점을 맞추었으나14,15,16,17,18,

OV 치료와 다른 장기 또는 조직 간의 상호작용은 거의 탐구되지 않았다.

Much evidence indicates that deregulated microbiota plays a critical role in affecting cancer progression and immunotherapy19,20,21,22,23,24,25. For example, a recent report shows that microbial metabolite enhances immune checkpoint inhibitor efficacy by modulating T cell stemness in pan-cancer26. Another report suggested that oral reovirus reshapes the gut microbiome and enhances antitumor immunity in colon cancer27. Oncolytic immunotherapy regarding microbiota deregulation and its impact on tumor microenvironment in HCC has not been reported.

Here, we show that the efficacy of oncolytic virotherapy is attenuated by the presence of microbiota and subsequent compromised antitumor immune response. Indeed, oncolytic virus VSVΔ51 treatment leads to altered gut microbiome homeostasis, reflecting that microbiota is involved in regulating the efficacy of oncolytic virotherapy. Microbiota profiling shows that specific bacteria strains changes during VSVΔ51 treatment. Especially, VSVΔ51 affect L. acidophilus attachment and colonization by downregulating the expression‎ of SLC20A1, the receptor from the intestinal epithelium for L. acidophilus. VSVΔ51-mediated decrease of L. acidophilus attachment leads to the augmentation of other commensal bacteria. Supplementation of L. acidophilus reprogrammed microbiome homeostasis, thereby enhancing multiple OVs’ antitumor effect, including VSVΔ51, HSV-1 armed with PD-1 antibody and Vaccina virus. In this work, we reveal a previously uncharacterized role of microbiota in oncolytic virotherapy, which provides a possible mechanistic basis for future strategies to improve the efficacy of oncolytic virotherapy.

수많은 증거는 조절되지 않은 미생물 군집 deregulated microbiota 이

암 진행과 면역치료에 영향을 미치는 데

핵심적인 역할을 한다는 점을 시사한다19,20,21,22,23,24,25.

예를 들어,

최근 보고서에 따르면

미생물 대사산물이 범암종에서 T 세포 줄기세포성을 조절함으로써

면역 체크포인트 억제제 효능을 증강시킨다26.

또 다른 보고서는 경구용 레오바이러스가

장내 미생물군집을 재구성하고 대장암에서 항종양 면역력을 증강시킨다고 제안했다27.

간세포암종(HCC)에서 미생물군집 조절 장애와

종양 미세환경에 미치는 영향에 관한

종양용해 면역요법은 아직 보고된 바 없다.

본 연구에서는

미생물군집의 존재와 이에 따른 항종양 면역 반응 저하가

종양용해 바이러스 치료의 효능을 약화시킨다는 사실을 입증하였다.

실제로 종양용해 바이러스 VSVΔ51 치료는

장내 미생물군집 항상성 변화를 유발하며,

이는 미생물군집이 종양용해 바이러스 치료 효능 조절에 관여함을 반영한다.

미생물군 프로파일링 결과,

VSVΔ51 치료 중 특정 박테리아 균주의 변화가 관찰되었다.

특히 VSVΔ51은

장 상피세포의 L. acidophilus 수용체인 SLC20A1 발현을 하향 조절함으로써 

L. acidophilus의 부착 및 정착에 영향을 미쳤다.

VSVΔ51에 의한 L. acidophilus 부착 감소는

다른 공생 박테리아의 증식으로 이어졌다. 

L. acidophilus 보충은

미생물군집의 항상성을 재구성하여 VSVΔ51, PD-1 항체를 탑재한 HSV-1,

백신 바이러스를 포함한 다중 종양용해바이러스(OV)의 항종양 효과를 증강시켰다.

본 연구는

종양용해바이러스 치료에서 미생물군집의 기존에 알려지지 않은 역할을 규명함으로써,

향후 종양용해바이러스 치료의 효능을 향상시키기 위한 전략의 가능한 기전적 근거를 제공한다.

Results

Gut microbiota affect the efficacy of VSVΔ51-mediated oncolytic virotherapy in HCC

To explore the role of microbiota in oncolytic virotherapy, we established subcutaneous and orthotopic liver cancer mice models (4 groups in H22, Hepa1-6) using an antibiotic cocktail (ABX) of vancomycin, cilastatin, neomycin and amphotericin to abolish the microbiota composition: (i) PBS, (ii) ABX, (iii) VSVΔ51 (OVs), (iv) VSVΔ51 plus ABX (Fig. 1a). The data shows that VSVΔ51 plus ABX group had better antitumor efficacy than VSVΔ51 in terms of decreased tumor burden and proliferation, smaller tumor area and better survival (Fig. 1b–d and Supplementary Fig. 1a). Tumors from VSVΔ51 plus ABX group had reduced Ki-67 and elevated cleaved-Caspase-3 staining (Supplementary Fig. 1b, c). H22 tumor oncolytic virotherapy experiments were performed in germ-free (GF) mice and conventional specific pathogen-free (SPF) mice. Like oral antibiotics treated mice (to abolish microbiota), oncolytic virotherapy is much more efficient in GF mice than SPF mice (Fig. 1e) in terms of relative tumor proliferation rate (T/C%) and tumor mass measurement (Fig. 1f and Supplementary Fig. 1d). Tumors from GF mice had significant reduced Ki-67 and elevated cleaved-Caspase-3 staining (Supplementary Fig. 1e). Furthermore, Fecal Microbiota Transplantation (FMT) of VSVΔ51 treated tumor-bearing mice feces (FMTVSVΔ51) to ABX-treated mice for oncolytic virotherapy experiments (Fig. 1g) results in compromised antitumor effect of OVs, suggesting that feces from VSVΔ51 treated tumor-bearing mice contained microbiome that reduced efficacy of oncolytic virotherapy (Fig. 1h and Supplementary Fig. 1g).

결과

장내 미생물군이 간세포암종(HCC)에서 VSVΔ51 매개 종양용해 바이러스 치료의 효능에 미치는 영향

미생물군이 종양용해 바이러스 치료에서 하는 역할을 탐구하기 위해, 우리는 반코마이신, 실라스타틴, 네오마이신 및 암포테리신으로 구성된 항생제 칵테일(ABX)을 사용하여 미생물군 구성을 제거한 피하 및 동종 이식 간암 마우스 모델(H22, Hepa1-6에서 4개 그룹)을 확립했습니다: (i) PBS, (ii) ABX, (iii) VSVΔ51 (OVs), (iv) VSVΔ51 + ABX (그림 1a). 데이터에 따르면 VSVΔ51 + ABX 군은 종양 부하 및 증식 감소, 종양 면적 축소, 생존율 향상 측면에서 VSVΔ51 단독군보다 우수한 항종양 효능을 보였다(그림 1b–d 및 보충 그림 1a). VSVΔ51 + ABX 군의 종양에서는 Ki-67 염색이 감소하고 절단된 카스파제-3 염색이 증가했다(보충 그림 1b, c). H22 종양에 대한 종양용해성 바이러스 치료 실험은 무균(GF) 마우스와 일반적 특정 병원체 무감염(SPF) 마우스에서 수행되었다. 경구 항생제 치료 마우스(미생물군 제거 목적)와 마찬가지로, 종양용해 바이러스 치료는 상대적 종양 증식률(T/C%) 및 종양 질량 측정(그림 1f 및 보충 그림 1d) 측면에서 SPF 마우스보다 GF 마우스에서 훨씬 더 효율적이었습니다(그림 1e). 무균 마우스의 종양은 Ki-67 염색이 현저히 감소하고 절단된 카스파제-3 염색이 증가했습니다(보충 그림 1e). 또한, 항암 바이러스 치료 실험을 위해 항생제 처리 마우스에 VSVΔ51 처리 종양 보유 마우스의 분변(FMTVSVΔ51)을 분변 미생물 이식(FMT)한 결과(그림 1g), 항암 바이러스의 항종양 효과가 저하되었는데, 이는 VSVΔ51 처리 종양 보유 마우스의 분변에 항암 바이러스 치료의 효능을 감소시키는 미생물 군집이 포함되어 있음을 시사한다 (그림 1h 및 보충 그림 1g).

Fig. 1: Depletion of the microbiota enhances anti-tumor effects of VSVΔ51.

a Schematic of the experimental design. Antibiotics (ABX) or water were intragastric (i.g.) injected 3 times and subsequently added into drinking water per os (p.o.) after tumor cells injection; VSVΔ51 were intravenously administrated (1 × 107 PFUs, i.v.) 3 times after ABX treatment. Tumor weight was calculated. Created in BioRender. Yang, X. (2025) https://BioRender.com/d33a515. b H22 and Hepa1-6 were subcutaneously injected into mice and tumor growth was measured (n = 5/group). c Tumor mass were shown (n = 5/group). d H22 was orthotopically injected into the liver of mice. Representative images of tumor (n = 3/group). Kaplan-Meier survival curves were shown (n = 5/group). e H22 tumor growth curves in GF or SPF mice (n = 5-6/group). f The relative tumor proliferation rate (T/C%) and tumor mass were shown (n = 5/group). g Schematic diagram showing administration strategies of FMT. Created in BioRender. Yang, X. (2025) https://BioRender.com/k99v786. h Representative images of tumor were shown. Dashed yellow lines represent tumor areas and the areas were calculated. Kaplan-Meier survival curves were shown (n = 5/group). i Immunohistochemistry analysis for the VSV-G in tumor samples (Scale bar, 200 μm). Every spot presents a mean counting number from three fields (n = 5/group). j Biodistribution of VSVΔ51 in tumor and normal tissues 3 days post infection (n = 3/group). k, l Comparison of dendritic cells (F4/80- MHCII+ CD11c+), CD8+ T cells (CD8+), Granzyme B+ cells, IFN-γ+ cells and TNF-α+ cells among CD8+ T cells, mean fluorescence intensity (MFI) of PD-1+, TIGIT+ and TIM-3+ cells among CD8+ T cells and regulatory T cells (CD25+ FOXP3+) among CD4+ T cells in the TME were presented (n = 5/group). b–l Unpaired, two-tailed Student’s t-test. c, f One-way ANOVA with Tukey post-hoc test were used. d, h Kaplan-Meier survival curves, and statistical differences between curves were analyzed using the log-rank test. Results are presented as mean ± S.D. Each spot represents one subject. ns, not significant. Source data are provided in the Source Data file. See also Supplementary Fig. 1.

Full size image

We found that VSVΔ51 replication was more efficient in tumor tissues of GF mice than that of SPF group based on VSV-G expression‎ (Fig. 1i), suggesting that microbiota may have deferred impact on viral replication. Indeed, we found that OVs loading is significantly higher in various tissues from GF mice/ABX-treated mice than that of SPF/control mice (Fig. 1j). Notably, we found that the viral RNA accumulates in tumor samples than that of other tested organs or tissues by up to ~2000 times including heart, spleen, and lung, suggesting that the enhanced viral replication by absence/reduction of microbiome is selectively enriched in tumor (Fig. 1j). H&E staining shows that VSVΔ51 treatment did not induce significant pathology changes in major organs (heart, liver, spleen, lung, kidney, and brain) from both SPF/GF and control/ABX mice, suggesting the safety of the therapeutics (Supplementary Fig. 1h).

Oncolytic viruses can induce a potent systemic and potentially durable antitumor immunity14,28. We observed that the enhanced replication of VSVΔ51 induced by microbiome depletion stimulates stronger antitumor immunity. Dendritic cells (DCs) can process and present tumor antigens to T cells for initiating anti-tumor response; we then found that VSVΔ51 treatment led to more DCs infiltration in GF mice when compared with SPF mice (1.86 × vs 1.17 ×) (Fig. 1k). Next, while both GF and SPF mice do not increase the number of CD8+ T cells, cytotoxic effector CD8+ T cells (Teff) are augmented, including Granzyme B+ (GF mice 2.43 × vs SPF mice 1.84 ×), IFN-γ+ (GF mice 1.33 × vs SPF mice 1.19 ×) and TNF-α + (GF mice 1.61 × vs SPF mice 1.43 ×) CD8+ T cells. By contrast, exhausted CD8+ T cells (Tex) are further decreased, including PD-1+ (GF mice 0.28 × vs SPF mice 0.59 ×), TIGIT+ (GF mice 0.49 × vs SPF mice 0.76 ×) and TIM-3+ (GF mice 0.24 × vs SPF mice 0.53 ×) CD8+ T cells (Fig. 1k). Regulatory CD4+ T cells (Treg) were also decreased (GF mice 0.67 × vs SPF mice 0.71 ×) (Fig. 1l and Supplementary Fig. 1i).

Taken together, all these data suggest that the existence of microbiota can decrease VSVΔ51 replication and compromise oncolytic effect in mouse liver cancer models.

VSVΔ51 alters gut microbiome homeostasis including gut transit time, metabolome, and microbiota composition

Since microbiota altered by VSVΔ51 restricts the antitumor effect, we hypothesized that the physiological outcomes caused by VSVΔ51–altered gut microbiome may be related to this compromised antitumor effect (Fig. 2a). We observed that VSVΔ51 treatment significantly decreased fecal output and pellet water content, suggesting that constipation occurred (Fig. 2b). We confirmed that small intestine transit is decreased, while whole gut transit time is increased (Fig. 2c). Moreover, VSVΔ51 treatment dramatically changes microbial metabolites based on metabolomic analysis (Fig. 2d, e). Amino acid metabolism and vitamins metabolism pathways were top ranked (Fig. 2f). Therefore, these data led us to study the composition of bacteria regulated by VSVΔ51 systematically.

Fig. 2: VSVΔ51 treatment modulates gut homeostasis, changing gut transit time, metabolome, and microbiota composition.

a Schematic of the experimental design. Created in BioRender. Yang, X. (2025) https://BioRender.com/r48z536. b Fecal output and water content in feces were shown (n = 4/group). c Small intestine transit (%) and whole gut transit time were shown (n = 4/group). d Score plot presenting the metabolic separation for pair splitting between groups through oPLS-DA (n = 4/group). e Volcano plot showing the differential metabolites in the comparison of VSVΔ51 to control. f Overview of the top 20 terms in the metabolic enrichment analysis of the differential metabolites. g Unsupervised PCoA clustering analysis of the feces bacteria using unifrac distance. MRPP test P-value is 0.0001. h Alpha-diversity Chao1 index of 16S sequencing data (n = 9 for ABX; n = 10 for other groups). Data are presented as box plots (center line at the median, upper and lower bounds are 75th and 25th percentile with whisker at 1.5 IQR). i Stacked bar plot showing the bacteria relative abundance at the phylum and genus level in various samples (n = 9 for ABX; n = 10 for other groups). j, k Proportional Chord Diagram showing the top-ranked significantly changed bacteria in two repeated experiments. l, m LEfSe analysis showed differences in bacterial profiles among the two groups (n = 9–10/group) in two repeated experiments. b, c Unpaired, two-tailed Student’s t-test. e, f One-tailed Student’s t-test. h One-way ANOVA with Tukey post-hoc test were used. Results are presented as mean ± S.D. Each spot represents one subject. Source data are provided in the Source Data file.

Full size image

Microbiota composition was analyzed by 16S ribosomal RNA (rRNA) gene sequencing from the indicated mouse groups. Principal coordinate analysis (PCoA) showed that the microbiota composition of VSVΔ51 plus ABX group was significantly distinct from the control and VSVΔ51 groups (Fig. 2g). Bacteria richness and evenness were estimated using the Chao1 index. The data show that the alpha diversity was significantly increased in VSVΔ51 treatment group. Meanwhile, ABX treatment decreased the alpha diversity (Fig. 2h). Furthermore, the differentially distributed microbial community was characterized among the groups. When comparing the microbiota of VSVΔ51 treatment and non-treated mice via taxonomic profiling, we observed apparent changes at genus and phylum level (Fig. 2i). The richness of top ranked bacteria was shown (Fig. 2j, k). We performed linear discriminant analysis (LDA) effect size (LEfSe) and found that VSVΔ51 treatment increased the richness of lots of bacteria, including Barnsiella, Clostridium_XIVa, Akkermansia, Desulfovibrio, Acetatifactor, Clostridium_IV, Alistipes and Rikenella. In contrast, Lactobacillus, Blautia and Klebsiella were significantly decreased after VSVΔ51 treatment (Fig. 2l). We confirmed the observation and found that most of the bacteria could be affected by VSVΔ51 treatment in the repeated experiments, such as Clostridium_XIVa, Acetatifactor, Clostridium_IV, Alistipes, Lactobacillus and Blautia (Fig. 2m). Collectively, these data suggest a strong impact on the gut microbiota composition after oncolytic virotherapy.

The presence of gut microbiome is attenuating the antitumor effect of VSVΔ51

Despite gut microbiota deregulation contributes to the tumor progression, tumor-resident microbiota is an emerging tumor component that has been documented for a variety of cancer types29,30. To distinguish the contribution of tumor-resident bacteria and gut microbiota in affecting the antitumor effect of VSVΔ51, we use distinct combinations of antibiotics and administration routes to selectively eliminate tumor-resident bacteria or gut microbiota according to previous report31 (Fig. 3a). Mice with orthotopic H22 tumors were treated with (i) PBS control, (ii) intravenous ABX (i.v.), (iii) gavage ABX (i.g.), (iv) intravenous VSVΔ51, (v) a combination of intravenous VSVΔ51 plus ABX i.v., (vi) a combination of intravenous VSVΔ51 plus ABX i.g. (Supplementary Fig. 2a). We collected the feces and liver cancer tissues and performed 16S rRNA sequencing. Principal coordinate analysis (PCoA) showed that the gut microbiota composition of (ABX i.g.) co-clustered with the VSVΔ51 plus ABX i.g. samples very well but was significantly distinct from the other four groups (Supplementary Fig. 2b). However, due to the abundance of microbiome in liver is relatively low, all of the six groups from liver tumors were co-clustered (Supplementary Fig. 2b). Next, alpha diversity analysis shows that ABX (i.g.) can eliminate both gut and tumor-resident microbiota. However, ABX (i.v.) majorly eliminates tumor-resident microbiota while the gut microbiota is left intact (Fig. 3b). Bacteria composition analysis confirmed that ABX (i.g.) can eliminate both gut microbiota and tumor-resident microbiota when compared with ABX (i.v.), which only eliminate tumor-resident microbiota (Fig. 3c, d, Supplementary Fig. 2c and 2d). We next tested the localization of bacteria in the liver cancer by 16S fluorescence in situ hybridization (FISH) analysis. We observed that VSVΔ51 treatment increased the tumor-resident bacteria. Importantly, both ABX (i.g.) and ABX (i.v.) treatments eliminated tumor-resident bacteria (Fig. 3e), which is consistent with the 16S rRNA sequencing results. LEfSe analysis shows that VSVΔ51 treatment increased the richness of lots of bacteria in feces, such as Alistipes, Barnsiella, Clostridium_XIVa, Clostridium_IV, Rikenella, and Acetatifactor. Importantly, Lactobacillus was significantly decreased after VSVΔ51 treatment (Fig. 3f and Supplementary Fig. 2e). Meanwhile, lots of liver cancer tissue-resident bacteria were increased after VSVΔ51 treatment, such as Barnsiella, Bacteroides, Oscillospira, Allobaculum, Clostridium_XIVa, Coprococcus, Alistipes and Oscillibacter. Abundance of Robinsoniella was decreased after VSVΔ51 treatment (Fig. 3g and Supplementary Fig. 2f). Together, these data suggest that VSVΔ51 treatment can affect the composition of both gut and tissue-resident microbiota.

Fig. 3: The presence of gut microbiome compromises the antitumor effect of VSVΔ51.

a Schematic diagram showing administration strategies of antibiotics and the influences on the gut and tissue microbiota. Created in BioRender. Yang, X. (2025) https://BioRender.com/q52l414. b Alpha-diversity Chao1 index of feces (n = 10/group) and liver tumor-resident bacteria (n = 6 for control, i.g., VSVΔ51 and i.g. + VSVΔ51, n = 7 for other groups). Data are presented as box plots (center line at the median, upper and lower bounds are 75th and 25th percentile with whisker at 1.5 IQR). c, d Heatmap representation of differentially abundant bacterial species in feces and liver tumor. e 16S FISH analysis of orthotopic tumor. Red, 16S FISH probe. Blue, DAPI. Scale bar, 20 μm (n = 5/group). f, g LEfSe analysis showing differences in bacterial profiles among feces and liver tumor between the two groups. h Representative images of tumor from each group were shown (n = 4/group). Dashed yellow lines represent tumor areas and the areas were calculated. Kaplan-Meier survival curves were shown (n = 5/group). i At the end of the experiment, the mice were sacrificed, and immune cells from TME were analyzed by flow cytometry. Statistical data for the proportion of dendritic cells, CD8+ T cells, Granzyme B+ cells, IFN-γ+ cells and TNF-α+ cells among live cells in the TME were presented (n = 5/group). b–i One-way ANOVA with Tukey post-hoc test were used, Kaplan-Meier survival curves, and statistical differences between curves were analyzed using the log-rank test. Results are presented as mean ± S.D. Each spot represents one subject. Source data are provided in the Source Data file. See also Supplementary Fig. 2.

Full size image

With these different antibiotic treatment strategies, we next aimed to address the role of gut microbiota and tumor-resident microbiota in oncolytic virotherapy. The results show that: (i) i.v. (ABX) treatment and i.g. (ABX) treatment can both enhance the oncolytic effect of VSVΔ51; (ii) i.g. (ABX) treatment enhanced antitumor effects of VSVΔ51 is better than i.v. (ABX) treatment (Fig. 3h); (iii) i.g. (ABX) treatment significantly induces smaller tumor size, better mice survival, increased infiltration of DCs (F4/80- MHCII+ CD11c+) and cytotoxic CD8+ T cells (Granzyme B+, TNF-α+, and IFN-γ+), decreased infiltration of regulatory T cells (CD25+ FOXP3+) and exhausted CD8+ T cells (TIGIT+), as well as decreased Ki-67 staining and increased cleaved-Caspase-3 staining when compared with i.v. (ABX) during VSVΔ51 oncolytic virotherapy (Fig. 3i and Supplementary Fig. 2g-2i). These data collectively demonstrate that gut microbiota plays a much crucial role in compromising the antitumor effect of VSVΔ51 when compared with the tumor-resident microbiota.

Strain-by-strain profiling shows that specific bacteria strains mitigate VSVΔ51 antitumor effect

Next, we sought to examine which strain can modulate OVs’ antitumor effect. We designed a co-culture screen in which HuH7 cell line was co-cultured with bacterial strains from 10 phylum (27 bacterial strains) that were regulated by VSVΔ51 in vivo (refer to Fig. 2l). (i) Tumor cell viability was assessed after treatment with increasing multiplicity of infection (MOI) of VSVΔ51 plus various bacterial strains. EC50 (median effect dose) shift and differences in the area under the curve (DAUCs) for each bacterial strain were calculated. (ii) Viral fluorescence intensity (GFP, flow cytometry) and (iii) mRNA levels (qPCR) were applied to identify intracellular viral content. (iv) TCID50 was performed to detect the secreted viral particles (Fig. 4a).

Fig. 4: Oncolytic virotherapy-regulated bacteria can mitigate VSVΔ51 antitumor effect.

a An outline of bacteria screening protocol. HuH7 cells were treated with increasing MOI of VSVΔ51 with or without bacterium from the community. Cell viability (at VSVΔ51 for 48 h), viral GFP, viral RNA (VSV-G) and viral titer (at VSVΔ51 for 12 h) were determined. The DAUC was calculated according to the following formula: (AUC VSVΔ51 − AUC VSVΔ51 + Bacteria)/AUC VSVΔ51 + Bacteria; FCM, Flow Cytometry. Created in BioRender. Yang, X. (2025) https://BioRender.com/j49g663. b Phylogenetic tree of Bacteria and the screening results were shown. c Correlation analysis of DAUCs with viral mRNA (95% CI = [0.0000, 0.8760]), viral GFP (95% CI = [0.2784, 0.6318]) or titer (95% CI = [0.0000, 0.8120]), used Pearson correlation coefficient. Source data are provided in the Source Data file. See also Supplementary Fig. 3.

Full size image

Remarkably, the results show that most bacteria upregulated by VSVΔ51 can attenuate the oncolytic efficacy (defined as DAUC < -0.3, bacteria strain ratio: 13/27) (Fig. 4b and Supplementary Fig. 3). These bacteria characterized are Alistipes strains, Rikennella microfusus, Klebsiella oxytoc, Ligilactobacillus animals, Lactobacillus salivarius strain, Lactobacillus fermentum, Clostridium methylpentosum, Clostridium scindens, Clostridium boltei and Blautia hansenii. Accordingly, intracellular viral loading and secreted viral particles were also decreased. By contrast, Barnesiella intestinihominis and Lactobacillus apodemi can enhance OVs’ antitumor effect (defined as DAUC > 0.3, bacteria strain ratio: 2/27) (Supplementary Fig. 3), with concurrent elevation of viral replication. The other bacteria have no impact on antitumor effect of OVs (defined as -0.3 < DAUC < 0.3, bacteria strain ratio: 12/27) (Fig. 4b and Supplementary Fig. 3). We observed that DAUC is positively correlated with viral mRNA, viral GFP and viral titer, suggesting that bacteria-mediated oncolytic effect regulation is dependent on viral replication (Fig. 4c). On the basis of the study, we found that most of the strains analyzed can limit the oncolytic effect. Together, these data strongly demonstrate that based on bacteria strain profiling, most bacteria strains can mitigate oncolytic virotherapy while few bacterial strains are posing effect to enhance oncolytic virotherapy.

L. acidophilus supplementation boosts up the antitumor efficacy of oncolytic virothrapies

Given that lots of bacteria attenuated the antitumor effect of VSVΔ51 in oncolytic virotherapy assay, we wondered whether potential probiotics could enhance oncolytic virotherapy through harnessing microbiota homeostasis32. Microbial profiling reveals that Lactobacillus is negatively correlated with lots of bacteria, including Barnsiella, Blautia, Clostridium_XIVa, Akkermansia, Desulfovibrio, Acetatifactor, Clostridium_IV, Alistipes and Rikenella (Fig. 5a). Furhter searching probiotics from Lactobacillus genus, we found that Lactobacillus acidophilus (L. acidophilus) is one of the most widely used probiotics owing to its regulating role of the balance of intestinal flora33,34. FISH assay revealed that VSVΔ51 treatment increases total bacteria abundance while reduces L. acidophilus abundance in mouse intestines (Fig. 5b). This observation led us to hypothesize whether L. acidophilus supplementation can reprogram the gut microbiota homeostasis and enhance oncolytic effect of VSVΔ51.

Fig. 5: L. acidophilus supplementation enhances oncolytic effect of oncolytic virohterapy.

a Mantel test analysis of the correlations between the VSVΔ51 up-regulated and down-regulated microbiota. L. acidophilus negatively correlates with other bacteria (Highlighted with red line). b The colonic abundance of L. acidophilus as determined by FISH. 16S was stained with green fluorescence (EUB338) and L. acidophilus was stained with red fluorescence (specific probe) (n = 8/group). c H22 cells were subcutaneously injected into C57 mice, VSVΔ51 were intravenously administrated (1 × 107 PFUs, i.v., 3 times) after L. acidophilus or L. acidophilus (per-heated) (2 × 108 PFUs, i.g., total 5 times) treatment 4 times. Tumor growth was measured (n = 6/group). d Mice body weight were shown (n = 6/group). e, f Gross tumor and tumor mass were shown (n = 6/group). g H22 cells were subcutaneously injected into C57 mice, HSV-1-anti-PD1 were intratumoral (i.m.) administrated (1 × 107 PFUs, i.m., 3 times) after L. acidophilus (2×108 PFUs, i.g., total 5 times) treatment 4 times. Tumor growth was measured (n = 6/group). h Mice body weight were shown (n = 6/group). i, j Gross tumor and tumor mass were shown (n = 6/group). k H22 cells were subcutaneously injected into C57 mice, VACV(WR)-GFP were Intratumoral administrated (1 × 107 PFUs, i.m., 3 times) after L. acidophilus (2 × 108 PFUs, i.g., total 5 times) treatment 4 times. Tumor growth was measured (n = 6/group). Tumor growth was measured (n = 6/group). l Mice body weight were shown (n = 6/group). (m and n) Gross tumor and tumor mass were shown (n = 6/group). (b) Unpaired, two-tailed Student’s t-test. c–l Two-way ANOVA. f–n One-way ANOVA with Tukey post-hoc test. Results are presented as mean ± S.D. Each spot represents one subject. Source data are provided in the Source Data file.

Full size image

As previous report shows that either alive or heat inactivated probiotics can regulate gut microbiota composition35, we next tested that if pre-heated L. acidophilus and/or alive L. acidophilus could enhance antitumor efficacy of VSVΔ51. Tumor bearing mice (C57 mice were subcutaneously injected H22 cell line) were treated with (i) control, (ii) L. acidophilus, (iii) VSVΔ51, (iv) L. acidophilus + VSVΔ51, (v) L. acidophilus (pre-heated) and (vi) L. acidophilus (pre-heated) + VSVΔ51. The results showed that mice treated with L. acidophilus + VSVΔ51 or L. acidophilus (pre-heated) + VSVΔ51 had significant slower tumor growth (Fig. 5c) when compared with other groups, suggesting that either alive or pre-heated L. acidophilus can significantly enhance the antitumor efficacy of VSVΔ51-mediated oncolytic virotherapy. At the same time, all of the treatment does not affect mice weight (Fig. 5d). From tumor size observation at the end point, we observed that alive L. acidophilus + VSVΔ51 and L. acidophilus (pre-heated) + VSVΔ51 had smallest tumor size compared with other group (Fig. 5e, f). However, the antitumor effect of alive L. acidophilus + VSVΔ51 is better than L. acidophilus (pre-heated) + VSVΔ51.

Next, we wondered whether harnessing micriobiota could affect the antitumor effect of other classical oncolytic viruses, tumor bearing mice were treated with HSV-1 armed with anti-PD-1 antibody or VACV(WR)-GFP with or without L. acidophilus. Our results shows that L. acidophilus could significantly enhance antitumor efficacy of HSV-1-anti-PD-1 or VACV(WR)-GFP, as evidenced by decreased tumor growth curve, reduced tumor size and without affecting the mice weight at the end point (Fig. 5g–n). These data strongly indicate that L. acidophilus supplementation could significantly enhance the antitumor effect of oncolytic viruses.

L. acidophilus supplementation facilitates antitumor immune response of VSVΔ51 through reprogramming microbiome

Next, we asked whether enhanced oncolytic virotherapy by L. acidophilus is dependent on the existence of gut microbiota. We established two HCC mice models, H22 cell line intra-liver injection orthotopic model and a hydrodynamic tail vein injection (HDTI) genetic HCC mice model. HA-myr-Akt, N90-β-catenin, and Sleeping Beauty (SB) transposon constructs were delivered into the liver of C57 mice by hydrodynamic tail vein injection to establish HCC model36,37. Mice were treated with (i) control, (ii) L. acidophilus, (iii) VSVΔ51, (iv) L. acidophilus + VSVΔ51 in the presence or absence of antibiotics (Supplementary Fig. 4a).

The results showed that in the presence of microbiota (-ABX), mice treated with the L. acidophilus + VSVΔ51 combination had the smallest tumor area and prolonged survival when compared with other groups in both H22 orthotopic model and HDTI model (Fig. 6a–c). Strikingly, the combination group shows total tumor regression in HDTI model (Fig. 6d–f). The combination group had decreased tumor area as assayed by H&E staining, decreased Ki-67 level, and increased cleaved-Caspase-3 level (Supplementary Fig. 4b and 4c). Meanwhile, the combination group had increased cytotoxic CD8+ T cells (TNF-α+, IFN-γ+ and Granzyme B+) (Fig. 6g), decreased exhausted CD8+ T cells, and decreased regulatory CD4+ T cells (CD25+ FOXP3+) in the tumor microenvironment (Supplementary Fig. 4d). In the presence of ABX, the L. acidophilus + VSVΔ51 combination group was no different from VSVΔ51 group in terms of antitumor effect as evidenced by tumor area, mice survival, effector CD8+ T cells, exhausted CD8+ T cells and regulatory CD4+ T cells (Fig. 6a–g, Supplementary Fig. 4b–d).

Fig. 6: L. acidophilus supplementation enhances antitumor immune response of VSVΔ51 via changing the landscape of microbiota.

a Orthotopic mice models were established (H22 cell line). Representative images of tumor from each group were shown (n = 2/group). b Dashed yellow lines represent tumor areas and the areas were calculated (n = 4). c Kaplan-Meier survival curves were shown (n = 5/group). d Orthotopic mice models were established (spontaneous HCC by HDTI). Representative images of tumor from each group were shown (n = 2/group). e Dashed yellow arrow represent tumor areas and the areas were calculated (n = 4/group). f Kaplan-Meier survival curves were shown (n = 5/group). g At the end of the experiment, the mice were sacrificed, and immune cells from TME were analyzed by flow cytometry. Statistical data for the proportion of CD8+ T cells among CD3+ T cells, CD4+ T cells (CD3+ CD4+), Granzyme B+ cells (CD3+ CD8+ GZMB+), IFN-γ+ (CD3+ CD8+ IFN-γ+), TNF-α+ (CD3+ CD8+ TNF-α+) were shown (n = 5/group). h Unsupervised PCoA clustering analysis of the HDTI mice (without ABX treatment) feces bacteria using unifrac distance. MRPP test p-value is 0.0001(n = 9/group). i Alpha Diversity analysis was shown. Data are presented as box plots (center line at the median, upper and lower bounds are 75th and 25th percentile with whisker at 1.5 IQR). j Heatmap showing differentially abundant bacterial species in mice. b–i One-way ANOVA with Tukey post-hoc test. c, f Kaplan-Meier survival curves, and statistical differences between curves were analyzed using the log-rank test. Results are presented as mean ± S.D. Each spot represents one subject. Source data are provided in the Source Data file. See also Supplementary Figs. 4 and 5.

Full size image

Moreover, through 16S rRNA sequencing, L. acidophilus supplementation in VSVΔ51-treated mice changes microbiota composition when compared with VSVΔ51-treated mice as demonstrated by PCoA analysis and alpha diversity analysis (Fig. 6h, i). Strikingly, L. acidophilus supplementation in VSVΔ51-treated mice diminished the abundance of bacteria upregulated by VSVΔ51-treated mice, including Alistipes, Acetatifactor, Rikenella, Clostridium_IV and Clostridium_XIVa (Fig. 6j). RNA sequencing and Gene Set Enrichment Analysis (GSEA) analyses from VSVΔ51-treated mouse gut epithelial tissues were performed. We found that cell adhesion-related pathways were enriched in VSVΔ51-treated mice group based on KEGG analysis and Gene Ontology analysis (Supplementary Fig. 5a-5d). VSVΔ51-treated mice markedly decreased the expression‎ of tight junction mRNA, including Tjp1 (ZO-1), Ocln (Occludin), and Cdh1 (E-cadherin) (Supplementary Fig. 5e). L. acidophilus supplementation in VSVΔ51-treated mice restores the gene expression‎ of these three genes downregulated by VSVΔ51 (Supplementary Fig. 5e). Accordingly, Occludin1 and E-cadherin expression‎ diminished by VSVΔ51 was rescued by L. acidophilus supplementation in VSVΔ51-treated mice (Supplementary Fig. 5f). Detailed assessment using transmission electron microscopy (TEM) demonstrated that L. acidophilus supplementation in VSVΔ51-treated mice restored the fragmented/destroyed cell-cell junction caused by VSVΔ51 (Supplementary Fig. 5g). Taken together, these data indicate that L. acidophilus reduced by VSVΔ51 might lead to the deregulation of gut barrier (tight junction) function and alteration of microbiota composition, thereby compromising the efficacy of oncolytic virotherapy. Supplementation of L. acidophilus can strengthen oncolytic virotherapy efficacy of VSVΔ51 through modulating microbiota composition and restore gut barrier function.

VSVΔ51 downregulates SLC20A1 via IL-6-STAT3 axis to reduce the attachment of L. acidophilus to intestinal tissue and subsequent gut colonization

To investigate how VSVΔ51 reduces L. acidophilus colonization in the intestine, we collected mice epithelial tissues after VSVΔ51 treatment and performed RNA-sequencing (RNA-seq). Meanwhile, we performed biotinylated L. acidophilus pull-down assay to characterize L. acidophilus interacting epithelial proteins by mass spectrometry38,39. Based on the identification of the candidate L. acidophilus-interacting proteins (including membrane proteins) and characterization of VSVΔ51-mediated downregulated genes (including membrane genes) by RNA-seq, we revealed one major membrane protein SLC20A1 could be the candidate membrane protein interacting with L. acidophilus and was downregulated by VSVΔ51 (Fig. 7a, Supplementary Fig. 6a, b). Far-Western results showed that a potential L. acidophilus -interacting protein near 75 kDa (SLC20A1) was decreased after VSVΔ51 treatment (Fig. 7b).

Fig. 7: VSVΔ51 downregulates SLC20A1, thereby decreasing the attachment of L. acidophilus to intestinal tissue and subsequent gut colonization.

a Schematic of the screening of Lactobacillus receptor. RNA-seq was performed from mice intestinal epithelial tissues with or without VSVΔ51 treatment and differentially expressed genes (DEGs) were analyzed. L. acidophilus proteins were collected and labeled with biotin. Streptavidin pull-down and mass spectrometry were performed. Venn diagram was performed based on DEGs, pull-down proteins and membrane proteins, which SLC20A1 was identified.Created in BioRender. Yang, X. (2025) https://BioRender.com/s68y847. b Biotinylation-based Far-Western detected a protein band loss in VSVΔ51 treatment group. c–f KD of SLC20A1 showing reduced attachment of L. acidophilus on NCM460 cells as determined by adhesion assay (n = 3/group), TEM (n = 5/group), bacteria staining with IF (n = 5/group) and SEM (n = 4/group). Quantitation data of each assay for the bacterial count were shown. Dox, doxycycline (200 ng ml-1). g TOPCONS analysis of SLC20A1 membrane protein topology. h The second domain (78-158 aa) of SLC20A1 was bound to the L. acidophilus. NCM460 protein lysis was subjected to SDS-PAGE; Biotin-labeled L. acidophilus proteins were incubated and far-western was performed. i The attachment of L. acidophilus as determined by adhesion assay(n = 4/group). j SLC20A1 binding proteins from L. acidophilus protein were determined by IP-MS. k Immunoprecipitation analysis of the binding between SLC20A1 and CdpA in HEK-293T cells. l The second domain (78–158 aa) of SLC20A1 binds to CdpA in HEK-293T cells based on Immunoprecipitation analysis. m Modeling the binding between the second domain of SLC20A1 and CdpA complex by SWISSMODEL.EXPASY and PyMOL. c, i Two-tailed Student’s t-test. f One-way ANOVA. Results are presented as mean ± S.D. Each spot represents one subject. Source data are provided in the Source Data file. See also Supplementary Fig. 6.

Full size image

To address the role of SLC20A1 on L. acidophilus colonization, we found that Dox-induced knockdown of SLC20A1 attenuated the attachment of L. acidophilus to NCM460 cell, an intestinal epithelial cell line, based on colony formation assay (Fig. 7c and Supplementary Fig. 6c). Of note, SLC20A1 KD does not affect the cell viability (Supplementary Fig. 6d and 6e). Significantly, we observed that the attachment of L. acidophilus to NCM460 cell was indeed compromised when SLC20A1 gene was knocked down in NCM460 cells based on TEM, immunofluorescence, and scanning electron microscope (SEM) studies (Fig. 7d-f). Membrane protein topology analysis shows that SLC20A1 is a sodium-dependent phosphate transporter with 5 outside membrane domains40 (Fig. 7g). We demonstrated that the second domain (78-158 amino acids, aa) of SLC20A1 is critical for L. acidophilus binding based on a domain mapping experiment (Fig. 7h). We demonstrated that the attachment of L. acidophilus to NCM460 cell transfected with SLC20A1Δ2 was decreased when compared with wildtype (WT) SLC20A1 (Fig. 7i). Several proteins of L. acidophilus that can interact with SLC20A1 were characterized by immunoprecipitation-mass spectrometry (IP-MS) including a cell wall protein CdpA (Fig. 7j). CdpA is the candidate that binds strongly with SLC20A1 among other possible candidates (Fig. 7k and Supplementary Fig. 6f). Moreover, the binding between SLC20A1 with CdpA depends on the second domain of SLC20A1 as SLC20A1Δ2 reduces its binding to CdpA (Fig. 7l). The binding between SLC20A1 second domain (78-158 aa) and CdpA was modeled using PyMOL program41 (Fig. 7m). These observations suggest that SLC20A1 of intestinal cells, which binds cell wall protein CdpA of L. acidophilus, is crucial for L. acidophilus attachment and subsequent gut colonization.

SLC20A1 was identified during intestine tissue RNA sequencing screening. To determine the mechanism behind which VSVΔ51 downregulates SLC20A1 expression‎, we first examine SLC20A1 mRNA and protein levels of NCM460 cells infected with VSVΔ51 in a dose-dependent manner. Surprisingly, VSVΔ51 treatment does not downregulate SLC20A1 mRNA and protein levels in NCM460 cell context (Supplementary Fig. 6g, h). By contrast, we confirmed that VSVΔ51 treatment indeed decrease the Slc20a1 expression‎ in intestinal tissues context as evident by qPCR and Western blot (Supplementary Fig. 6i, j). In addition to direct tumor-killing, OVs (VSVΔ51) are well-known to induce bystander killing through cytokine storms42,43. Thus, the seemingly contradictory data in different context led us to hypothesize that Slc20a1 expression‎ might be regulated by circulating inflammatory factors.

To this end, cytokines array was performed using serum from VSVΔ51-treated mice. As the data show, IL-12p40, G-CSF, IL-6 and IL-10 were significantly decreased after VSVΔ51 treatment (Supplementary Fig. 6k). Importantly, exogenous addition of these inflammatory factors, including IL-6, to cells could upregulate SLC20A1 expression‎ (Supplementary Fig. 6l). Interestingly, IL-6, G-CSF, IL-12B and IL-10 are upstream of STAT3 signaling44,45,46. We found that IL-6-mediated upregulation of SLC20A1 can be blocked with JAK-1/2-STAT3 inhibitor Ruxolitinib, suggesting that IL-6-STAT3 axis is involved in regulating SLC20A1 expression‎ (Supplementary Fig. 6m). Indeed, SLC20A1 promoter has STAT3-binding motif based on JASPAR program analysis47 (Supplementary Fig. 6n). Chromatin immunoprecipitation (ChIP) assay confirmed that IL-6 signaling enhances binding of phosphorylated-STAT3 and RNA polymerase 2 (Pol2) to the SLC20A1 promoters, suggesting the transcriptional activation of SLC20A1 by IL-6-STAT3 axis (Supplementary Fig. 6n). Together, these results demonstrated that VSVΔ51-mediated downregulation of IL-6 attenuates STAT3 signaling, thereby reducing the expression‎ of SLC20A1 and subsequent loss of L. acidophilus attachment (colonization) to intestinal cells. In summary, the oncolytic virotherapy elicits host responses, which shapes microbiome populations and inhibits viral replication. VSVΔ51 infection caused downregulation of IL-6, which leads to repression of SLC20A1 gene expression‎ through blocking STAT3 signaling. Thus, L. acidophilus attachment to SLC20A1 and subsequent colonization in gut tissue decreased, thereby reshaping the landscape of microbiota in the gut (Supplementary Fig. 7).

L. acidophilus의 정착에 대한 SLC20A1의 역할을 규명하기 위해, 도크소루비신(Dox) 유도 SLC20A1 노크다운이 장 상피 세포주인 NCM460 세포에 대한 L. acidophilus의 부착을 약화시킨다는 것을 콜로니 형성 분석을 통해 확인하였다(그림 7c 및 보충 그림 6c). 주목할 점은, SLC20A1 KD가 세포 생존율에는 영향을 미치지 않는다는 것이다(보충 그림 6d 및 6e). 중요한 것은, NCM460 세포에서 SLC20A1 유전자를 노크다운했을 때 L. acidophilus의 NCM460 세포 부착이 실제로 저해되었음을 TEM, 면역형광 및 주사전자현미경(SEM) 연구를 통해 관찰했다는 점이다(그림 7d-f). 막 단백질 토폴로지 분석에 따르면 SLC20A1은 5개의 외부 막 도메인을 가진 나트륨 의존성 인산염 수송체이다40 (그림 7g). 도메인 매핑 실험을 통해 SLC20A1의 두 번째 도메인(78-158 아미노산, aa)이 L. acidophilus 결합에 중요함을 입증하였다(그림 7h). SLC20A1Δ2를 발현하는 NCM460 세포에 L. acidophilus의 부착이 야생형(WT) SLC20A1에 비해 감소함을 입증하였다(그림 7i). 면역침전-질량분석법(IP-MS)을 통해 L. acidophilus의 여러 단백질들이 SLC20A1과 상호작용할 수 있음을 확인했으며, 여기에는 세포벽 단백질 CdpA가 포함되었습니다(그림 7j). CdpA는 다른 후보들 중 SLC20A1과 강력하게 결합하는 후보 단백질입니다(그림 7k 및 보충 그림 6f). 또한 SLC20A1과 CdpA 간의 결합은 SLC20A1의 두 번째 도메인에 의존하는 것으로 나타났는데, SLC20A1Δ2 돌연변이체는 CdpA와의 결합이 감소하였다(그림 7l). SLC20A1 두 번째 도메인(78-158 aa)과 CdpA 간의 결합은 PyMOL 프로그램41을 사용하여 모델링하였다(그림 7m). 이러한 관찰 결과는 장 세포의 SLC20A1이 L. acidophilus의 세포벽 단백질 CdpA와 결합하여 L. acidophilus의 부착 및 후속 장 내 정착에 결정적 역할을 함을 시사한다.

SLC20A1은 장 조직 RNA 시퀀싱 스크리닝 과정에서 확인되었다. VSVΔ51이 SLC20A1 발현을 하향 조절하는 기전을 규명하기 위해, 먼저 VSVΔ51에 감염된 NCM460 세포에서 SLC20A1 mRNA 및 단백질 수준을 용량 의존적으로 조사하였다. 놀랍게도, VSVΔ51 처리는 NCM460 세포 환경에서 SLC20A1 mRNA 및 단백질 수준을 하향 조절하지 않았다(보충 그림 6g, h). 반면, qPCR 및 웨스턴 블롯 분석을 통해 장 조직 환경에서 VSVΔ51 처리가 실제로 Slc20a1 발현을 감소시킨다는 사실을 확인했습니다(보충 그림 6i, j). 직접적인 종양 사멸 외에도, OV(VSVΔ51)는 사이토카인 폭풍을 통해 주변 세포 사멸을 유도하는 것으로 잘 알려져 있습니다42,43. 따라서 서로 다른 맥락에서 나타난 상반된 데이터는 Slc20a1 발현이 순환 염증 인자에 의해 조절될 수 있다는 가설을 제시하게 했다.

이를 위해 VSVΔ51 처리 마우스 혈청을 이용한 사이토카인 배열 분석을 수행했다. 데이터가 보여주듯, VSVΔ51 처리 후 IL-12p40, G-CSF, IL-6 및 IL-10이 유의하게 감소했다(보충 그림 6k). 중요한 점은, IL-6을 포함한 이러한 염증 인자들을 세포에 외인성으로 추가하면 SLC20A1 발현을 상향 조절할 수 있다는 것이다(보충 그림 6l). 흥미롭게도, IL-6, G-CSF, IL-12B 및 IL-10은 STAT3 신호전달의 상류에 위치한다44,45,46. 우리는 JAK-1/2-STAT3 억제제 룩솔리티닙으로 IL-6 매개 SLC20A1의 상향 조절을 차단할 수 있음을 발견했으며, 이는 IL-6-STAT3 축이 SLC20A1 발현 조절에 관여함을 시사합니다(보충 그림 6m). 실제로, JASPAR 프로그램 분석에 따르면 SLC20A1 프로모터에는 STAT3 결합 모티프가 존재합니다47 (보충 그림 6n). 크로마틴 면역침전(ChIP) 분석을 통해 IL-6 신호전달이 인산화-STAT3 및 RNA 중합효소 2(Pol2)의 SLC20A1 프로모터 결합을 강화함을 확인하였으며, 이는 IL-6-STAT3 축에 의한 SLC20A1의 전사 활성화 가능성을 시사한다(보충 그림 6n). 종합하면, 종양용해성 바이러스 치료법은 숙주 반응을 유발하여 미생물군집을 형성하고 바이러스 복제를 억제한다. VSVΔ51 감염은 IL-6의 하향조절을 유발하여 STAT3 신호전달을 차단함으로써 SLC20A1 유전자 발현을 억제한다. 이로 인해 L. acidophilus의 SLC20A1 부착 및 장 조직 내 정착이 감소하여 장내 미생물 군집 구성을 재편한다(보충 그림 7).

Discussion

Hepatocellular carcinoma, presenting a very dismal prognosis, is usually resistant to conventional chemotherapy and radiotherapy. Oncolytic virotherapy has been explored in HCC. Despite increasing interest of oncolytic virotherapy, the underlying biology and pharmacology of OVs are not fully understood. A better understanding regarding these issues is needed to fully employ therapeutic potential of OVs in cancer patients. We demonstrate that the existence of microbiota might be a prominent factor impacting oncolytic virotherapy.

Microbiota influence the efficacy of VSVΔ51-mediated oncolytic virotherapy

Studies show that microbial signals are required in the development of cancer, which is reminiscent of the observation that hematopoietic malignancies were compromised by antibiotic treatment and also failed to develop in germ-free mice48. Another similar observation is that germ-free condition or antibiotic-treatment significantly compromises lung cancer development in mice model49. Using antibiotic cocktail to abolish the microbiota composition can potentiate the efficacy of VSVΔ51 oncolytic virotherapy, suggesting that the existence of microbiota may be a hurdle for treatment. One can imagine that the presence of commensal bacteria (microbiota) can stimulate Myd88-dependent IL-1β and IL-23 production from myeloid cells, thereby activating T cells to produce IL-17/other effector molecules and promoting inflammation and tumor cell growth49. It is then clear that oncolytic virotherapy of VSVΔ51 encounters a barrier to preclinical effectiveness due to the presence of microbiota. Thus, tumor may become resistant, resulting in treatment failure. Indeed, we show that specific bacteria strains mitigate VSVΔ51 antitumor effect. This limitation of inducing complete tumor regression by VSVΔ51 has prompted us to explore the mechanism and new approaches targeting tumors that are resistant to oncolytic virotherapy of VSVΔ51.

Although we observed encouraging results in mice models, it should be noted that the composition of human microbiome is relatively different to mice microbiome. Gut microbiota plays an important role in liver homeostasis and HCC progression50,51. The further alteration of microbiome composition after oncolytic virotherapy in HCC patients and other cancer patients deserves further investigation. Moreover, patient-relevant models are also needed to confirm‎ our observations in future studies. Although these concerns exist, the potential clinical implications of this research are noteworthy, as our study provides a clear rationale for complementing OVs with L. acidophilus for improvement of OVs applications in HCC therapy. L. acidophilus supplementation not only restores gut microbiome balance after VSVΔ51-induced dysbiosis but also improves gut barrier function and modulates host immunity, which collectively enhances the replication and efficacy of the OVs. In all, our results uncovered an unrecognized feature of OVs: microbiome alteration plays an important role in oncolytic virotherapy.

L. acidophilus improves barrier integrity disturbed by VSVΔ51

L. acidophilus is one of the most widely used and studied probiotics owing to its favorable safety profile and efficacy52. L. acidophilus exhibits important technical properties, i.e., thermostability and retaining of activity at a wide pH range along with strong inhibitory actions against food spoilage and pathogenic bacteria make them an important class of biopreservatives53. Many report shows that L. acidophilus can regulate the balance of intestinal flora by reducing the intestinal pH and producing metabolites33,34.

During treatment, VSVΔ51 causes disturbances in the tightness of gut barrier (loss of Occludin and E-cadherin), possibly leading to the potential penetration of toxins/harmful bacterial products into the system, which leads to more systemic infections and inflammation, compromising oncolytic effect. It is not clear whether VSVΔ51-mediated elevation of TNF-α may lead to this intestinal barrier dysfunction. However, few VSVΔ51-reduced bacterial strains, such as L. acidophilus, are able to enhance VSVΔ51 oncolytic virotherapy. It is known that L. acidophilus-intestinal barrier regulation can result in the prevention of gut-barrier defects although mechanisms are not well defined54. We found that L. acidophilus indeed ameliorates the disturbance of barrier integrity by restoring the expression‎ of tight junction proteins such as Occludin and E-cadherin, with concurrently high efficacy of VSVΔ51 oncolytic virotherapy.

L. acidophilus attachment can boost up efficacy of VSVΔ51

Mechanistically, L. acidophilus binds SLC20A1/PiT1 of gut tissue for attachment. We show that VSVΔ51–mediated blocking of IL-6-STAT3 signaling is critical for SLC20A1 downregulation, thereby reducing L. acidophilus attachment to gut tissue. Significantly, L. acidophilus in combination with VSVΔ51 improves therapeutic outcomes. We find that L. acidophilus improves barrier integrity, increases cytotoxic CD8+ T cells, and decreases exhausted CD8+ T cells in the tumor microenvironment; therefore, providing preclinical proof-of-principle for enhancing oncolytic virotherapy in liver cancer.

SLC20A1/PiT1 protein is a Na (+)-dependent cotransporters for transporting phosphate across cell membranes, involving in maintaining Pi homeostasis. It is shown that SLC20A1/PiT1-depleted cells are more sensitive to the proapoptotic activity of TNF-α55. Thus, VSVΔ51-mediated down regulation of SLC20A1/PiT1 may also facilitate TNF-α-mediated cell apoptosis. Although SLC20A1 transporter is characterized as a receptor of L. acidophilus, it remains to be determined whether L. acidophilus can act through SLC20A1 transporter to potentiate some signaling pathways, such as TNF-α, for enhancing the efficacy of VSVΔ51. L. acidophilus attachment to the sites on the gut epithelium may also be affected by the VSVΔ51-mediated elevation of other symbionts, which may also directly antagonize L. acidophilus through contact-dependent inhibition, the type VI secretion system (T6SS) or secreted molecules such as bacteriocins56,57,58,59. However, these issues warrant further investigation. As for the L. acidophilus CdpA protein that interacts with SLC20A1, it is not clear that L. acidophilus CdpA has any other functions other than attachment. Whether it uses its cyclic di-GMP phosphodiesterase activity for regulating host functions warrants further investigation. For example, Burkholderia pseudomallei CdpA is characterized to be involved in auto aggregation, flagellum synthesis, motility, biofilm formation, cell invasion, and cytotoxicity60.

Taken together, these studies enable the potential development of employing combination of microbiota modifier that reprograms the tumor microenvironment, triggering better anti-tumor immune responses to enhance the efficacy of oncolytic virotherapy and opens a great avenue for future medical practice.

토론

간세포암종은 매우 불량한 예후를 보이며,

일반적으로 기존 화학요법 및 방사선 치료에 내성을 가집니다.

간세포암종(HCC)에서

종양용해성 바이러스 치료법이 연구되어 왔습니다.

종양용해성 바이러스 치료법에 대한 관심이 증가하고 있음에도 불구하고,

OV의 근본적인 생물학적 및 약리학적 메커니즘은 완전히 이해되지 않았습니다.

암 환자에서 OV의 치료 잠재력을 완전히 활용하기 위해서는

이러한 문제들에 대한 더 나은 이해가 필요합니다.

본 연구는

미생물군집의 존재가

종양용해성 바이러스 치료에 영향을 미치는

주요 요인일 수 있음을 입증한다.

미생물군집이 VSVΔ51 매개 종양용해성 바이러스 치료의 효능에 미치는 영향

연구에 따르면 미생물 신호는

암 발생에 필수적이며,

이는 항생제 치료로 조혈계 악성종양이 억제되고

무균 마우스에서는 발생하지 않는다는 관찰 결과와 유사하다48.

또 다른 유사한 관찰 결과는 무균 상태나 항생제 치료가

생쥐 모델에서 폐암 발병을 현저히 저해한다는 점이다49.

항생제 칵테일을 사용하여 미생물군집 구성을 제거하면

VSVΔ51 종양용해 바이러스 치료의 효능을 증강시킬 수 있으며,

이는 미생물군집의 존재가 치료의 장애물이 될 수 있음을 시사한다.

공생 세균(미생물군집)의 존재가

골수계 세포에서 Myd88 의존적 IL-1β 및 IL-23 생산을 자극하여 T 세포를 활성화시키고,

이로 인해 IL-17/기타 효과기 분자를 생성하며 염증과 종양 세포 성장을 촉진할 수 있다고 추측할 수 있다49.

따라서

VSVΔ51의 종양용해 바이러스 치료법이

미생물군집의 존재로 인해 전임상적 효과에 장벽을 마주한다는 점이 분명해진다.

따라서

종양이 내성을 획득하여 치료 실패로 이어질 수 있습니다.

실제로 특정 균주가 VSVΔ51의 항종양 효과를 약화시킨다는 점을 확인했습니다.

VSVΔ51에 의한 완전한 종양 퇴행 유도 한계는

VSVΔ51 종양용해 바이러스 치료에 내성을 보이는 종양을 표적하는 메커니즘과 새로운 접근법 탐구를 촉발시켰습니다.

마우스 모델에서 고무적인 결과를 관찰했지만, 인간 미생물군집의 구성이 마우스 미생물군집과 상대적으로 다르다는 점을 유의해야 한다. 장내 미생물군집은 간 항상성 및 간세포암(HCC) 진행에 중요한 역할을 한다50,51.

간세포암 환자와 다른 암 환자에서 종양용해성 바이러스 치료 후

미생물군집 구성이 어떻게 변화하는지는 추가 연구가 필요하다.

또한 향후 연구에서 우리의 관찰 결과를 확인하기 위해서는 환자 관련 모델도 필요하다.

이러한 우려가 존재함에도 불구하고,

본 연구는 간세포암 치료에서 온콜리틱 바이러스(OV) 적용을 개선하기 위해 

L. acidophilus로 보완하는 명확한 근거를 제시함으로써 잠재적 임상적 함의가 주목할 만하다. 

L. acidophilus 보충은

VSVΔ51 유발 미생물군집 불균형 후

장내 미생물군집 균형을 회복시킬 뿐만 아니라

장 장벽 기능을 개선하고 숙주 면역 반응을 조절하여,

종합적으로 온콜리틱 바이러스의 복제 및 효능을 증진시킨다.

종합하면,

우리의 결과는 OV의 알려지지 않은 특징을 밝혀냈습니다:

미생물군집 변화는 종양용해 바이러스 치료에서 중요한 역할을 합니다.

L. acidophilus는

VSVΔ51에 의해 손상된 장벽 무결성을 개선한다

L. acidophilus는

우수한 안전성과 효능으로 인해 가장 널리 사용되고 연구된

프로바이오틱스 중 하나이다52. 

L. acidophilus는

열안정성과 넓은 pH 범위에서의 활성 유지라는 중요한 기술적 특성을 나타내며,

식품 부패 및 병원성 세균에 대한 강력한 억제 작용으로 인해

중요한 생물학적 보존제 계열을 이룬다53.

많은 연구에서 

L. acidophilus가 장 내 pH를 낮추고 대사 산물을 생성함으로써

장내 미생물 군집의 균형을 조절할 수 있음을 보여줍니다33,34.

치료 중 VSVΔ51은 장 장벽의 밀착성(Occludin 및 E-cadherin 손실)을 교란시켜 독소/유해한 세균 생성물이 체내로 침투할 가능성을 높이며, 이는 더 많은 전신 감염과 염증을 유발하여 종양 용해 효과를 저해합니다. VSVΔ51에 의한 TNF-α 증가가 이러한 장 장벽 기능 장애를 유발하는지는 명확하지 않다. 그러나 L. acidophilus와 같은 소수의 VSVΔ51 감쇠 균주는 VSVΔ51 종양용해 바이러스 치료 효과를 향상시킬 수 있다. L. acidophilus의 장 장벽 조절이 장 장벽 결손 예방으로 이어질 수 있다는 점은 알려져 있으나, 그 기전은 명확히 규명되지 않았다54. 본 연구에서 L. acidophilus가 실제로 Occludin 및 E-cadherin과 같은 밀착 연결 단백질 발현을 회복시켜 장벽 무결성 장애를 개선하며, 동시에 VSVΔ51 종양용해 바이러스 치료의 높은 효능을 유지함을 확인했습니다.

L. acidophilus 부착은 VSVΔ51의 효능을 증진시킬 수 있습니다

기전적으로, 

L. acidophilus는 장 조직의 SLC20A1/PiT1에 결합하여 부착한다.

우리는

VSVΔ51에 의한 IL-6-STAT3 신호전달 경로의 차단을 통해 SLC20A1 발현이 하향 조절되며,

이로 인해 L. acidophilus의 장 조직 부착이 감소함을 보여주었다.

특히, 

L. acidophilus와 VSVΔ51의 병용은

치료 결과를 개선한다. 

L. acidophilus는 장벽 무결성을 개선하고,

종양 미세환경에서 세포독성 CD8+ T 세포를 증가시키며, 소진된 CD8+ T 세포를 감소시킴으로써

간암에서의 종양용해 바이러스 치료법 강화에 대한 전임상적 원리 입증(proof-of-principle)을 제공한다.

SLC20A1/PiT1 단백질은 세포막을 가로질러

인산염을 수송하는 Na(+)-의존성 공동수송체로, 인산염 항상성 유지에 관여한다.

SLC20A1/PiT1이 제거된 세포는

TNF-α의 세포사멸 촉진 활성에 더 민감한 것으로 나타났습니다55.

따라서

VSVΔ51에 의한 SLC20A1/PiT1의 하향 조절은

TNF-α 매개 세포 사멸을 촉진할 수도 있습니다.

SLC20A1 수송체는 

L. acidophilus의 수용체로 특징지어지지만, 

L. acidophilus가 SLC20A1 수송체를 통해 TNF-α와 같은 신호 전달 경로를 활성화하여

VSVΔ51의 효능을 증강시킬 수 있는지 여부는 아직 밝혀지지 않았다.

장 상피 세포 부위에 대한 L. acidophilus의 부착은

VSVΔ51에 의해 증가된 다른 공생균들에 의해 영향을 받을 수도 있으며,

이들 공생균들은 접촉 의존적 억제,

제6형 분비 시스템(T6SS) 또는 박테리오신과 같은 분비 분자56,57,58,59를 통해 

L. acidophilus를 직접적으로 억제할 수도 있습니다.

그러나 이러한 문제들은 추가 연구가 필요합니다. SLC20A1과 상호작용하는 L. acidophilus CdpA 단백질의 경우, 부착 이외의 다른 기능이 있는지 여부는 명확하지 않습니다. 주기적 디-GMP 포스포디에스테라아제 활성을 이용해 숙주 기능을 조절하는지 여부는 추가 연구가 필요합니다. 예를 들어, Burkholderia pseudomallei CdpA는 자가 응집, 편모 합성, 운동성, 생물막 형성, 세포 침입 및 세포 독성에 관여하는 것으로 알려져 있습니다60.

종합하면,

이러한 연구들은 종양 미세환경을 재프로그래밍하여

항종양 면역 반응을 촉진함으로써

종양 용해 바이러스 치료의 효능을 향상시키고,

향후 의료 실천을 위한 탁월한 길을 열어주는 미생물군 조절제 조합 활용의 잠재적 개발을 가능케 합니다.

Methods

Mouse modelsMouse experiments

Germ-free (GF) C57BL/6 J mice were bred and housed at the Shenzhen Gnotobio Biotechnology Co., Ltd. GF status was confirmed through 16S qPCR analysis before used for relative experiments, which were also carried out at Shenzhen Gnotobio Bio-technology Co., Ltd.

Conventionally housed 6-week-old C57BL/6 J wild-type mice were purchased from Guangdong GemPharmatech Co., Ltd, GUANGDONG MEDICAL LABORATORY ANIMAL CENTER and Sun Yat-sen University. Mice were fed with normal diet (PicoLab 5053, Purina) and maintained on a 12 h light/dark cycle (6 a.m–6 p.m.). Studies involving animals were approved by the Animal Ethical and Welfare Committee of Sun Yat-sen University (SYSU-IACUC-2023-002133) and Shenzhen Gnotobio Biotechnology (JTAW20240806-1, JTAW20241106-1 and JTAW20240605-1) and were carried out in accordance with the approved protocols.

Measurement of fecal output, water content, whole gut transit time and small intestinal transit

The fecal output was recorded for 3 h to detect the gut motility of mice. Each mouse’s average number of fecal pellets per hour was measured for inter-individual comparisons. Fresh fecal pellets from each mouse were collected in a separate sterile EP tube. After calculating the wet weight, each sample was subject to a freeze dryer for 24 h to get the dry weight. The water content was calculated as the difference between the wet and dry weight of the representative fecal pellet. For whole gut transit time, mice were fasted overnight and were given an oral gavage of 250 μL of black ink. The mice were immediately transferred to a new, clean, empty individual cage. The whole gut transit or defecation time was calculated as the time required to defecate the first black stool pallet. For small intestinal transit, gastrointestinal motility was measured according to a previous report61. Briefly, mice were fasted overnight and were given an oral gavage of 250 μL of black ink. After 10 min, the abdomen of each mouse was opened, and the entire small intestine (from the pylorus to the cecum) was carefully taken out. The distance moved by ink and the total length of the small intestine were measured.

HCC mouse models

Hydrodynamic tail vein injection (HDTI) was performed based on the previous report62. HA-myr-AKT, N90-β-catenin, Sleeping Beauty (SB) and sgClock/Vector were injected in 2 mL PBS within 7 s. For orthotropic mice model, H22 cells (50 μL, ~1 × 106 cells) were collected in PBS and intra-liver injected into 6-week-old female C57BL/6 J mice. The organs or tissues of each mouse were dissected and then fixed in 4% formaldehyde and embedded in paraffin. For subcutaneous mice model, H22 and Hepa 1–6 (1 × 106 cells per mouse) cells were inoculated subcutaneously into the hind-flanks of 6-week-old female C57BL/6 J mice. Tumor length and width were measured, and the volume was calculated according to the formula (length × width2)/2. According to the Animal Ethics and Welfare Committee, the maximum tumor load allowed by the ethics committee was 1500 mm3. The humane end point was when the tumor volume reached 1500 mm3. For subcutaneous mice model, VSVΔ51 were intravenously administrated (1 × 107 PFUs) 3 times after L. acidophilus or L. acidophilus (per-heated) 3 times (2 × 108 PFUs, i.g., total 5 times). HSV-1-anti-PD1 and VACV(WR)-GFP administered by intratumoral injection (1 × 107 PFUs).

T/C% defined as the relative tumor proliferation rate, and T/C% values of <40% are considered to indicate an effective response. The T/C% was calculated as follows: T/C% = TRTV /CRTV 100%; Vt, tumor volume after treatment; V0, tumor volume before treatment; RTV, relative tumor volume, RTV = Vt /V0; TRTV, RTV in the treatment group; CRTV, RTV in the control group. At the end of experiments, tumor mass was calculated and tumor tissues were dissected and fixed in 4% formaldehyde.

Fecal microbiota transplantation (FMT)

Orthotropic mice model with H22 cells were established. Feces from mice treated with VSVΔ51 (tail vein injection) or not were collected and resuspended in PBS at a concentration of 0.125 g mL-1. Mice were pretreated with antibiotics (ABX (500 μL each mouse), containing vancomycin (50 mg mL-1; V105495, Aladdin, China), imipenem/cilastatin (25 mg mL-1; C304719, I304258, Aladdin, China), neomycin (10 mg mL-1; N412785, Aladdin, China), and amphotericin (1 mg mL-1; A105482, Aladdin, China)) for 5 times. After administration of 0.15 mL fecal suspension from Control or VSVΔ51-treated mice for 3 days, the microbiota-recipient mice were subjected to 1 × 107 PFUs/mouse VSVΔ51 treatment. At the end of experiments, tumor areas were calculated, and tumor tissues were dissected and fixed in 4% formaldehyde.

Cell culture, viruses and transfection

HEK-293T, NCM460, Hepa1-6, HuH7, Vero and BHK-21 cells were cultured at 37 °C in Dulbecco’s Modified Eagle Medium (DMEM)(10-013-CV, Corning Cellgro, USA) supplemented with 10% fetal bovine serum (FBS)(35-010-CV, Corning Cellgro, USA). H22 cells were cultured at 37 °C in Roswell Park Memorial Institute 1640 (RPMI 1640) (10-040-CV, Corning Cellgro, USA) supplemented with 10% FBS.

The cell lines have been authenticated by the short tandem repeat (STR) assay and were confirmed to be without mycoplasma contamination. VSVΔ51-GFP was provided by Prof. Haipeng Zhang (School of Medicine, Jinan University). HSV-1-anti-PD1 was provided by Prof. Minfeng Shu (School of Basic Medical Sciences, Fudan University). VACV(WR)-GFP was purchased from VectorBuilder company. VSVΔ51 was propagated in BHK-21 cells, VACV-GFP and HSV-1 were propagated in Vero cells. Virus titer was determined by TCID50 assay using BHK-21 cells and converted to PFU.

All transient transfections of plasmids and shRNA into cell lines followed the standard protocol for Lipofectamine 3000 Transfection Reagent (L3000001, Thermo Fisher, USA). Cell viability was measured by 2-(2-methoxy-4-nitrobenzene)-3-(4-nitrobenzene)-5-(2, 4-disulfonyl benzene)-2h-tetrazole monosodium salt bromide (CCK8) (K1018, APExBIO, USA) assay after 48 h post-infection.

Bacteria and culture conditions

Clostridium aldenense (DSM 19262), Clostridium bolteae (ATCC BAA-613), Clostridium lavalense (DSM 19851), Clostridium scindens (CCUG 45363), Clostridium polysaccharolyticum (DSM 1801), Barnesiella viscericola (DSM 18177), Barnesiella intestinihominis (DSM 21032), Blautia hansenii (DSM 20583), Blautia wexlerae (DSM 19850), Desulfovibrio oxamicus (DSM 1925), Desulfovibrio simplex (DSM 4141), Desulfovibrio piger (DSM 749), Rikenella microfusus (DSM 15922), Acetatifactor muris (DSM 23669), and Lactobacillus apodemi (DSM 16748) were purchased from Ning bo testobio Co., Ltd. Klebsiella oxytoca (GDMCC 1.134), Alistipes shahii (DSM 19121 = ATCC BAA-1179 = CCUG 48947), Alistipes onderdonkii subsp. Onderdonkii (DSM 19147 = ATCC BAA-1178 = CCUG 48946), Alistipes finegoldii (DSM 17242), Alistipes indistinctus (DSM 22520), Ligilactobacillus animalis (CGMCC 1.1852), Lacticaseibacillus paracasei (ATCC 334), Lactobacillus fermentum (ATCC 11739), Clostridium methylpentosum (ATCC 43829), Lactobacillus salivarius (ATCC 11741), Lactobacillus acidophilus (CICC 20244) and Akkermansia muciniphila (ATCC BAA-835) were purchased from Guangdong Microbial Culture Collection Center (GDMCC).

Akkermansia muciniphila and Clostridium lavalense cultured in Brain Heart Infusion (BHI) broth (MZM0090, MINGZHOUBIO, China) at 37 °C under aerobic conditions for 48 h before use. Desulfovibrio piger, Desulfovibrio oxamicus, and Desulfovibrio simplex were cultured in Desulfovibrio Medium Base (MZM2256, MINGZHOUBIO, China) at 37 °C under aerobic conditions for 48 h before use. Klebsiella oxytoca cultured in Nutrient Agar broth (HB8274-1, Hopebio, China) at 37 °C under aerobic conditions for 48 h before use. Rikenella microfusus cultured in Columbia Blood Agar Base broth (CP0160, HuanKai Microbial, China) at 37 °C under aerobic conditions for 48 h before use. Alistipes shahii, Alistipes finegoldii, Alistipes indistinctus and Alistipes onderdonkii subsp. Onderdonkii cultured, Clostridium aldenense, Clostridium scindens, Barnesiella viscericola, and Barnesiella intestinihominis in Chopped Meat Broths (MZM5638, MINGZHOUBIO, China) at 37 °C under aerobic conditions for 48 h before use. Clostridium bolteae in Gifu Anaerobic Medium broth (TYSW33987, Ning bo testobio Co., Ltd, China) at 37 °C under aerobic conditions for 48 h before use. Clostridium lavalense and Clostridium polysaccharolyticum, Clostridium methylpentosum and cultured in Chopped Meat Carbohydrate Broth (CMC) (KDM150, MINGZHOUBIO, China) at 37 °C under anaerobic conditions. Blautia hansenii, Blautia wexlerae, and Acetatifactor muris were cultured in PYG MEDIUM (modified) (KDM144, MINGZHOUBIO, China) at 37 °C under anaerobic conditions. Lacticaseibacillus paracasei, Lactobacillus apodemi, Ligilactobacillus animalis, Lactobacillus fermentum, and Lactobacillus salivarius, Lactobacillus acidophilus cultured in MRS Medium (27315, HuaiKai Microbial, China) at 37 °C under anaerobic conditions, respectively. Chopped Meat Broths, Gifu Anaerobic Medium broth and Chopped Meat Carbohydrate Broth The configuration of a required an additional 20% minced meat pellet, Vitamin K1 Solution (500 mg mL-1) and Heme chloride solution (50 mg mL-1).

Cell-bacteria co-culture

As we reported previously, the bacteria were collected for the co-culture of epithelial cells with bacteria. HuH7 cell were seeded on 48-well plates at 1 × 105 cells per well for 24 h and treated with bacteria at different MOI (100 MOI, 200 MOI) for 2 h. After 2 h, the supernatant was aspirated. Cells were washed 5 times using sterile PBS. Fresh antibiotics (1% penicillin/streptomycin, 15140-122, GIBCO, USA) containing medium. VSVΔ51 (0.01 MOI) was used to infect cells in DMEM supplemented with 10% FBS containing serum for 2 h. Afterward, the supernatant was removed and replaced with fresh medium. Cell viability was assessed 48 h later. Cell viability was measured by CCK8 (K1018, APExBIO, USA) assay.

Antibiotics treatment

To ablate the gut microbiome and/or the tumor microbiome, 6-week-old mice were administered an antibiotic cocktail (ABX) as described31. Briefly, mice started to be administered with high dose of antibiotics (300 μL each mouse), containing vancomycin (50 mg mL-1; V105495, Aladdin, China), imipenem/cilastatin (25 mg mL-1; C304719, I304258, Aladdin, China), neomycin (10 mg mL-1; N412785, Aladdin, China), and amphotericin (1 mg mL-1; A105482, Aladdin, China), by oral gavage daily for 3 consecutive days. Then the mice were treated with a low-dose ABX containing vancomycin (0.5 mg mL-1), imipenem/cilastatin (0.5 mg mL-1), neomycin (1 mg mL-1), and amphotericin (0.5 mg mL-1), by the sterile drinking water until endpoints. Solutions and bottles were changed every other day due to the short half-life of imipenem. To specifically eliminate the intratumoral microbiota, we chose an intravenous injection of 300 μL of the ABX suspension daily (vancomycin 10 mg mL-1; imipenem/cilastatin 4 mg mL-1; Neomycin 1.5 mg mL-1).

Feces 16S rRNA sequencing and data analysisDNA extraction and PCR amplification

Briefly, DNA extraction was performed using BGI-NGS-TQ-DNA method (Meta DNA protocol). DNA library preparation and 16S rRNA gene sequencing were performed by The Beijing Genomics Institute (BGI, China) and CHI BIOTECH CO., LTD. The V3-V4 regions of 16S rRNA genes were amplified using specific primer (341F-‘CCTAYGGGRBGCASCAG’and 806R-‘GGACTACNNGGGTATCTAAT’) together with the barcode.

Illumina sequencing

Purified amplicons were pooled in equimolar amounts and paired-end sequenced on an MGISEQ-2000 platform (BGI, Beijing, China) according to the standard protocols by BGI.

Amplicon sequence processing and contamination DNA filtration

The 16S rRNA sequencing data were quality-filtered and analyzed using QIIME2 (version 2019.4.0; default parameters) software. The sequencing errors and replicated sequences were determined by Deblur algorithm with default parameters. Raw data are filtered, and high-quality clean reads are generated. After chimera sequences filtering, the dereplicated sequences were classified taxonomically through Ribosomal Database Project (RDP) 16S rRNA gene reference database at a 99% identity cut-off by USEARCH (v7.0.1090) software. OTU representative sequences are aligned by RDP classifier (v2.2) software (sequence identity is set to 0.6).

Data analysis

Alpha diversity indices including Chao1, Shannon index, principal coordinate analysis (PCoA) based on Bray-curtis dissimilarity and the PERMANOVA test were all calculated by Vegan v2.6-4 package. Proportional Chord Diagram analysis was all calculated by “circlize” package. The linear discriminant analysis (LDA) effect size (LEfSe) (http://huttenhower.sph.harvard.edu/LEfSe) was all calculated by “microeco” package (LDA score > 2, P < 0.05). Stacked bar plot analysis was all calculated by “MicrobiotaProcess” package. Mantel test analysis was calculated by “linkET” package.

Tissues 5 R (5 regions) 16S rRNA sequencing and data analysisDNA extraction and PCR amplification

Liver tumor samples were collected, and total microbial genomic DNA was extracted using the FastPure Stool DNA Isolation Kit (Magnetic bead) (MJYH, shanghai, China) according to manufacturer’s instructions. The quality and concentration of DNA were determined by 1.0% agarose gel electrophoresis and a NanoDrop® ND-2000 spectrophotometer (Thermo Scientific, USA).

16S rRNA amplification and sequencings were done by amplifying 5 regions on the 16S rRNA gene in multiplex by T100 Thermal Cycler (BIO-RAD, USA). The negative controls contain sampling negative controls, DNA extraction controls and no-template PCR amplification controls.

Illumina sequencing

Purified amplicons were pooled in equimolar amounts and paired-end sequenced on an Illumina NextSeq 2000 platform (Illumina, San Diego, USA) according to the standard protocols by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China).

Amplicon sequence processing and contamination DNA Filtration

Reads were demultiplexed per sample, filtered and aligned to each of the five amplified regions based on the primers’ sequences. To combine read counts from the five regions into a coherent profiling result, the Short MUlitiple Regions Framework (SMURF) method was applied. The GreenGenes database (May 2013 version, with some improvements) was used as reference29,63. To reduce the noise variation, samples with <1000 normalized reads (including negative controls) and species with relative abundances of <10–4 were discounted from further analysis. Any species that was preval‎ent across >7.5% of negative DNA/PCR/Sequencing controls or >7.5% of empty paraffin controls was completely removed64.

Data analysis

Same as Feces 16S rRNA sequencing data analysis.

Measurement of viral mRNA levels and titers

VSVΔ51-infected mice were euthanized 3 days post infection. Organs or tissues such as heart, liver, spleen, lung, kidney, brain, anker and tumor were harvested, weighed, and homogenized. The remaining cells were resuspended in TRIzol (15596018, ThermoFisher, USA), and RNA was extracted according to the manufacturer’s recommendation. VSVΔ51 RNA levels were determined by one-step quantitative reverse transcriptase PCR (qRT–PCR) using an RT-PCR kit (A1250, PROMEGA, USA) on a Roche Instrument using standard cycling conditions. For some samples, the viral titers were determined by TCID50 assay on BHK-21 cells as described65.

Phylogenetic analysis for bacteria

The 16S rRNA sequences of Bacteria were downloaded from NCBI and used as a reference genome. To construct a phylogenetic tree through identified strains, the Basic Local Alignment Search Tool (BLAST) was used to analyze the obtained sequence with organisms in the GeneBank database. The Neighbor Joining algorithm in MEGAX software was used to construct the phylogenetic tree and molecular evolutionary analyses. The percentage of replicate trees where the associated taxa clustered together in the bootstrap test (1000 replicates) was shown next to the branches. The Maximum Composite Likelihood method was used to compute the evolutionary distances66. The bacterial 16S rRNA sequences used are listed in Supplementary Data 2.

Flow cytometry

Tumors samples were harvested after perfusion with PBS. Tumor samples were obtained, subjected to single cell suspension and passed through a 70 mm filter. Erythrocytes were lysed through Red Blood Cell Lysis Buffer (R874904, MACKLIN, USA), and the remaining cells were resuspended in PBS supplemented with 2% fetal bovine serum (35-010-CV, Corning, USA) and 1 mM EDTA (E762089, MACKLIN, USA) and maintained on ice. Fluorochrome-conjugated anti-mouse antibodies from BioLgend were used. Utilizing cell stimulation cocktail (1:500, 00-4975-93, ThermoFisher, USA) induction followed by intracellular detection of cytokines and secreted proteins in ex vivo cells. Dead cells were stained with Zombie Red™ Fixable Viability kit (423110, BioLegend, USA). The staining for cell surface antigens was performed. Cells were fixed using True-Nuclear™ Transcription Factor Buffer Set (424401, BioLegend, USA) and Cyto-Fast™ Fix/Perm Buffer Set (BioLegend, 426803, USA) processed for flow cytometry assay (Beckman Coulter). Data were acquired with CytoFLEX and analyzed using FlowJo. The gate strategy was set based on isotype control. Technicians acquiring and gating the data were blinded to the treatments. Schematic gating strategy is provided in Supplementary Fig. 8.

Western blot

Cells or tissues were lysed using NP-40 buffer36. The protein in SDS-PAGE (10%-12%) was then transferred onto polyvinylidene difluoride (0.45 µm, PVDF) membranes (ZY101123, Millipore, USA) for about 260 V 2.5 h, which was then blocked with 5% non-fat milk in 0.05% Tris-based saline-Tween 20 for 1 h at room temperature. The membranes were incubated with primary antibodies overnight at 4 °C and then with secondary antibody at room temperature for 1 hr. The antibodies used in this study are shown in the essential resources Supplementary Data 1. For immunoprecipitation, cell lysates were incubated with indicated antibodies for 3 h at 4 °C.

Antibodies and reagents

Antibodies used in this study are listed as follows: GAPDH (60004-1-Ig, Proteintech, 1:10000), HA Tag (51064-2-AP, Proteintech, 1:4000), Myc-Tag (9B11) (2276 s, Cell Signaling Technology, 1:2000), FLAG M2 Sigma F1804 1:2000), Cleaved Caspase-3 (Asp175) (5A1E) (9664 P,Cell Signaling Technology, 1:200 (IHC)), Ki67 (34330, Cell Signaling Technology, 1:800 (IHC)), SLC20A1 (12423-1-AP, Proteintech, 1:1000 (WB) 1:200 (IF)), Occludin (502601, ZEN-BIOSCIENCE, 1:1000), E-Cadherin (3195 s, Cell Signaling Technology, 1:1000), Stat3 (D3ZZ2G) (12640, Cell Signaling Technology, 1:1000), Phospho-Stat3 (Tyr705) (D3A7) (9145, Cell Signaling Technology, 1:1000), VSV-G[8G5F11] (EB0010, Kerafast,1:1000 (WB) 1:100 (IHC)), Stat1 (D1K9Y) (14994, Cell Signaling Technology, 1:1000), Phospho-Stat1 (Tyr701) (58D6) (9167, Cell Signaling Technology, 1:1000), Goat anti-Rabbit IgG (H + L) Secondary Antibody HRP (31460, Thermo Fisher, 1:10000), Goat anti-Mouse IgG (H + L) Secondary Antibody HRP (31430, Thermo Fisher, 1:10000), Ultra Streptavidin-HRP (RG235220, Thermo Fisher Scientific, 1:10,000), CD45.2 APC/Fire 750 (109852, BioLegend, USA, 1:200), CD3 FITC (100204, BioLegend, 1:200), CD4 Brilliant Violet 421 (100438, BioLegend, 1:200), CD8a Brilliant Violet 510 (100752, BioLegend, 1:200), IFN-γ Brilliant Violet 605 (505804, BioLegend, 1:200), TNF-α APC (506308, BioLegend, 1:200), Granzyme B PE (372208, BioLegend, 1:200), FOXP3 PF (126404, BioLegend, 1:200), CD25 PE/Cyanine7 (102016, BioLegend, 1:200), CD11b APC (101211, BioLegend, 1:200) and F4/80 PE/Cyanine7 (123114, BioLegend, 1:200)

Immunoprecipitation and pull-down assays

For FLAG/Myc-tagged protein immunoprecipitation, transfected cells were collected and lysed in NP-40 buffer supplemented with protease and phosphatase inhibitor cocktail (B15002, Bimake, USA). Cell lysates were incubated with indicated antibodies (M2, A222, Sigma-Aldrich /Myc beads, B26302, Biotool, USA) for 3 h at 4 °C. The beads were washed and boiled with 50 μL loading buffer and were subjected to SDS-PAGE.

Reverse transcription-quantitative PCR (RT-qPCR)

Total RNA was extracted using TRIzol (15596018, ThermoFisher, USA), and reverse transcription was performed from 2 μg total RNA and reverse transcribed using RevertAid Reverse Transcriptase (22948, PROMEGA, USA) according to the manufacturer’s recommendation. qRT-PCR was performed with SuperReal PreMix SYBR Green (A4004S, BIMAKE, USA) using a Roche480 Real-Time PCR Detection System (Roche480, Roche, USA). All genes were normalized to β-ACTIN/β-Actin. The primer sequences used are listed in Supplementary Table 1.

RNA sequencing and data processing

Total RNA was extracted from the cells using Trizol (ThermoFisher) according to the manual and sent to Pangu Medical Technology Co., Ltd (Guangdong, China) for further processing and RNA-seq analysis. In brief, total RNA was qualified and quantified using a NanoDrop and Agilent 2100 bioanalyzer (Thermo Fisher Scientific, USA), and qualified library sequenced using illumina Novaseq 6000. The raw reads with adapters and low-quality bases were filtered by fastp67. Fragments Per Kilobase of transcript per Million mapped reads (FPKM) of each gene was calculated based on the length of the gene and the number of reads mapped to this gene.

Differential gene expression‎ and enrichment analysis

Differential gene expression‎ between the two groups of different treatments were analyzed using the DESeq2 R package68. The P-values were adjusted using the Benjamini&Hochberg method. A corrected P-value < 0.05 (FDR = 5%) and absolute fold change ≥ 2 was set as the threshold for significantly different expression‎ genes (DEGs). Cluster Profiler tool was used to perform gene enrichment analysis. GSEA enrichment scores were calculated using the R package GSVA69.

Lentivirus production and infection

All of the plasmids were purchased from Addgene. We screened ~4 hairpin shRNAs targeting specific transcript and found at least two independent sequences that reduced mRNA levels by >70%. These shRNAs were in the pLKO.1 vector. Lentiviruses were produced in HEK-293T cells using 10 μg shRNA plasmid, 5 μg psPAX2 plasmid and 5 μg pMD2.G. The supernatants containing viral particles were harvested at 48 and 72 h after transfection and were filtered through Millex-GP Filter Unit (0.45 μm pore size, Millipore). Cells were introduced with lentivirus, 200 μL FBS, 5 µg mL-1 polybrene (TR-1003-G, Millipore, USA) for 48 h. Cells were selected with 2 µg mL-1 puromycin for 3–4 days to increase the KD efficiency. After selection, 200 ng mL-1 doxycycline (Dox) was added to induce KD of the target gene.

DNA construction and mutagenesis

PCR-amplified SLC20A1, Cdpa, LBA1838, SlpA, LBA1838 and LBA1839 were cloned into pcDNA3.1. All the mutations were generated by Mut Express II Fast Mutagenesis Kit (C214-01/02, Vazyme, China) according to the manufacturer’s instructions. All constructions with the desired insertions or mutations were confirmed by sequencing.

Immunofluorescence (IF) and Fluorescence in Situ Hybridization (FISH) assay

Immunofluorescence staining was performed on paraffin-embedded tissue sections. Tissue sections were incubated with primary antibodies at a dilution of 1:100, fluorescence dye-conjugated secondary antibodies and DAPI according to standard protocols. Cy5.5-labeled L. acidophilus probe (sequence: 5’-TCTTTCGATGCATCCACA-3’) was used to detect L. acidophilus colonization in paraffin-embedded intestine sections. FITC-labeled 16S EUB338 probe (sequence: 5’-GCTGCCTCCCGTAGGAGT-3’) was used to detect bacteria colonization in paraffin-embedded intestine sections. FITC-labeled Live/Dead Bacterial Double Stain Kit (share-bio, SB-L6060L, USA) was used to detect bacteria adhesion in NCM460. Images were acquired with a confocal microscope (Leica).

Transmission electron microscopy (TEM)

NCM460 cells with or without SLC20A1 KD were seeded in sterile coverslips and co-cultured with L. acidophilus (MOI = 100) for 2 h. Cell samples were collected with centrifugation. Colonic tissues were dissected into small pieces. All the samples were fixed in 2.0% glutaraldehyde in 0.1 M sodium cacodylate (G1102-100ML, Electron Microscopy Sciences, Servicebio, China). Samples were then submitted to the Electron Microscopy Facility (Servicebio) for standard transmission electron microscopy ultrastructural analysis.

Scanning electron microscopy (SEM)

NCM460 cells with or without SLC20A1 KD were seeded in sterile coverslips and co-cultured with L. acidophilus (MOI = 100) for 2 h. Cells were then washed with freshly prepared Sorensen’ phosphate buffer (0.1 M, pH 7.2) three times and were fixed with 2.5% glutaraldehyde fixative overnight. After dehydration and coating with gold-palladium, specimens were examined by a Cold Field Scanning Electron Microscope (Hitachi SU8100, Guang Tenng Technology CO.,LTD).

Chromatin immunoprecipitation assay (ChIP)

NCM460 cells pre-treated with or without IL-6 were crosslinked with 1% formaldehyde for 10 min at room temperature and quenched by glycine. After cell lysis, the chromatin was fragmented into 100–500 bp by Bioruptor Sonicator (Diagenode) and protein-DNA complexes were immunoprecipitated (IP) by 5 mg mouse monoclonal antibody against human Phospho-STAT3 (Tyr705) (9145, CST, USA), RNA Polymerase II, clone CTD448 or anti-IgG antibody (17-371, Millipore, USA). Dynal magnetic bead was added and mixed on rotator at 4 °C overnight. Contaminating RNA was removed by treating with 10 μg mL-1 RNase A. After washing and reversal of crosslinks, the IP and input DNA was amplified by real-time PCR with primers near the putative IL-6-binding site in the SLC20A1 promoter.

Cytokines array

Mice sera were collected and send to the Shanghai Universal Biotech Co., Ltd company for detection of 23 cytokine group I factors using the Bio-Plex 200 assay kit.

Hematoxylin-eosin staining (H&E) and Immunohistochemistry (IHC) assay

Tissue pathological observations, cell death and proliferation markers were measured by H&E staining and IHC staining, respectively. The expression‎ of Ki-67, cleaved-Caspase-3 (34330, 9664 P, Cell Signaling Technology, USA) and VSV-G (EB0010, Kerafast, USA) in tumors was assessed by IHC. Briefly, for H&E staining, tissues were immobilized with 10% neutral-buffered formalin and embedded in paraffin. The sections (5 mm) were deparaffinized with xylene and rehydrated through washing with graded ethanol. For IHC staining, the sections were dewaxed in xylene, hydrated in descending concentrations of ethanol, immersed in 0.3% H2O2-methanol for 30 min. The sections were further washed with phosphate-buffered saline and probed with indicated antibodies or isotype control at 4 °C overnight. After washing, biotinylated goat anti-rabbit or anti-mouse IgG were incubated at room temperature for 2 h. Immunostaining was visualized with streptavidin/peroxidase complex and diaminobenzidine, and sections were then counterstained with hematoxylin. All slides were observed and captured under a microscope (SQS-1000, TEKQSRAY, China) and at least three fields for each slide were randomly selected for qualification by ImageJ software.

L. acidophilus attachment assay

NCM460 cells were co-cultured with L. acidophilus (MOI = 200) for 2 h under aerobic conditions. After co-culture, the cells were washed 5 times with PBS and subsequently lysed for 5 min with 500 μL ice-cold PBS with Tween 80 (0.01%). The samples were then diluted 10-, 100-, and 1000-fold and plated onto a BHI agar plate. After overnight incubation at 37 °C, the number of colony formation units (CFU) was counted.

L. acidophilus surface proteins extraction

To prepare the L. acidophilus surface proteins, L. acidophilus was first cultured at 37 °C under aerobic conditions for 24 h. The bacteria cells were then collected by centrifugation at 2000 g for 15 min at 4 °C. The bacteria pallets were washed 3 times by PBS and the wet pallet was weighted. Approximately 3 mL lysis buffer (50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 50 mM NaCl, 1 mM phenylmethylsulfonylfluoride) with lysozyme (300 μg mL-1) for each wet gram of L. acidophilus pallet was added to resuspend the pallet. The whole suspension was stirred for 30 min at 4 °C, which was then treated with 1% triton x-100 (v/v) and applied for ultrasound sonication. The sonicated L. acidophilus suspension was further incubated at room temperature with DNase1 (10 mg kg-1) and MgCl2 (10 mM) for 15 min to remove the viscous nucleic acid. The suspension was centrifuged at 10,000 g for 15 min at 4 °C to remove the pallets, and the supernatant was collected and used as the surface membrane proteins for further identification of the protein of interest.

Biotin pull-down assay and protein identification

The extracted biotinylated L. acidophilus protein and NCM460 lysates were incubated with streptavidin magnetic beads (88816, Thermo Fisher Scientific, USA) for 4 h at 4 °C. The streptavidin bound L. acidophilus proteins were incubated with NCM460 proteins overnight at 4 °C, and subsequently washed with lysis buffer 5 times. The NCM460-L. acidophilus interacting proteins were eluted by heating in 50 μL of 2 × loading buffer at 100 °C. Eluted proteins were subjected to SDS-PAGE and the whole gel was then subjected to Coomassie brilliant blue G250 method staining (P0017, Beyotime, China) following the manufacturer’s instructions. The extraction, detection, and quantitative analysis of protein profiling in the samples were performed by Wuhan Metware Biotechnology Co., Ltd. (www.metware.cn). LC was performed on a nanoElute UHPLC (Bruker Daltonics, Germany). About 200 ng peptides were separated within 40 min at a flow rate of 0.3 μL min-1 on a commercially available reverse-phase C18 column with an integrated CaptiveSpray Emitter (25 cm × 75 μm ID, 1.6 μm, Aurora Series with CSI, IonOpticks, Australia). The LC was coupled online to a hybrid timsTOF Pro2 (Bruker Daltonics, Germany) via a CaptiveSpray nano-electrospray ion source (CSI). The capillary voltage for mass spectra and mass spectrum acquisition was set to 1500 V and ranged from 100 to 1700 m/z. MS raw data were analyzed using DIA-NN(v1.8.1) with library-free method the uniprot-proteome_UP000000589_M database(A total of 55319 sequences) was uesed to creat a spectra library with deep learning algrithms of neural networks.The MBR method was used to create a spectral library from the DIA data, which was then used for reanalysis. FDR of search results was adjusted to <1% at both protein and precursor ion levels, the remaining identifications were used for further quantification analysis. Mass spectrometry results show in Supplementary Data 3 and 4.

Far western assay and protein identification

For the screening of the NCM460 surface receptor that binds to L. acidophilus, the NCM460 proteins were separated by SDS-PAGE and transferred onto a PVDF membrane. The PVDF membrane was blocked with 5% BSA for 1 hr and incubated with biotinylated L. acidophilus surface proteins (final concentration 0.05 µg µl-1) overnight at 4 °C, Biotin-labeled proteins were detected using Ultra Streptavidin-HRP (RG235220, Thermo Fisher Scientific, USA) (1:10000).

Targeted metabolomics assay

To extract metabolites from the samples, 800 μL of cold Methanol/Acetonitrile/Water (2:2:1, v/v) extraction solvent was added to 100 mg sample, and adequately vortexed. For absolute quantification of the metabolites, stock solutions of stable-isotope internal standards were added to the extraction solvent simultaneously. Then the samples were under vigorous shaking for 2 min at 4 °C incubated on ice for 20 min, and then centrifuged at 14,000 g for 20 min at 4 °C, the supernatant was collected and flowed through a 96-well protein precipitation plate, and then the elution was collected and dried in a vacuum centrifuge at 4 °C. For LC-MS analysis, the samples were re-dissolved in 100 μL Acetonitrile/Water (1:1, v/v) solvent and centrifuged at 14000 g at 4°C for 15 min, then the supernatant was injected. LC-MS/MS analyses were performed using an UHPLC (1290 Infinity LC, Agilent Technologies) coupled to a QTRAP MS (6500 + , Sciex) in Shanghai Applied Protein Technology Co., Ltd. HILIC separation was then initiated at a flow rate of 300 μL min-1. RPLC separation was then initiated at a flow rate of 400 μL min-1. MultiQuant or Analyst was used for quantitative data processing. After and normalization, the data were uploaded to SIMCA-P (v14.1, Umetrics, Umea, Sweden) for multivariate data analysis: orthogonal partial least squares discriminant analysis (OPLS-DA). The stability of the model was eval‎uated using 7-fold cross validation and response permutation test. The importance (VIP) value of each variable in the OPLS-DA model was calculated to indicate its contribution to classification. Significance was determined by unpaired t-tests. P < 0.05 was considered statistically significant.

Statistics and reproducibility

Statistical significance was performed using GraphPad Prism version 9.0 (La Jolla, CA). The specific test for each dataset is indicated in respective figure legends and was selected by the number of comparison groups and variance of the data. Significance was determined by unpaired one- or two-tailed Student’s t-test, Fisher’s Exact test, Mann–Whitney U-test, or ANOVA with Tukey post-hoc test. one-tailed Fisher’s Exact test was used for enrichment analysis of GO terms and KEGG pathways. The number of independent biological replicates are provided in the figures and figure legends. No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All data presented in the manuscript and Supplementary Figures are provided in the Excel file as the Source Data file, which is linked to this manuscript. The raw sequence data reported in this study are available in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in National Genomics Data Centre (Nucleic Acids Res 2022), China National Centre for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences database under the Bioproject number PRJCA027376 (GSA: CRA017533, CRA022843, CRA017504, CRA0175502 and CRA017489) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa. The protein mass spectrometry data generated in this study have been deposited in the ProteomeXchange database under accession code PXD061574 and PXD061642 [http://www.proteomexchange.org]. The authors declare that the remaining data generated or analyzed during this study are available within the article, Supplementary Information, or Source Data file. Source data are provided with this paper.

References

1.  
2.  
3.  
4.  
5.  
6.  
7.  
8.  
9.  
10.  
11.