To validate the impact of ATO on the human gut microbiota, a comprehensive bioinformatic re-analysis was performed using publicly available data from the National Center for Biotechnology Information (NCBI) Sequence Read Archive. Raw 16S rRNA gene sequencing data of gut microbiota were obtained from BioProject ID PRJNA935705, which included 22 APL patients receiving ATO-based regimens. 20 of 22 fecal samples data were obtained and further analyzed, including pre-treatment (APL group, n = 5) and post-ATO treatment (ATO group, n = 15). Raw sequencing data were processed using the QIIME 2 pipeline (version 2023.2, https://qiime2.org/). After quality control, denoising, and chimera removal via the DADA2 plugin, amplicon sequence variants (ASVs) were generated. Taxonomic classification was performed using the SILVA database (version 138.1, https://www.arb-silva.de/).

Alpha diversity indices, including Chao1 (species richness) and Shannon (richness and evenness), were calculated to assess within-sample diversity. Group differences were evaluated using the Mann–Whitney U test. Beta diversity was visualized using principal coordinate analysis (PCoA) based on weighted UniFrac distances. Differentially abundant bacterial genera between groups were identified using R package “DESeq2” (v1.38.3, https://www.bioconductor.org/packages/release/bioc/html/DESeq2.html). Spearman’s rank correlation analysis was conducted among the top 10 most abundant genera, and a correlation heatmap was generated to explore interactions, especially involving key beneficial taxa such as Bifidobacterium.

The human APL cell line NB4 (ATCC® CRL-2047™) was cultured in RPMI-1640 (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Gibco), 1% penicillin/streptomycin (Gibco), and 1% L-glutamine (Gibco) at 37 °C in a 5% CO humidified incubator (Thermo Fisher Scientific, Waltham, MA, USA). ATO (GBW08666, purchased from the National Institute of Metrology, Beijing, China) was dissolved in sterile phosphate-buffered saline (PBS) to a final concentration of 200 µg/mL for in vivo assays. All cell lines were validated by STR profiling and tested negative for mycoplasma.

Six-week-old female C57BL/6J mice were maintained under specific pathogen-free (SPF) conditions (22 $\pm$ 1 °C; 55% humidity; 12 h light/dark cycle). To establish an APL model, the mice received 450 cGy total-body irradiation (PRECISION X-RAY SmART+, Precision X-ray Inc, Madison, CT, USA) and were intravenously transplanted with 1 $\times$ 10⁷ NB4 cells. After 20 days, successful engraftment was confirmed by peripheral blood smear analysis via tail vein sampling and defined as $\ge$20% leukemic cells by microscopy. Subsequently, the mice were treated with ATO $\pm$ BP bacteria: BP (BeNa Culture Collection, BNCC135158, Beijing, China; 1 $\times$ 10⁹ CFU/mL, 200 µL/dose orally, thrice weekly) and ATO (1 mg/kg, intraperitoneal injection, daily for 14 days) were co-administered for 14 days. Control groups included APL mice receiving PBS alone. Mice were euthanized by pentobarbital sodium = 100 mg/kg and monitored for loss of the righting reflex and corneal reflex. Once unresponsive, euthanasia was immediately performed by cervical dislocation; death was confirmed by the absence of respiration, palpable heartbeat, and pupillary light reflex. All manipulations were conducted by trained personnel using appropriate personal protective equipment (PPE; e.g., gloves, lab coat, eye protection) and aseptic technique. After confirmed death, orbital sinus blood was collected in a non-survival procedure. The animal was placed in lateral recumbency, and the medial canthus and periorbital skin were disinfected with 75% ethanol. A sterile, blunt-ended capillary tube (0.5 $\times$ 100 mm) was inserted at a 30°–45° angle relative to the nasal plane, advanced approximately 2–3 mm along the posterior orbital wall until resistance was felt, then gently rotated ~180° to lacerate the orbital venous plexus. Blood was collected by capillary action into EDTA-coated microtubes. Tibiae, spleen and colon were collected for subsequent studies, with details described later. All steps complied with the AVMA Guidelines and were approved by the Institutional Animal Ethics Committee of the First Affiliated Hospital of Harbin Medical University (IACUC No.2025032).

Fecal samples were collected from mice in the ATO and ATO+BP groups (n = 8 per group). Genomic DNA was extracted using the DP328 Fecal Genomic DNA Extraction Kit (Centrifugal Column Type, TIANGEN, Beijing, China), and DNA quality was assessed based on 260/280 nm and 260/230 nm absorbance ratios.

Library preparation was performed using the Rapid Plus DNA Library Prep Kit for Illumina (RK20208, ABclonal, Wuhan, China). DNA fragmentation was carried out with a Covaris LE220R-plus Ultrasonicator (Covaris, Woburn, MA, USA), and sequencing was conducted on an Illumina NovaSeq X Plus System (PE150, Illumina, San Diego, CA, USA).

Following ATO treatment, alpha diversity (Chao1, Shannon, and Simpson) and beta diversity (Bray–Curtis dissimilarity) indices were calculated to assess microbial community richness, evenness, and structural variation.

Bone marrow aspirates were collected from tibial bone marrow cavities using 1$\times$ DPBS (Thermo Fisher Scientific, 14190144) supplemented with 10 µg/mL DNase I (Sigma-Aldrich, 11284932001) on ice. Cell suspensions were filtered through 40 µm nylon cell strainers (Corning Incorporated, Corning, NY, USA, 352340) and treated with 1$\times$ Red Blood Cell Lysis Solution (Thermo Fisher Scientific, 00-4333-57) to eliminate erythrocytes. After washing with 1$\times$ DPBS containing 2% fetal bovine serum (FBS; Gibco, 10082147), cell viability was assessed using 0.4% Trypan Blue (Thermo Fisher Scientific, 14190144) via Countess® II Automated Cell Counter (Thermo Fisher Scientific).

CD11b immune cells were isolated using CD45 microbeads and MACS LS columns (Miltenyi Biotec, Bergisch Gladbach, Germany, 130-042-301) following the manufacturer’s protocol. Single-cell libraries were generated using the 10$\times$ Genomics Chromium Next GEM Single Cell 3^′ Kit v3.1 (10$\times$ Genomics, Pleasanton, CA, USA, 1000121) according to the CG000206 Rev D protocol. Briefly, cells were partitioned into droplets using the Chromium Controller, followed by reverse transcription, cDNA amplification, and library preparation. Final libraries were quantified via High Sensitivity DNA Analysis (Agilent Bioanalyzer 2100, Agilent Technologies, Santa Clara, CA, USA) and Qubit dsDNA HS Assay (Thermo Fisher Scientific), then sequenced on an Illumina NovaSeq 6000 platform (2 $\times$ 150 bp paired-end configuration).

Raw reads were processed using Cell Ranger v2.1.0 (10$\times$ Genomics) with default parameters for alignment to the GRCm38 mouse genome (STAR aligner v2.7.8a, https://github.com/alexdobin/STAR). Gene-barcode matrices were generated after Unique Molecular Identifier (UMI) counting and barcode filtering. Data integration and batch correction were performed by R package “Seurat” (version 2.3.0, https://cran.r-project.org/web/packages/Seurat/index.html).

The normalized data (NormalizeData function in Seurat package) was performed for extracting a subset of variable genes. Variable genes were identified while controlling for the strong relationship between variability and average expression. Next, we integrated data from different samples after identifying “anchors” between datasets using FindIntegrationAnchors and IntegrateData in the Seurat package. The clusters were visualized by Uniform Manifold Approximation and Projection (UMAP). In order to understand the BMME in APL, HALLMARK analysis was used to evaluate immune cells in samples [17]. The difference in the abundance of immune cell infiltration was analyzed.

Spleens and distal colons were fixed in 10% neutral-buffered formalin (Leica Biosystems, Nussloch, Germany) for 24–48 h, processed routinely, embedded in paraffin, and sectioned at 5 µm. Sections were stained with hematoxylin and eosin (H&E) for histopathological evaluation or subjected to immunohistochemistry (IHC) for Ki67 detection (DAKO, Glostrup, Denmark).

For immunofluorescence (IF), paraffin-embedded sections were deparaffinized, rehydrated, and subjected to antigen retrieval (citrate buffer, pH 6.0, 95 °C for 20 min). After blocking, sections were incubated with primary and fluorescent-conjugated secondary antibodies, followed by DAPI counterstaining. Fluorescence microscopy images were acquired using a Zeiss AXIO Imager M2 system (ZEISS, Oberkochen, Denmark), while bright-field (white light) images were captured using an Olympus CKX53 microscope (OLYMPUS, Tokyo, Japan). Antibody panels including anti-Ki67 antibody (DAKO, Glostrup, Denmark, 1:200) and, for immunofluorescence, Claudin-1 (Servicebio, Wuhan, China, GB11032, 1:500), Occludin (Abcam, Cambridge, UK, ab53032, 1:500), and ZO-1 (Servicebio, Wuhan, China, GB111402, 1:500) were employed. Secondary antibodies (ZSGB-BIO, Beijing, China) and all reagents were prepared following the manufacturers’ instructions.

Bone marrow cells were isolated by flushing the bilateral tibiae and femurs with PBS supplemented with 2% FBS, followed by filtration through a 70 µm nylon mesh for FCM analysis. Bone marrow and peripheral blood cells were treated with red blood cell (RBC) lysis buffer and then filtered through 70 µm nylon strainers (BD Falcon, Franklin Lakes, NJ, USA). Antibodies for FCM were purchased from Thermo Fisher Scientific, MA, USA. The panels included (1) APL cells: anti-hCD11b (11-0118-42) and hCD33 (12-0339-42); (2) monocytes: anti-mCD11b (12-0112-81) and Ly6C (53-5932-82); (3) T cells: anti-mCD3 (11-0032-80), CD8 (17-0081-82), and CD4 (12-0041-81). Each panel included the live/dead stain eBioscience™ Fixable Viability Dye eFluor™ 780 (65-0865-14). Data were acquired on a BD FACSCanto™ II (Becton, Dickinson and Company, NJ, USA) and analyzed using FlowJo software (version 10.8.1, https://www.flowjo.com) .

Data are presented as mean $\pm$ SEM. Normality was assessed via the Shapiro–Wilk test. Parametric comparisons used Student’s t-test (two groups) or one-way ANOVA with Tukey’s post hoc (multiple groups). Non-parametric data were analyzed by the Mann–Whitney U test. Permutational multivariate analysis of variance (PERMANOVA) was used to test for significant compositional differences between groups in PCoA analysis. The PERMANOVA results were obtained based on Euclidean distance with 9999 permutations. Correlations were evaluated via Spearman’s rank ($\alpha$ = 0.05). Statistical software included GraphPad Prism (version 10.1.0, https://www.graphpad.com/) and R (version 4.3.1, https://cran.r-project.org/). For microbiota analysis, DESeq2 (v1.38.3) was used to obtain normalized raw counts, and taxonomic profiling was performed using SILVA 138.1. Raw and processed data are available in the NCBI Sequence Read Archive and the corresponding author’s institutional repository.

Building upon previous reports [8], we analyzed 16S rRNA sequencing data from 20 APL patients treated with ATO based regimens (BioProject ID: PRJNA935705). Our analysis revealed that ATO therapy induced significant alterations in gut microbiota composition, including reduced microbial diversity and depletion of beneficial taxa such as Bifidobacterium. A significant difference in Shannon diversity was observed between ATO-based treatment patients (ATO group) and the non-treated group (APL group), which demonstrated a significantly lower diversity in the feces of ATO group (Fig. 1A). Principal coordinate analysis (PCoA) demonstrated a significant difference in $\beta$-diversity using the weighted UniFrac distance between the ATO and APL groups, suggesting that the microbial composition in APL patients was significantly shifted regarding operational taxonomic units (OTUs) (Fig. 1B). Notably, the abundance of beneficial gut microbiota—including Romboutsia, Blautia, and Bifidobacterium—was significantly decreased following ATO treatment (Fig. 1C), and Spearman correlation analysis further revealed an inverse association between Bifidobacterium and other probiotic taxa (Fig. 1D); these findings raise two critical questions: whether the reduction in these protective bacterial populations is mechanistically linked to the therapeutic efficacy of ATO, and whether similar microbiota alterations occur in animal models where ATO’s pharmacological effects are well-characterized.

Fig. 1.

The diversity and composition of the gut microbiota are significantly altered in APL patients after treatment with ATO. Data of 16S rRNA genes sequenced from APL patients after or before ATO treatment. (A) The diversity and richness of the gut microbiota in APL patients before (APL n = 5) and after ATO treatment (ATO n = 15). (B) PCoA of a weighted UniFrac distance analysis (n = 20), R² = 0.1169, p = 0.0391. (C) Relative taxon abundance comparison among the APL and ATO groups, with Bifidobacterium showing a significant change (n = 17). (D) Spearman correlations between the intestinal content of the 10 genera in the APL and ATO groups (red: positive correlation; blue: negative correlation). *: p 0.05, **: p 0.01, ***: p 0.001. APL, acute promyelocytic leukemia; ATO, arsenic trioxide; PCoA, principal coordinate analysis.

To investigate whether ATO affects the gut microbiota in murine models, we established an APL mouse model and collected fecal samples from both ATO-treated (ATO group) and control (APL group) mice before euthanasia (Fig. 2A). A total of 16 fecal samples (8 per group) were subjected to metagenomic sequencing. Consistent with the human data, murine APL models recapitulated similar dysbiotic patterns. ATO-treated mice showed a relative reduction in OTUs compared to controls (Fig. 2B–D). Principal component analysis (PCA) of UniFrac distances confirmed microbiota shifts (Fig. 2E), and the heatmap showed a decrease in Bifidobacterium abundance, especially BP (Fig. 2F).

Fig. 2.

The diversity and composition of the gut microbiota are significantly altered in APL mouse models after treatment with ATO. (A) Schematic diagram of the mouse experimental process. (B–D) Alpha and beta diversity based on data of bacterial metagenomics from APL murine models after or before ATO treatment (n = 8, respectively), *p 0.05 vs. ATO. (E) PCA of a weighted UniFrac distance analysis (n = 8, respectively). (F) The heatmap indicates a significant reduction in the abundance of BP after ATO treatment. PCA, principal component analysis.

Although LEfSe analysis did not identify significant abnormalities based on our metagenomic sequencing data, we observed abundance differences of BP after ATO treatment, consistent with evidence from the previous literature suggest that ATO treatment markedly reduces Bifidobacterium abundance [8]. To investigate the functional role of BP bacteria during ATO therapy, we supplemented ATO administration with BP. After 14 days of treatment, bone marrow and peripheral blood samples were collected for FCM analysis (Fig. 3A). Notably, the combination of BP supplementation and ATO (BP+ATO) significantly reduced APL cell burden. Specifically, in ATO+BP-treated mice, peripheral blood leukemic blasts were reduced by 72% (4.57% $\pm$ 0.49% vs. 16.53% $\pm$ 7.09%; p = 0.043) (Fig. 3B), while in bone marrow (CD11b$\pm$/CD33+), they were reduced by 64% (2.76% $\pm$ 0.94% vs. 7.39% $\pm$ 1.07%; p = 0.005) (Fig. 3C). Histopathological analysis of spleens revealed fewer leukemic infiltrates in the ATO+BP group (Fig. 3D). These findings collectively indicate that BP supplementation synergistically enhances the therapeutic efficacy of ATO against APL in murine models.

Fig. 3.

BP alleviated APL cell burden in mouse models. (A) Schematic diagram of the mouse experimental process. (B)Leukemia cells (CD11b+/CD33+ cells) in peripheral blood from ATO mice (n = 4) and ATO+BP mice (n = 4). p = 0.043. (C) Leukemia cells (CD11b+/CD33+ cells) in bone marrow from ATO mice (n = 4) and ATO+BP mice (n = 4). p = 0.005. (D) H&E histopathology sections and Ki-67 IHC of a representative spleen: control, ATO group, and ATO+BP group. All microscopic analyses were performed (Scale bar = 500 µm, left and middle; scale bar = 200 µm, right). BP, Bifidobacterium pseudolongum; H&E, hematoxylin and eosin.

Here, we conducted paired scRNA-seq and scTCR/BCR-seq profiling on 6 bone marrow (BM) aspirates (3 with and 3 without ATO) (Fig. 4A). As the mice had undergone irradiation prior to leukemia cell inoculation, both groups exhibited markedly low hematopoietic stem cell (HSC) proportions compared to clinical APL patients. ATO treatment induced profound remodeling of the immune microenvironment. Unsupervised immune cell clustering and scoring revealed widespread transcriptional alterations across major immune populations (Fig. 4B,C). Notably, while CD8+ T cells showed reduced abundance, post-ATO treatment. This depletion might reflect exhaustion following APL remission, suggesting that strategies to preserve or expand CD8+ T cells could amplify ATO’s immunotherapeutic efficacy. In contrast, B cells and basophils remained stable. HALLMARK analysis demonstrated commonly enhanced inflammatory responses especially in neutrophilsm monocytes and CD8+ T cells (Fig. 4D).

Fig. 4.

Landscape of Microenvironment in APL murine models with ATO treatment by single-cell RNA sequencing and the functional states of APL subtypes. (A) UMAP plot of BM samples from murine models with or without ATO. (B) Heatmap showing the scaled expression levels of genes in specific subtypes. (C) Proportion comparison of cell fraction between APL and ATO group. (D) Comparative immunophenotypic analysis of ATO-treated versus untreated mice revealed distinct immune modulation patterns. ATO administration induced significant immunomodulatory effects across multiple immune cell subsets: monocytes: p = 0.0029, CD8+ T cells: p = 0.057, BAS: p = 0.57, M1 macrophages: p = 1.1 $\times$ 10^-7, M2 macrophages: p = 0.62, B cells: p = 0.37, Neutrophils: p = 2.2 $\times$ 10^-16 and Neutrophil precursor: p = 4.9 $\times$ 10^-11. UMAP, Uniform Manifold Approximation and Projection; BM, bone marrow; BAS, Basophil.

Previous studies have demonstrated that probiotics can mitigate intestinal barrier dysfunction caused by pathological conditions or pharmacological interventions [8, 10, 18]. In this study, we investigated whether BP supplementation could counteract ATO-induced intestinal barrier impairment. Using immunofluorescence analysis, we quantified the expression of tight junction proteins (Claudin-1, Occludin, and ZO-1) in colonic epithelial cells. ATO treatment alone compromised intestinal barrier function, as evidenced by the decreased expression of tight junction proteins Claudin-1, Occludin, and ZO-1 (Fig. 5A, Supplementary Fig. 1A–C). Strikingly, BP supplementation in ATO-treated mice restored tight junction integrity, as shown in Fig. 5A. Interestingly, previous studies have demonstrated that BP bacteria modulate the tumor microenvironment, particularly by increasing CD8+ T cell proportions and enhancing antitumor immunity [19]. Consistently, our flow cytometry analysis of peripheral blood and bone marrow from ATO $\pm$ BP-treated mice revealed that BP co-administration significantly elevated monocyte and CD8+ T cell frequencies as FCM results shown. Monocytes exhibited the most pronounced alteration, demonstrated by a 1.42-fold increase in peripheral blood (71.90% $\pm$ 6.25% vs. 48.47% $\pm$ 4.51%; p = 0.006; Fig. 5B) and a 2.18-fold expansion in bone marrow (16.23% $\pm$ 5.95% vs. 7.44% $\pm$ 1.48%; p = 0.068; Fig. 5C). Notably, despite their low abundance, BP co-treatment significantly modulated CD8+ T cell numbers; in peripheral blood, there was a 1.69-fold increase (1.69% $\pm$ 0.12% vs. 0.99% $\pm$ 0.24%; p = 0.012; Fig. 5D) and in bone marrow a 2.21-fold elevation (3.18% $\pm$ 0.52% vs. 1.43% $\pm$ 0.38%; p = 0.009; Fig. 5E).

Fig. 5.

BP remodeled the immune microenvironment and repaired the intestinal barrier in murine models. (A) Representative IF images showing Claudin-1, Occludin, and ZO-1 (all in green) localization with DAPI nuclear counterstain (blue) (n = 4 per group), quantification data were shown in Supplementary Fig. 1. Scale bar: 200 µm. (B) Monocytes (CD11b+Ly6c+ cells) in peripheral blood from ATO mice (n = 4) and ATO+BP mice (n = 4). p = 0.006. (C) Monocytes (CD11b+Ly6c+ cells) in bone marrow from ATO mice (n = 4) and ATO+BP mice (n = 4). p = 0.068. (D) CD8+ T cells (CD3+/CD8+ cells) in peripheral blood from ATO mice (n = 4) and ATO+BP mice (n = 4). p = 0.012. (E) CD8+ T cells (CD3+/CD8+ cells) in bone marrow from ATO mice (n = 4) and ATO+BP mice (n = 4). p = 0.009. IF, immunofluorescence; DAPI, 4^′,6-diamidino-2-phenylindole.