Controlled production of bacterial outer membrane vesicles in intestine by orally administered bacteria as an effective tumor vaccine

The orally administered tumor vaccines are still limited, and the major challenges are the complex gastrointestinal environment and intestinal epithelial barriers. Here, we present an effective tumor vaccine based on the OMVs which are generated by orally administered genetically engineered bacteria in intestine. A tumor antigen was fused with the surface protein ClyA on OMVs, whose expression was controlled by an arabinose-inducible promoter. Through oral administration of the modified bacteria and arabinose, in situ controllable production of OMVs loaded with tumor antigen in the intestine was achieved. The OMV-based tumor vaccine not only overcame the intestinal epithelial barriers to reach the immune cells in the lamina propria, but also stimulated dendritic cell maturation to facilitate a potent antitumor adaptive immunity. The oral bacteria generated OMV-based tumor vaccine significantly inhibited the lung metastatic melanoma and subcutaneous colon cancer in mouse models. Furthermore, a robust immune memory was generated, offering long-term protection against tumor challenge. Our work provides a proof-of-concept for oral bacteria generated OMV-based tumor vaccine that may of the surface protein, ClyA, of OMVs (ClyA-Ag-mFc). We hypothesized that the mFc would enhance the recognition and uptake of OMVs by DCs via the interaction between Fc and the neonatal Fc receptor (FcRn) 30-32 . In addition, in order to realize the strict vaccination route and avoid immune tolerance caused by long-term antigen stimulation, we introduced an arabinose-inducible promoter to control the fusion protein expression 33,34 . Through oral administration of the modified bacteria and expression inducer arabinose (Ara), we achieved in situ controllable production of OMVs loaded with tumor antigen (OMV-Ag-mFc) in the intestine. These OMV-Ag-mFc effectively crossed the intestinal epithelial barriers and were taken up by DCs in the lamina propria, followed by lymph node drainage and tumor antigen presentation. Tumor antigen-specific immune activation led to a significant inhibition of tumor growth and resistance to tumor challenge in multiple murine cancer models. T cells in the splenocytes of immunized mice by flow cytometry, using IFN- γ and OVA tetramer as markers of activation and antigen specificity, respectively 37 . As shown in Figure 4c-4f , the proportions of either IFN- γ + or OVA tetramer + in CD3 + CD8 + T cells in splenocytes re-stimulated with OVA 257–264 antigen peptide were the highest in the oral ClyA-OVA-mFc vaccine group. In addition, the levels of lung-infiltrating CD8 + CTLs were also the highest in the oral ClyA-OVA-mFc vaccine group ( Figure 4g and 4h ). These results demonstrate that the ClyA- OVA-mFc is an effective tumor vaccine that can successfully elicit antitumor immunity. these results verify that the oral bacteria-derived OMV-based tumor vaccine can induce effective immune memory, which is critical for a long-term prevention of tumor recurrence.


Fig. 1 | Schematic illustration of the genetically engineered bacteria-derived OMV-based oral tumor vaccine.
Engineered E. coli were obtained by transformation with a plasmid expressing ClyA (a surface protein on OMVs) fused with a tumor antigen and Fc fragment of mouse IgG (ClyA-Ag-mFc). An arabinose-inducible promoter (araBAD) was introduced to control the fusion protein expression. After oral administration of engineered E. coli and expression inducer, Ara, OMVs loaded with tumor antigen (OMV-Ag-mFc) are produced in the intestine of mice. By virtue of the OMVs' natural behavior, OMV-Ag-mFc effectively penetrate the intestinal epithelial barriers, and are recognized and taken up by DCs in lamina propria via the interaction between mFc and FcRn. Due to the amount of PAMPs in OMVs, DCs rapidly mature, which then drain to the lymph node and complete tumor antigen presentation, resulting in expansion of antigen-specific cytotoxic T lymphocytes and memory T cells.

Biodistribution analysis of engineered E. coli after oral administration in vivo
To visualize the biodistribution of the bacteria, we first engineered E. coli (Top10 strain) to express luciferase fused with an HA tag (Luc-HA) under the control of an araBAD promoter, which is activated by Ara (pBAD-Luc-HA). As shown in Figure 2a, using western blot analysis, the Luc-HA expression was detected only in the pBAD-Luc-HA E. coli after Ara induction. After adding the luciferase substrate (fluorescein potassium), fluorescence was observed in the pBAD-Luc-HA E. coli group after adding Ara (Figure 2a). Next, we administered the engineered E. coli orally to mice, followed by oral Ara. After 12 h, the gastrointestinal tissues from different locations and feces were collected for bioluminescence detection. Bioluminescent signals were observed in the contents of the cecum, colon, and feces from mice in the pBAD-Luc-HA E. coli + Ara group (Figure 2b). These results demonstrate that the orally administered, engineered E. coli were able to reach the intestine and tolerate the intestinal environment, which ensured the successful expression of the target fusion protein when Ara was also taken orally. In addition, the extent of the induction in the engineered bacteria was Ara concentration-dependent, and 20 g/L of Ara was chosen for the further experiments (Figure 2c).
Next, we evaluated the biodistribution and pharmacokinetics of the engineered E. coli. As shown in Figures 2d and 2e, after oral administration with pBAD-Luc-HA E. coli and Ara, the bioluminescent signals mainly appeared in the cecum at 2 h after administration, and gradually moved to the colon within 12 h. After 24 h, the bioluminescent signal became weak, due to the metabolic clearance of the bacteria, which is important for the biosafety of an oral vaccine.

Figure 2 | Biodistribution analysis of engineered E. coli after oral administration. a,
Representative western blot analysis and in vitro bioluminescence images of engineered E. coli expressing luciferase fused with the HA tag (Luc-HA) under the control of Ara induction. b, In vivo bioluminescence images of intestinal contents at 12 h after oral administration of engineered E. coli with or without oral Ara solution. c, Optimization of Ara concentration for in vivo induction. After oral administration of pBAD-Luc-HA E. coli, mice were provided the indicated concentrations of Ara. Feces were collected after 12 h and the bioluminescence intensities of luciferase were analyzed (n = 3). d-e, The biodistribution and pharmacokinetics of pBAD-Luc-HA E. coli in vivo. The biodistribution of pBAD-Luc-HA E. coli in the gastrointestinal tract at different timepoints post oral administration with 20 g/L Ara was analyzed using in vivo bioluminescence imaging (d). The bioluminescence intensities of luciferase in different locations along the gastrointestinal tract were quantified (n = 3; e). Data are presented as the mean ± SD.

Epithelial penetration and immune stimulation of the oral vaccine
To investigate whether the oral vaccine was able to stimulate the immune system, we fused an epitope of ovalbumin (OVA) [OVA257-264 (SIINFEKL)] and/or mFc to the C-terminal of the surface protein, ClyA, of OMVs to generate ClyA-mFc, ClyA-OVA and ClyA-OVA-mFc, all of which were labelled with an HA tag. Using western blot analysis to evaluate the HA-tag expression, we verified the Ara-controlled expression of these fusion proteins in the engineered bacteria and their secreted OMVs (Supplementary figure 1a). In the transmission electron microscopy (TEM) observation, all OMVs exhibited a uniform spherical morphology with a bilayer structure (Supplementary figure 1b). The diameters of the OMVs detected by dynamic light scattering (DLS) in control, ClyA-mFc, ClyA-OVA and ClyA-OVA-mFc groups were 21.5, 22.0, 23.9 and 30.3 nm, respectively (Supplementary figure 1c).
We proceeded to investigate the ability of OMVs to penetrate the intestinal epithelial barriers in vitro and in vivo. To this end, we co-cultured colon cancer Caco-2 and HT29 cells in the upper chamber of a transwell system for 21 days to simulate the intestinal epithelial barriers 35 . Next, DC2.4 cells were seeded in the lower chamber, and OMVs from different engineered bacteria were added into the upper chamber (Supplementary   figure 2a). The percentages of mature DCs (using the co-stimulatory molecules, CD80 + CD86 + , as markers) were analyzed by flow cytometry after 12 h. OMVs effectively penetrated the epithelial barriers and activated DCs in the lower chamber. As expected, OMVs expressing mFc (ClyA-mFc and ClyA-OVA-mFc) exhibited a higher activation efficiency (Supplementary figure 2b). Next, we performed an enema experiment to examine the penetration ability of OMVs in vivo. The mice were anesthetized, the colon was ligated, and different OMVs were injected into the ligated intestinal cavity. After a 2 h incubation, the ligated colon was excised for immunofluorescence analysis. OMVs were detected using an anti-HA antibody (red), and DCs were labelled with an anti-CD11c antibody (green). As shown in Figures 3a and 3b, there was apparent penetration of the intestinal epithelial barriers by the OMVs in ClyA-mFc, ClyA-OVA and ClyA-OVA-mFc groups. Additionally, OMVs with mFc exhibited greater penetration (Figure 3b) and DC affinity (Figure 3a, white arrows) than OMVs without mFc. In summary, the OMVs secreted by engineered E. coli were able to penetrate the intestinal epithelial barriers, and the ClyA fusion with mFc was beneficial for this process.
We next examined the ability of the oral bacteria-derived vaccine to stimulate the immune system in vivo.
Mice received vaccination through oral administration with different engineered bacteria, and consumed the Ara solution for 12 h. After a single vaccination, the immune cells in the lamina propria were isolated and analyzed by flow cytometry (Supplementary figure 3). As shown in Figure 3c, compared with the oral phosphate buffered saline (PBS) group, the proportions of CD80 + CD86 + cells in CD11c + DCs significantly increased in all the oral bacteria-derived vaccine groups, even in the group that received the oral ClyA-OVA-mFc bacteria without Ara induction [ClyA-OVA-mFc (-Ara)]. The proportions of CD80 + CD86 + cells in CD11c + DCs were higher in the oral ClyA-mFc and ClyA-OVA-mFc vaccine groups than that in the oral ClyA-OVA vaccine group (Figure 3c), which may be due to the mFc-mediated epithelial penetration enhancement. These results suggest that the immune cells in the lamina propria of mice tolerate the intrinsic symbiotic bacteria, while the orally administered foreign bacteria stimulated the DCs in the lamina propria via an OMV-mediated mechanism. We studied the antigen-specific immune responses after 3 rounds of immunization with different oral bacteria-derived vaccines (days 0, 3 and 8). On day 12, the splenocytes were collected and stimulated with an OVA257-264 antigen peptide. As shown in Figure 3d and Supplementary figure 4, the oral ClyA-OVA-mFc vaccination stimulated the highest number of antigen-specific T cells (OVA tetramer + in CD3 + CD8 + cells).
Compared with that in the oral ClyA-OVA-mFc vaccine group, the markedly lower number of antigen-specific T cells in the oral ClyA-OVA-mFc (-Ara) vaccine group verifies that the immune stimulation of this oral bacteriaderived vaccine could be controlled by oral administration of Ara (Figure 3d). We also evaluated the cytotoxic effects of these collected splenocytes on ovalbumin-expressing melanoma B16 (B16-OVA) and colon cancer MC38 cells. As shown in Figure 3e, the splenocytes from the ClyA-OVA-mFc vaccine group exhibited a stronger cytotoxic effect against B16-OVA cells than those from the ClyA-OVA vaccine group, indicating the importance of the epithelial penetration enhancement by mFc modification. The cytotoxic effect in all groups disappeared when using MC38 cells, which do not express the OVA antigen, thus demonstrating the antigen specificity of the immune stimulation by the oral bacteria-derived vaccine (Figure 3f). Taken together, the novel oral vaccine was able to produce OMVs loaded with tumor antigen in the intestine under the control of Ara. The in situ OMVs vaccine could penetrate the intestinal epithelial barriers and activate antigen-specific T cells against tumor cells. Mice were anesthetized, the colon was ligated and different OMVs were injected into the ligated intestinal cavity. After a 2 h incubation, the ligated colon was excised for immunofluorescence analysis (a). The cell nucleus was stained with DAPI (blue), OMVs were detected using an anti-HA antibody (red) and DCs were labelled with an anti-CD11c antibody (green). The white arrows indicate the interaction between OMVs and DCs. The numbers of penetrating HA-labelled OMVs in each area were quantified (n = 4; b). Scale bar, 100 μm. c-f, Immune stimulation analysis of the indicated oral vaccines. C57BL/6 mice were immunized with an oral vaccine or PBS on days 0, 3 and 8. On day 1, after a single vaccination, the immune cells in the lamina propria were isolated.

Antitumor effects in a lung metastatic melanoma model
Considering that the majority of immune cells in the entire immune system reside in the intestine, a specific immune stimulation in the intestine is expected to elicit a robust antitumor effect 17,36 . To evaluate the antitumor effects of the oral vaccine on metastatic melanoma, we inoculated mice intravenously with B16-OVA cells on We also assessed the degree of activation of antigen-specific T cells in the splenocytes of immunized mice by flow cytometry, using IFN-γ and OVA tetramer as markers of activation and antigen specificity, respectively 37 . As shown in Figure 4c-4f, the proportions of either IFN-γ + or OVA tetramer + in CD3 + CD8 + T cells in splenocytes re-stimulated with OVA257-264 antigen peptide were the highest in the oral ClyA-OVA-mFc vaccine group. In addition, the levels of lung-infiltrating CD8 + CTLs were also the highest in the oral ClyA-OVA-mFc vaccine group (Figure 4g and 4h). These results demonstrate that the ClyA-OVA-mFc is an effective oral tumor vaccine that can successfully elicit antitumor immunity. . c-f, Antigen-specific immune response analysis. The splenocytes were collected on day 17 and stimulated with OVA peptide. The percentages of IFN-γ + (c,d) or OVA tetramer + (e,f) in CD3 + CD8 + cells in the splenocytes were analyzed by flow cytometry (n = 5). g-h, Immunohistochemical staining of CD8 + cells (brown staining) in the lung tissues collected on day 17 (g) and the quantitation of CD8 + cells in each field (n = 3; h). Scale bar, 2 mm in the upper panels and 100 μm in the lower panels. The data are presented as the mean ± SD and were analyzed by one-way analysis of variance (ANOVA) with GraphPad Prism software. N.S., no significance; * , P < 0.05; ** , P < 0.01; *** , P < 0.001; **** , P < 0.0001. Tumor volumes were recorded every two days until day 19. As shown in Figure 5b and Supplementary figure 6a, the mice in the oral ClyA-Adpgk-mFc vaccine group exhibited a stronger inhibition of tumor growth compared to the mice in other groups. The tumors were harvested on day 19 (Supplementary figure 6b), and the tumor weight in the oral ClyA-Adpgk-mFc vaccine group was also the lowest compared with that in the other groups (Figure 5c). We calculated the tumor inhibition rate according to the tumor volume at the endpoint (Figure 5d). Compared with the oral PBS group, the tumor inhibition rate in the oral ClyA-Adpgk-mFc vaccine group was 81.2%, which was higher than the 64.1% in the subcutaneous Poly(I:C) + Adpgk vaccine groups.

Antitumor effects in a subcutaneous colon cancer model
In addition, we examined the body weight and morphology of the major organs in the different groups (Supplementary figures 7 and 8), all of which demonstrate that these oral vaccines were well tolerated, with no apparent toxicity.
We evaluated the antigen-specific immune response in blood and splenocytes, respectively. Increased proportions of Adpgk tetramer + cells in CD3 + CD8 + T cells in blood were found in the oral ClyA-Adpgk-mFc vaccine and subcutaneous Poly(I:C) + Adpgk vaccine groups; this effect was stronger in mice immunized with the oral ClyA-Adpgk-mFc vaccine (Figure 5e). The gating strategies and representative flow plots were presented in Supplementary figures 9a and 9b. Next, the enzyme-linked immunospot (ELISPOT) assay was used to examine IFN-γ secretion from splenocytes re-stimulated with Adpgk antigen peptide (Figure 5f). As expected, most IFN-γ was produced by the splenocytes from the mice immunized with the oral ClyA-Adpgk-mFc vaccine (Figure 5g). These results further confirm that our oral vaccine successfully activated the antigenspecific immune response more strongly than the common formulation used in clinical trials.
Next, we detected the infiltrating immune cells in the tumor tissues by flow cytometry. The cells we assessed included CD3 + T cells, CD3 + CD4 + T cells, CD3 + CD8 + T cells (Supplementary figures 10a and 10b),

Evaluation of long-term immune memory
To investigate the long-term immune memory and benefits of treatment with the oral vaccine formulations, we treated healthy mice 3 times with the various oral vaccinations on days 0, 3 and 8 (Figure 6a). Splenocytes were isolated on day 50 and the proportions of central memory T cells and effector memory T cells were analyzed by flow cytometry (Figure 6b and Supplementary figure 14). As shown in Figure 6c, the proportion of effector memory T cells (CD3 + CD8 + CD44 + CD62L -) in the oral ClyA-OVA-mFc vaccine group was significantly greater than that in the other oral vaccine and PBS groups. Although a decrease in central memory T cells (CD3 + CD8 + CD44 + CD62L + ) was observed in the oral ClyA-OVA-mFc vaccine group compared with that in the PBS group, similar phenomena were observed in the other oral vaccine groups (Figure 6d). Next, the immunized mice were challenged with B16-OVA cells by tail vein injection on day 50, and the lungs were harvested on day 65 and imaged (Figure 6e). In contrast to the dense metastases in the PBS, oral ClyA-mFc vaccine and oral ClyA-OVA-mFc (-Ara) vaccine groups, the pulmonary metastases in the oral ClyA-OVA vaccine and oral ClyA-OVA-mFc vaccine groups were significantly reduced, with the protective effect against tumor challenge stronger in the oral ClyA-OVA-mFc vaccine group (Figure 6f). Collectively, these results verify that the oral bacteria-derived OMV-based tumor vaccine can induce effective immune memory, which is critical for a long-term prevention of tumor recurrence. . e-f, Evaluation of resistance to tumor challenge. The lungs were collected on day 65 (e) and the number of tumor nodules were enumerated (f; n = 8). Scale bar, 1 cm. The data are presented as the mean ± SD and were analyzed by one-way analysis of variance (ANOVA) with GraphPad Prism software. N.S., no significance; *** , P < 0.001; **** , P < 0.0001.

Conclusions
Here we present an oral tumor vaccine that differs from previous nanomaterials-based oral vaccine delivery technologies in that it comprises bacteria-membrane vesicle-derived particles 4 . Inspired by the natural phenomenon that commensal bacteria in the gastrointestinal tract can interact with the host immune cells through OMVs, we engineered bacteria to develop an oral tumor vaccine 23 . In some previous studies, engineered bacteria were directly explored for tumor treatment, achieving targeted proliferation and drug delivery to tumor tissue through the complex genetic engineering 40,41 . In the present study, the working environment of the oral vaccine was the intestine, which, as an application strategy, is more biocompatible given the large number of commensal bacteria in the organ. In addition, we exploited bacteria-secreted OMVs as messengers for tumor antigen delivery. This natural mechanism-mediated epithelial penetration is similar to that in live attenuated virus vaccines, however the non-viral delivery vehicles, OMVs, possess much enhanced biosafety profile. This natural OMV-inspired epithelial penetration strategy may have implications for the development of mucosal vaccines at other sites.
Behavioral controllability is a major challenge for in vivo applications of engineered bacteria 42 . Although we found that most of the engineered bacteria were cleared after 24 h, it was still necessary to control the production of OMVs loaded with tumor antigens to avoid immune tolerance due to long-term antigen stimulation. In some previous studies of bacterial applications in vivo, a common approach to control gene expression in engineered bacteria has been to introduce an environment-responsive promoter, such as one that responds to hypoxia or low pH in the tumor microenvironment 40,43,44 . However, an environment-responsive promoter is not a bona fide switch type of control method; once the engineered bacteria are exposed to the target environment, the gene expression remains in the open status. Therefore, we adopted an arabinoseinducible promoter to achieve switchable control of the engineered bacteria through inducer administration 45 .
In summary, we genetically engineered E. coli, one of the most abundant commensal bacteria in the gut, to establish a bacteria-derived OMV-based oral tumor vaccine. Using this strategy, we achieved in situ Pharmacokinetic study. Female C57BL6 mice were orally administered 10 9 engineered E. coli and 20 g/L arabinose water. The mice were then euthanized at 0, 2, 5, 12 and 24 h and the digestive tract tissue (from stomach to colon) were isolated. 2 mL 15 mg/mL D-Luciferin potassium solution was infused into the digestive tract for bioluminescence analysis.
OMVs preparation and characterization. TOP10 bacteria were cultured as described above and arabinose (2 g/L) was added to further induce protein expression when the OD600 reached 0.6-0.8. Bacteria was incubated overnight at 16℃ with shaking at 160 rpm and removed by centrifugation at 5000 ×g for 10 min at To analyze the intracellular IFN-γ and OVA tetramer positive T cells in the spleen, splenocytes collected at the end of the experiment were incubated with OVA peptide overnight to re-stimulate the T lymphocytes.
Splenocytes treated with ionomycin were used as a positive control group. For intracellular IFN-γ staining, the surface proteins CD3 and CD8 were first stained before fixation of the cells with a commercially available fixation buffer (BioLegend, USA, Catalog No. 420801). After fixation and permeabilization, the cells were subjected to IFN-γ staining. Finally, the cells were washed with PBS and analyzed using a flow cytometer. For OVA tetramer positive T cell staining, T-Select MHC Tetramer was added prior to the CD3 and CD8 flow cytometry antibodies. After incubation with T-Select MHC Tetramer for 30 min at 4℃, the cells were stained with CD3 and CD8 antibodies for another 30 min and analyzed using a flow cytometer.

Immunohistochemical analysis of T lymphocyte infiltration into lungs.
The lungs harvested at the end of the lung metastasis experiment were fixed with 4% paraformaldehyde and embedded in paraffin, which was cut into 7 μm sections for immunohistochemical staining of CD8. The sections were deparaffinized, rehydrated and treated with 3% H2O2 to eliminate the activity of endogenous peroxidase. The sections were subjected to antigen retrieval, blocking with 5% goat serum and incubation with an anti-CD8 antibody (Abcam, UK, catalog No. ab93278, dilution: 1:100) overnight. The sections were then incubated sequentially with a goat anti-rabbit IgG biotinylated antibody (Biorbyt, UK, catalog No. orb153693, dilution: 1:100) and HRP-conjugated streptavidin. Finally, the sections were stained with DAB for color development and counterstained with hematoxylin.
Antitumor efficacy in subcutaneous colon tumor models. In order to evaluate the antitumor effect of the oral vaccines in a solid tumor model, 2 × 10 6 murine colon adenocarcinoma (MC38) cells were injected subcutaneously into the right flank of C57BL/6 mice on day 0. The mice were orally administered with ClyA-HA-mFc, ClyA-HA-Adpgk or ClyA-HA-Adpgk-mFc engineered bacteria or PBS on days 3, 6 and 11 (n = 6).
Mice were provided 20 g/L arabinose water for 12 h after each vaccination, if applicable. Mice subcutaneously immunized with a mixture of 50 µg adjuvant Poly (I:C) + 50 µg Adpgk were used as a positive control group.
Mice treated with PBS were used as a negative control group. The tumor volume was measured every other day using Vernier calipers and calculated by the following formula: tumor volume = length × 1/2 width 2 . The mice were euthanized on day 19. The tumors were harvested, weighed and digested into single cell suspensions to analyze the infiltrating immune cells by flow cytometry and immunofluorescence. Spleens were collected for IFN-γ ELISPOT analysis. According to the manufacturer's instructions, splenocytes were seeded in a 96-well plate (1 × 10 5 cells/well), pre-coated with a mouse anti-IFN-γ antibody, and incubated with Adpgk peptide or ionomycin for 20 h. A biotinylated antibody specific for IFN-γ and alkaline-phosphatase conjugated to streptavidin were subsequently used to detect the IFN-γ secreted by the restimulated T cells. By adding a substrate solution, visual spots were formed at the sites of captured IFN-γ, and automated spot quantification was caried out by Dakewe Biotech Co., Ltd. The major organs, including heart, liver, spleen, lung and kidney, were collected for hematoxylin and eosin (H&E) staining. Immunofluorescence. Samples were collected and rapidly frozen in optimal cutting temperature (OCT) compound. They were then cut into sections (7 μm) and washed with PBS three times to remove OCT compound. The sections were incubated with 0.1% Triton X-100 for 30 min, blocked using 10% goat serum for 1 h and incubated overnight with an anti-CD11b antibody (GB11058, Servicebio, China), anti-HA-tag antibody (ab1424, Abcam, USA) or anti-CD8 antibody (ab203035, Abcam, USA), followed by washing three times with PBS and incubating with secondary antibodies conjugated with Alexa Fluor 488 for 2 h. Nuclei were stained with DAPI. The sections were analyzed using a confocal microscope (Zeiss LSM710, Germany).

Statistical analysis.
The data are presented as means ± SD. One-way analysis of variance (ANOVA) was used for multiple comparisons. GraphPad Prism 5 and FlowJo V10 were used to analyze experimental data.

Data availability
Source data are provided with this paper or available from the corresponding author upon reasonable request.