Engineered E. coli expressing bacterial hyaluronidase effectively degrades hyaluronan acid in human tissues
E. coli were genetically engineered to express the bacterial hyaluronidase by cloning the PelB signal sequence at the N-terminus of hyaluronidase (HAse) coding sequence, under the control of a rhamnose-inducible promoter, generating the E. coli-HAse bacterial strain. We determined the growth kinetics of E. coli-HAse expressing strains over 24 hours in non-induced and induced (0.2% rhamnose) conditions (Supplementary Fig. 1a). The HAse expression and activity were assessed by incubating E. coli-HAse in LB media, supplemented with 0.4 mg/ml purified hyaluronan in the presence and absence of rhamnose. LB media supplemented with 0.4 mg/ml purified hyaluronan was used as a negative control. The OD600 in each condition was measured over 6 h. Our findings show that under induced conditions, E. coli-HAse is capable of degrading purified hyaluronic acid, as demonstrated by cetyltrimethylammonium bromide turbidimetric method (CTM ) (Supplementary Fig. 1b). As the expressed Hyaluronidase contained a Flag tag, we were able to detect it by western blot in both bacterial pellets and conditioned media, confirming effective enzyme release (Supplementary Fig. 1c). Consistent with this, the supernatant collected from induced E. coli-HAse effectively degrades hyaluronic acid, whereas no HA degradation was observed in the supernatant collected from uninduced bacterial culture when compared to a control engineered strain expressing GFP under the same promoter (Supplementary Fig. 1d).
Next, we sought to examine the HAse ability to break down human hyaluronan in its native form. For this, we tested bacteria activity in tumor samples obtained from breast cancer patients, both as 2D tissue sections and 3D tissue explants. Tumor tissue explants from patients with triple-negative breast cancer were decellularized as previously described, in a process that is shown to maintain the structure, heterogeneity and composition of the ECM 24. E. coli-HAse was incubated with decellularized tissue explant material in the presence or absence of the hyaluronidase expression inducer molecule, rhamnose. After 4 h, the explant material was stained with biotylinated hyaluronic acid binding protein and imaged by light microscopy, showing that explants incubated with E. coli-HAse lose almost all hyaluronic acid content. E. coli-HA was further added to 2D tumor tissue sections at increasing CFU doses, showing dose-dependent degradation with maximum degradation occurring at the 107 CFU dose (Supplementary Fig. 1e-g).
E. Coli-Hase facilitates hyaluronic acid degradation through tumor-specific colonization
To validate the tumor-specific colonization of the programmable bacteria, we engineered a bioluminescent version of Neobe’s E. Coli bacterial chassis strain, constitutively expressing luciferase using the luxCDABE operon 25. This E. Coli-lux engineered strain enabled us to monitor the biodistribution of Neobe’s E. Coli chassis in the murine 4T1 triple-negative breast cancer model. Bacteria were administered at various doses (106, 107 and 108 CFUs) via a single intravenous injection, and mice were imaged at 24, 48 and 72 h post-administration using bioluminescence in vivo imaging. Mice administered the bioluminescent E. coli strain at doses of 106 CFU and 107 CFU maintained mean body weights comparable to those of mice administered the vehicle alone, while the higher dose of 108 CFU per mouse caused greater weight loss (Supplementary Fig. 2a, b). Importantly, all doses exhibited tumor-specific colonization with the highest dose being evident as early as 24 h. Given the weight loss in the group administered 108 CFU of the candidate strain, the level of colonization in the tumor and secondary organs was quantified in mice administered a dose of 107 CFU at 72 h post-administration. Luminescence was localized in the tumor and CFU counting showed a high concentration of bacteria in the tumor (108 CFU/g tissue) with residual levels of bacteria in secondary organs (around or below 104 CFU/g tissue, Supplementary Fig. 2c). These findings demonstrate that Neobe’s E. coli chassis is well-tolerated and localizes to and accumulates in tumor tissue.
Having confirmed the tumor colonizing ability of the chassis strain at well-tolerated doses, we proceeded to test the efficacy of the hyaluronidase-expressing strain in remodeling the tumor microenvironment. To ensure that payload expression is contained within the tumor, two different variants of the E. Coli-HAse different strains were generated expressing hyaluronidase under hypoxia-inducible promoters (C35R7 and C35R9) and tested together with an additional variant expressing hyaluronidase under a constitutive promoter (phelp). After confirming the tumor-specific colonization, we sought to investigate the potential of the three variants in remodeling the TME through HA degradation in orthotopic 4T1 tumors that are rich in HA and other ECM components 26,27. Tumor-bearing mice were intravenously injected with either phelp, C35R7 or C35R9 bacterial constructs at a dose of 106 and 107 CFUs, to identify the most effective candidate and dose for remodeling the TME. Changes in HA content were evaluated by immunofluorescence staining using hyaluronan binding protein (HABP) following tumor removal 7 days after bacterial administration (Fig. 1a). Interestingly, our data demonstrate that only the low dose of C35R7 can significantly reduce HA levels in 4T1 tumors, as indicated by the 2-fold reduction in HABP signal (Fig. 1b). As expected, none of the bacterial constructs had an antitumor effect, nor did they cause a reduction in body weight exceeding the 20% cutoff compared to the control group. The minimal weight loss observed the day after bacterial treatment is attributed to the shock to which mice were subjected to (Supplementary Fig. 3a, b). Nevertheless, we observed that administration of the high dose of phelp and C35R7 live biotherapy could cause spleen enlargement, potentially as an acute reaction to the initial bacterial dose (Supplementary Fig. 3c).
HA degradation leads to reduction of tumor stiffness and restoration of perfusion
HA abundance in desmoplastic tumors determines in large part the tumor’s mechanical microenvironment by inducing tissue stiffening, which promotes vessels collapsing and hypo-perfusion 26-28. To assess the role of the engineered bacteria in remodeling the TME, we employed the 4T1 tumor model and monitored in vivo the changes in the TME’s mechanical properties using ultrasound shear wave elastography (SWE) imaging. SWE measures tissue stiffness in real-time by generating a color map indicating different magnitudes of elastic modulus in kPa. SWE was performed before bacterial administration to establish baseline values, after 3 days of administration and at the end of the study (Fig. 2a, b). Interestingly, our findings demonstrate that the higher dose of phelp-HAse (107 CFUs) and both doses of C35R7-Hase (106 and 107 CFUs) can significantly decrease tumor stiffness, compared to the control group, as early as 3 days post-administration with the effect sustained even after 7 days, regardless of the dose employed (Fig. 2c). On the contrary, C35R9-HAse did not have any effect on tumor stiffness at either of the two doses administered in agreement with its inability to reduce HA levels, suggesting that either this promoter may be less effective at inducing Hase expression, or might exhibit a slow release of HAse, requiring more time to see an effect.
Given that alleviation of tumor stiffness is associated with improved perfusion across solid tumors 29, we investigated whether the bacteria could also improve vascular perfusion of 4T1 tumors. Tumor perfusion was assessed using contrast-enhanced ultrasound (CEUS) following a bolus injection of contrast agents (microbubbles). CEUS was employed twice: on the day of bacterial administration and at the study endpoint (Fig. 2a). Analysis of the ultrasound images of contrast agents at the time of peak intensity indicates that only the high concentration of phelp-HAse (107 CFUs) can enhance perfusion compared to the untreated group, while C35R7-Hase is beneficial at any of the doses (106 and 107 CFUs) employed (Fig. 2d,e). Moreover, we confirmed the existence of a linear correlation between tumor stiffness and perfusion values as measured at the end of the study consistent with previous studies (Fig. 2f)10,29. Considering its limited toxicity and its ability to effectively reduce HA levels and remodel the TME, we concluded that C35R7 at a dose of 106 CFUs is the ideal live biotherapeutic to continue our research with.
C35R7 enhances immunotherapy efficacy in triple-negative breast cancer murine models
The accumulation of HA in the TME serves as a physical barrier, compromising the delivery of ICIs and restricting access of immune cells to the tumor interior, highlighting the significance of ECM targeting as a potential approach to enhance the efficacy of immunotherapy. To address this, we conducted combination studies involving bacterial-mediated enzymatic stromal remodeling using the C35R7-Hase candidate and an antibody cocktail consisting of anti-PD-1 (10 mg/kg) and anti-CTLA-4 (5 mg/kg) in two orthotopic syngeneic models of breast cancer, namely 4T1 and E0771, which are known to be refractory to ICI26,27,30-32. 4T1 and E0771 tumor-bearing mice received the bacterial treatment only once when tumors reached an average size of 150 mm3. Given that bacteria can reduce stiffness as early as 3 days post-administration, the first cycle of the immunotherapy cocktail was administered 3 and 4 days post-bacterial treatment in 4T1 and E0771 studies, respectively. In total, mice received three cycles of the immunotherapy cocktail (via intraperitoneal injection) with a three-day interval between each cycle. Interestingly, for the 4T1 study, only mice treated with the combination therapy experienced tumor regression, while ICI monotherapy had no effect on tumor growth (Fig. 3a,c and Supplementary Fig. S4a). In the E0771 study, ICI alone had a modest but significant antitumor effect consisting of an initial plateauing of the tumor growth rate, which eventually resumed (p=0.036), while the combination of C35R7 with ICI exhibited a more potent antitumor response (p=0.0008), with 5 out of 10 mice in that group experiencing complete tumor regression (Fig. 3b, d and Supplementary Fig. S4b,c). To validate that the enhanced antitumor responses were attributed to the remodeling of the TME, we performed SWE and CEUS, before and 3 or 4 days after bacterial treatment, as described earlier. Consistent with our previous results, we confirmed that the groups receiving the bacteria had reduced intratumoral stiffness levels (Fig. 3e, f) and increased vascular perfusion compared to the tumors of groups treated with either ICI monotherapy or left untreated (Fig. 3g).
Combination treatment prolongs survival in triple-negative breast cancer models
The effectiveness of the engineered bacteria to boost immunotherapy efficacy was further supported by survival studies (Fig. 4a,b). To mimic the clinical scenario, tumors were surgically removed 3 days after the last ICI cycle and mice survival was closely monitored. For the 4T1 tumor model, our data demonstrate that 1 out of 9 mice treated with C35R7 + ICI achieved complete remission (Fig. 4a). For the E0771 study, mice of the untreated group were the first to die followed by the mice administered with only the bacteria 10 days apart, although the survival advantage noted was not statistically significant. One mouse out of 9 treated with ICI monotherapy experienced complete survival, while all mice of the combination treatment were cured (Fig. 4b, c). Survivors of both ICI monotherapy and the combination group were rechallenged with a second inoculation of E0771 cells and their growth curves were compared against 8 healthy control mice of the same age. Remarkably, tumors developed in all control mice (Fig. 4d) but none in the immunotherapy and combination-treated mice that survived the initial experiment (Fig. 4e, f), indicating the acquisition of durable long-term antitumor response and immunological memory to E0771 cancer cell antigens.
Programmable bacteria combined with ICI immunostimulates the TME
Given the pivotal role of the tumor immune microenvironment in determining sensitivity to ICI treatment, we aimed to examine the immune milieu of tumors following various treatments. Flow cytometry analysis revealed a reduction in the myeloid-derived suppressor cell (MDSC, CD45+GR1+CD11b+) population in both 4T1 and E0771 tumors treated with ICI alone or the combination therapy, whereas the levels of total lymphocyte remained unchanged across the different treatment groups (Fig. 5a, b, h, i). Consistent with this finding, the increase in the percentage of tumor-associated macrophages (TAMs, CD45+GR1-CD11b+F4/80+) and the elevation in the ratio of M1-like TAMs to M2-like TAMs, in the TME of E0771 tumors treated with ICI combination further suggests a shift from an immunosuppressive to an immunosupportive microenvironment (Fig. 5j, k). Furthermore, E0771-bearing mice treated with immunotherapy exhibited enhanced infiltration of helper and cytotoxic T cells (Fig. 5l, m), while the ratio of CD8+ T cells to immunosuppressive regulatory T cells (Tregs, TCRB+CD4SP+CD25hiCD127loFoxp3+) remained unaffected (Fig. 5n). The lack of statistical significance in the ICI combination group of the E0771 study is likely attributed to the small sample size. In the 4T1 study, we did not observe any changes in the abundance of TAMs or T cells, despite an increase in the CD8 T cell/Tregs ratio following ICI monotherapy. This lack of change is likely attributed to the fact that immunophenotyping of tumors was analyzed at the study endpoint, and might have missed detecting immune cell infiltration and activation during the earlier phase of the immune response. (Fig. 5c-g).
Antitumor potency of ICI combination therapy is dependent on baseline stiffness levels
Considering the positive impact of C35R7 live biotherapeutic in potentiating immunotherapy effectiveness in breast cancer, we wondered whether a less fibrotic tumor could respond similarly to the combined treatment. To address this, we employed the CT26 orthotopic colorectal cancer model by engrafting a small CT26 tumor fragment onto the caecum of mice, following established protocols 33. Mice were allowed to recover from surgery and then treated with a single i.v. injection of C35R7 at a concentration of 106 CFUs, as described earlier. Importantly, we observed that systemic administration of bacteria could effectively degrade HA levels in the CT26 tumor stroma without having any antitumor effects (Supplementary Fig. 6a-c), although the reduction noted was to a lesser extent compared to the 4T1 tumor model (Fig. 1). No significant adverse events including animal weight loss or splenomegaly were observed due to bacterial treatment (Supplementary Fig. 6d, e).
Having confirmed the enzymatic degradation of ECM in CT26 colon tumors, we next proceeded to ICI combination studies. Specifically, we employed a CT26 orthotopic tumor model stably expressing the luciferase reporter (CT26-Luc), allowing for real-time monitoring of tumor growth via bioluminescence imaging (BLI). Once the bioluminescence signal was detectable across all study mice (i.e., day 5 post-implantation), the bacterial treatment was administered. Then, 3 days after the C35R7 injection, mice received the first cycle of ICI cocktail, followed by two additional cycles. Photon flux (photons/sec) was used to quantify the bioluminescence signal, serving as an indicator of tumor progression. Notably, we observed a similar reduction in mean tumor burden in mice treated with the ICI alone or C35R7-Hase plus ICI combination (Fig. 6a-c). Individual BLI curves indicate that mice start responding to immunotherapy after the second cycle of treatment (day 11), with more mice in the combination group exhibiting a complete tumor regression compared to the monotherapy group (Fig. 6d-g), although not statistically significant. To assess whether the responsiveness of CT26 tumors to ICI is also reflected in the composition of the immune tumor microenvironment, we examined the prevalence of different immune cell populations following bacterial or ICI monotherapy, bacterial+ICI combination therapy, and in untreated mice. The intratumoral lymphocyte population remained unaffected across different treatment groups, consistent with our previous study in the triple-negative breast cancer models (Fig. 6h). Surprisingly, unlike the 4T1 and E0771 tumors, the levels of myeloid cells did not decrease after immunotherapy treatment, suggesting a tumor type-specific mechanism of action for ICI (Fig. 6i, j). T cell infiltration increased in groups receiving the ICI cocktail, but no change in CD8+ T cell to Tregs ratio was observed (Fig. 6k, l). Finally, we assessed the levels of antigen-presenting cells and confirmed their upregulation after ICI treatment (Fig. 6m). To support our hypothesis that ECM enzymatic degradation combined with ICI is more likely to benefit highly desmoplastic tumors refractory to ICI monotherapy, we assessed the tissue elasticity of CT26-luc tumors and compared them with the reported elastic moduli of 4T1 and E0771 tumors. This comparison was conducted through ex-vivo stress-strain experiments 34, as the in vivo measurement of CT26 tissue stiffness using SWE was not feasible with our ultrasound system. Interestingly, the ex-vivo elastic modulus measurements indicated that CT26 colon tumors are significantly softer compared to the triple-negative breast cancer models employed in the current study (Supplementary Fig. 7a). Immunofluorescence staining for the major ECM components HA and collagen confirms this finding (Supplementary Fig. 7b).