Tumor-seeking bacterial missiles

Abstract

In this exceptionally elegant and far-reaching study, the Arpaia lab at Columbia University engineer a probiotic Escherichia coli strain to eliminate cancer.¹

Working in two mouse models of cancer, CT26 colorectal cancer and B16F10 melanoma, Redenti et al.¹ start by identifying cancer-specific sequences. From these, they choose peptides containing (linked) MHCI and MHC II epitopes (about 25–30 amino acids long). They find that encoding a series of these in a plasmid, concatenated, but separated by five glycine-serine repeats, provides good expression of the neoantigens by E. coli.

They then turned to engineering EcN, a strain of E. coli isolated by Professor Alfred Nissle in 1917 from a young soldier resistant to infectious diarrhea.² Nissle marketed his discovery as a probiotic (Mutaflor®, still commercially available). Thus, Redenti et al. start with a safe, non-pathogenic E. coli, already extensively studied and widely used in humans (albeit orally).

Finding that their neoantigen plasmids express better in E. coli BL21 than in EcN, they set about modifying EcN to resemble BL21. Curing EcN of cryptic plasmids allows increased expression of the neoantigen-encoding plasmid. They then engineer deletions of two proteases: OmpT, which has roles in biofilm formation and the degradation of complement; and Lon which has pleiotropic roles within the bacterial cell, including oxygen-sensing.³ Hence, deleting these proteases, not only reduced degradation of the neoantigen peptides, but also attenuates EcN. Although not discussed by the authors, deletion of the Lon protease may impede the survival of EcN in normal tissues more than in the anoxic core of the tumor, further reducing potential off-target effects. Overall, the authors show that their engineered EcN bacteria show an 80-fold increase in expression of the neoantigen peptides compared with the original EcN, and a 1000-fold increased susceptibility to phagocytosis and clearance from the blood.

In another stroke of genius, the authors next insert the gene for Listeriolysin O (LLO), a pore-forming protein which allows Listeria to escape into the cytosol after phagocytosis. This has two benefits: improved loading of neoantigen epitopes on MHC class I, and skewing towards a T_H1 immune response after the sensing of intracytoplasmic bacteria.

The authors then turn to in vivo experiments to test their neoantigen-expressing, cryptic-plasmid cured, OmpT⁻, Lon⁻, LLO⁺ EcN.

In both of the tumor models, EcN injected i.v. could consistently be cultured from tumors (3–4 days after injection), but could not be cultured from any of the other tissues tested, including the tumor-draining lymph node (TdLN). Despite this, i.v. injection of the EcN bacteria in the CT26 model was shown to modulate the microenvironment in both the tumor and the TdLN.

In the TdLN, increased immunostimulatory capacity is implied by increased CD80/86 expression, and decreased PDL1 expression on classical type dendritic cells (both cDC1 and cDC2s). The percentage of cDC2 also increased. Within the tumor, treatment resulted in reduced frequencies of FoxP3⁺ regulatory T cells, as well as of PDL1⁺ neutrophils and macrophages, demonstrating alleviation of the immunosuppressive milieu. Increased IL-12 (typical of a Th1 response, and shown by the authors to be dependent on LLO expression) was also detected. These effects were independent of neoantigen expression and were also seen in response to bacteria expressing no, or irrelevant antigen. Whether this is due solely to bacterial PAMPs or whether the expression of (DNA damaging) colibactin by EcN potentiates these effects would be interesting to investigate.⁴

Neoantigen expression was, however, important for eliciting a tumor-specific T cell response: Tumor-infiltrating T cells collected 8 days after treatment more frequently showed IFNγ production in response to neoantigen peptides and could specifically kill the CT26 cell line in vitro. There is also a suggestion of epitope spreading beyond the 19 peptides originally targeted.

Turning to the B16F10 model, the authors demonstrate via in vivo depletion that both CD4⁺ and CD8⁺ T cells contribute to the anti-tumor response. After i.v. treatment with neoantigen-expressing EcN, intratumoral T cells increased in number, and showed increased proliferation, activation and killing capacity. Oddly, increases in the numbers and proportions of infiltrating cDC1, cDC2 and inflammatory monocytes, as well as increases in MHCII expression by these cells appear to be antigen dependent: they are not found after treatment with bacteria expressing irrelevant antigen. Repetition of these last experiments would be important.

Overall, the data paint a convincing picture: potent induction of the innate immune system by the attenuated intracellular bacteria sets the scene for priming of a strong antigen-specific T cell response, capable of eliminating both primary and metastatic cancers.

While the first reaction of many immunologists to injecting cancer patients with bacteria may be one of abject horror, there is precedence for this idea. Coley, 130 years ago, injected bacteria into cancer patients both i.t. and i.v. Live BCG (a strain of mycobacterium) infused into the bladder is a current, effective treatment for superficial bladder cancer. Our laboratory has injected a slow-release preparation of dead mycobacteria in advanced cancer patients.⁵

Although not widely appreciated, there is convincing evidence that two common cancer therapies, chemotherapy and checkpoint inhibitor therapies are also dependent on bacteria – in these cases, on the gut microbiota.^6-8 Both therapies are known to compromise gut integrity, and the efficacy of both is impaired by broad-spectrum antibiotics, which deplete gut bacteria. The trafficking of gut bacteria to the tumor after checkpoint inhibitor therapy was recently demonstrated by Choi et al.⁹ [and was first hypothesized in this journal by our laboratory!¹⁰]. Thus, treating cancer patients with injections of live bacteria, either i.t. or i.v. is a perfectly rational (and indeed inspired!) therapeutic concept, with considerably lower side effects than chemotherapy.

As human cancer patients all have different MHC alleles, and every cancer will have different mutations, translating Arpaia's protocol to humans will require a “personalized medicine” approach. This will necessitate MHC typing of each patient, biopsy and then sequencing of their tumor, computational identification of antigenic epitopes appropriate to the patient's MHCs, construction and transfection of the plasmid, and then injection of the personalized EcN into the patient. While the whole procedure will be labor-intensive and therefore expensive, it is feasible: sequencing of tumors and identification of “druggable” driver mutations is becoming more common, and similar neoantigen-identification strategies are currently under investigation for mRNA cancer vaccines.¹¹

Although Redenti et al.¹ ultimately focus on i.v. injections, their CT26 data suggest that i.t. delivery was more effective (and could cure both treated and distal tumors).¹ As the authors used the same dose for both treatment routes, the improved efficacy is likely a result of more bacteria reaching the tumor. With the majority of human cancers accessible to injection, and limitations on the number of bacteria deliverable i.v., the i.t. route should not be overlooked.

An extension of this idea might be to use “empty” EcN, or EcN engineered to express common cancer antigens able to bind promiscuously to human MHCs¹¹; either as a stand-alone treatment, or in combination with checkpoint therapies. Such a non-personalized approach would be especially useful when the time and/or cost to engineer personalized bacteria is prohibitive (e.g. in patients with poor prognoses, or for use in developing countries).

This remarkable publication provides proof of concept for a novel, imaginative and extremely clever approach to cancer treatment. It is time to add bacterial-based immunotherapies to the cancer immunotherapy arsenal!

George Cavic: Conceptualization; writing – original draft; writing – review and editing. Aude M Fahrer: Conceptualization; writing – original draft; writing – review and editing.

The authors declare no conflicts of interest.