Illuminating protein space with a programmable generative model.

Three billion years of evolution has produced a tremendous diversity of protein molecules1, but the full potential of proteins is likely to be much greater. Accessing this potential has been challenging for both computation and experiments because the space of possible protein molecules is much larger than the space of those likely to have functions. Here we introduce Chroma, a generative model for proteins and protein complexes that can directly sample novel protein structures and sequences, and that can be conditioned to steer the generative process towards desired properties and functions. To enable this, we introduce a diffusion process that respects the conformational statistics of polymer ensembles, an efficient neural architecture for molecular systems that enables long-range reasoning with sub-quadratic scaling, layers for efficiently synthesizing three-dimensional structures of proteins from predicted inter-residue geometries and a general low-temperature sampling algorithm for diffusion models. Chroma achieves protein design as Bayesian inference under external constraints, which can involve symmetries, substructure, shape, semantics and even natural-language prompts. The experimental characterization of 310 proteins shows that sampling from Chroma results in proteins that are highly expressed, fold and have favourable biophysical properties. The crystal structures of two designed proteins exhibit atomistic agreement with Chroma samples (a backbone root-mean-square deviation of around 1.0 Å). With this unified approach to protein design, we hope to accelerate the programming of protein matter to benefit human health, materials science and synthetic biology.

Tumor-localized catalases can fail to alter tumor growth and transcriptional profiles in subcutaneous syngeneic mouse tumor models.

Catalase is an antioxidant enzyme that catalyzes the rapid conversion of hydrogen peroxide to water and oxygen. Use of catalase as a cancer therapeutic has been proposed to reduce oxidative stress and hypoxia in the tumor microenvironment, both activities which are hypothesized to reduce tumor growth. Furthermore, exposing murine tumors to exogenous catalase was previously reported to have therapeutic benefit. We studied the therapeutic effect of tumor-localized catalases with the aim to further elucidate the mechanism of action. To do this, we engineered two approaches to maximize intratumoral catalase exposure: 1) an injected extracellular catalase with enhanced tumor retention, and 2) tumor cell lines that over-express intracellular catalase. Both approaches were characterized for functionality and tested for therapeutic efficacy and mechanism in 4T1 and CT26 murine syngeneic tumor models. The injected catalase was confirmed to have enzyme activity >30,000 U/mg and was retained at the injection site for more than one week in vivo. The engineered cell lines exhibited increased catalase activity and antioxidant capacity, with catalase over-expression that was maintained for at least one week after gene expression was induced in vivo. We did not observe a significant difference in tumor growth or survival between catalase-treated and untreated mice when either approach was used. Finally, bulk RNA sequencing of tumors was performed, comparing the gene expression of catalase-treated and untreated tumors. Gene expression analysis revealed very few differentially expressed genes as a result of exposure to catalase and notably, we did not observe changes consistent with an altered state of hypoxia or oxidative stress. In conclusion, we observe that sustained intratumoral catalase neither has therapeutic benefit nor triggers significant differential expression of genes associated with the anticipated therapeutic mechanism in the subcutaneous syngeneic tumor models used. Given the lack of effect observed, we propose that further development of catalase as a cancer therapeutic should take these findings into consideration.

Alum-anchored intratumoral retention improves the tolerability and antitumor efficacy of type I interferon therapies.

Effective antitumor immunity in mice requires activation of the type I interferon (IFN) response pathway. IFNα and IFNß therapies have proven promising in humans, but suffer from limited efficacy and high toxicity. Intratumoral IFN retention ameliorates systemic toxicity, but given the complexity of IFN signaling, it was unclear whether long-term intratumoral retention of type I IFNs would promote or inhibit antitumor responses. To this end, we compared the efficacy of IFNα and IFNß that exhibit either brief or sustained retention after intratumoral injection in syngeneic mouse tumor models. Significant enhancement in tumor retention, mediated by anchoring these IFNs to coinjected aluminum-hydroxide (alum) particles, greatly improved both their tolerability and efficacy. The improved efficacy of alum-anchored IFNs could be attributed to sustained pleiotropic effects on tumor cells, immune cells, and nonhematopoietic cells. Alum-anchored IFNs achieved high cure rates of B16F10 tumors upon combination with either anti-PD-1 antibody or interleukin-2. Interestingly however, these alternative combination immunotherapies yielded disparate T cell phenotypes and differential resistance to tumor rechallenge, highlighting important distinctions in adaptive memory formation for combinations of type I IFNs with other immunotherapies.

An affinity threshold for maximum efficacy in anti-PD-1 immunotherapy.

Monoclonal antibodies targeting the programmed cell death protein 1 (PD-1) remain the most prevalent cancer immunotherapy both as a monotherapy and in combination with additional therapies. Despite the extensive success of anti-PD-1 monoclonal antibodies in the clinic, the experimental relationship between binding affinity and functional potency for anti-PD-1 antibodies in vivo has not been reported. Anti-PD-1 antibodies with higher and lower affinity than nivolumab or pembrolizumab are entering the clinic and show varied preclinical efficacy. Here, we explore the role of broad-ranging affinity variation within a single lineage in a syngeneic immunocompetent mouse model. By developing a panel of murine anti-PD-1 antibodies with varying affinity (ranging from KD = 20 pM - 15 nM), we find that there is a threshold affinity required for maximum efficacy at a given dose in the treatment of the MC38 adenocarcinoma model with anti-PD-1 immunotherapy. Physiologically based pharmacokinetic modeling complements interpretation of the experimental results and highlights the direct relationship between dose, affinity, and PD-1 target saturation in the tumor.