Adeno-associated viruses (AAVs) are foundational gene delivery tools for basic science and clinical therapeutics. However, lack of mechanistic insight, especially for engineered vectors created by directed evolution, can hamper their application. Here, we adapt an unbiased human cell microarray platform to determine the extracellular and cell surface interactomes of natural and engineered AAVs. We identify a naturally-evolved and serotype-specific interaction between the AAV9 capsid and human interleukin 3 (IL3), with possible roles in host immune modulation, as well as lab-evolved low-density lipoprotein receptor-related protein 6 (LRP6) interactions specific to engineered capsids with enhanced blood-brain barrier crossing in non-human primates after intravenous administration. The unbiased cell microarray screening approach also allows us to identify off-target tissue binding interactions of engineered brain-enriched AAV capsids that may inform vectors' peripheral organ tropism and side effects. Our cryo-electron tomography and AlphaFold modeling of capsid-interactor complexes reveal LRP6 and IL3 binding sites. These results allow confident application of engineered AAVs in diverse organisms and unlock future target-informed engineering of improved viral and non-viral vectors for non-invasive therapeutic delivery to the brain.

Deep-learning-based methods for protein structure prediction have achieved unprecedented accuracy, yet their utility in the engineering of protein-based binders remains constrained due to a gap between the ability to predict the structures of candidate proteins and the ability toprioritize proteins by their potential to bind to a target. To bridge this gap, we introduce Automated Pairwise Peptide-Receptor Analysis for Screening Engineered proteins (APPRAISE), a method for predicting the target-binding propensity of engineered proteins. After generating structural models of engineered proteins competing for binding to a target using an established structure prediction tool such as AlphaFold-Multimer or ESMFold, APPRAISE performs a rapid (under 1 CPU second per model) scoring analysis that takes into account biophysical and geometrical constraints. As proof-of-concept cases, we demonstrate that APPRAISE can accurately classify receptor-dependent vs. receptor-independent adeno-associated viral vectors and diverse classes of engineered proteins such as miniproteins targeting the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike, nanobodies targeting a G-protein-coupled receptor, and peptides that specifically bind to transferrin receptor or programmed death-ligand 1 (PD-L1). APPRAISE is accessible through a web-based notebook interface using Google Colaboratory (https://tiny.cc/APPRAISE). With its accuracy, interpretability, and generalizability, APPRAISE promises to expand the utility of protein structure prediction and accelerate protein engineering for biomedical applications.

The blood-brain barrier (BBB) presents a major challenge for delivering large molecules to study and treat the central nervous system. This is due in part to the scarcity of targets known to mediate BBB crossing. To identify novel targets, we leverage a panel of adeno-associated viruses (AAVs) previously identified through mechanism-agnostic directed evolution for improved BBB transcytosis. Screening potential cognate receptors for enhanced BBB crossing, we identify two targets: murine-restricted LY6C1 and widely conserved carbonic anhydrase IV (CA-IV). We apply AlphaFold-based in silico methods to generate capsid-receptor binding models to predict the affinity of AAVs for these identified receptors. Demonstrating how these tools can unlock target-focused engineering strategies, we create an enhanced LY6C1-binding vector, AAV-PHP.eC, that, unlike our prior PHP.eB, also works in Ly6a-deficient mouse strains such as BALB/cJ. Combined with structural insights from computational modeling, the identification of primate-conserved CA-IV enables the design of more specific and potent human brain-penetrant chemicals and biologicals, including gene delivery vectors.

Delivering genes to and across the brain vasculature efficiently and specifically across species remains a critical challenge for addressing neurological diseases. We have evolved adeno-associated virus (AAV9) capsids into vectors that transduce brain endothelial cells specifically and efficiently following systemic administration in wild-type mice with diverse genetic backgrounds, and in rats. These AAVs also exhibit superior transduction of the CNS across non-human primates (marmosets and rhesus macaques), and in ex vivo human brain slices, although the endothelial tropism is not conserved across species. The capsid modifications translate from AAV9 to other serotypes such as AAV1 and AAV-DJ, enabling serotype switching for sequential AAV administration in mice. We demonstrate that the endothelial-specific mouse capsids can be used to genetically engineer the blood-brain barrier by transforming the mouse brain vasculature into a functional biofactory. We apply this approach to Hevin knockout mice, where AAV-X1-mediated ectopic expression of the synaptogenic protein Sparcl1/Hevin in brain endothelial cells rescued synaptic deficits.

Aim: We present multi-wavelength (MW) analytical ultracentrifugation (AUC) methods offering superior accuracy for adeno-associated virus characterization and quantification. Methods: Experimental design guidelines are presented for MW sedimentation velocity and analytical buoyant density equilibrium AUC. Results: Our results were compared with dual-wavelength AUC, transmission electron microscopy and mass photometry. In contrast to dual-wavelength AUC, MW-AUC correctly quantifies adeno-associated virus capsid ratios and identifies contaminants. In contrast to transmission electron microscopy, partially filled capsids can also be detected and quantified. In contrast to mass photometry, first-principle results are obtained. Conclusion: Our study demonstrates the improved information provided by MW-AUC, highlighting the utility of several recently integrated UltraScan programs, and reinforces AUC as the gold-standard analysis for viral vectors.

The blood-brain barrier (BBB) presents a major challenge to delivering large molecules to study and treat the central nervous system (CNS). This is due in part to the scarcity of effective targets for BBB crossing, the identification of which is the crucial first step of drug development. Here, we leveraged a panel of adeno-associated viruses (AAVs) previously identified through directed evolution for improved BBB transport to reverse engineer protein targets for enhanced BBB crossing. We identify both murine-restricted Ly6c1 and primate-conserved carbonic anhydrase IV (Car4; CA4) as novel receptors for crossing the BBB. We demonstrate how these receptors can unlock new experimental and computational target-focused engineering strategies by creating the enhanced Ly6c1-binding vector AAV-PHP.eC and by applying AlphaFold2-enabled in silico methods to rank capsids against identified receptors and generate capsid-receptor binding models. Here, with Car4, we add a completely new receptor to the very short list currently available for crossing the BBB in humans and, with Ly6c1, we validate a pipeline for receptor-targeted engineering. The identification of Car4/CA4 and structural insights from computational modeling provide new paths toward human brain-penetrant chemicals (drugs) and biologicals (including gene delivery).

Adeno-associated virus serotype 9 (AAV9) is a promising gene therapy vector for treating neurodegenerative diseases due to its ability to penetrate the blood-brain barrier. PHP.eB was engineered from AAV9 by insertion of a 7-amino acid peptide and point mutation of neighboring residues, thereby enhancing potency in the central nervous system. Here, we report a 2.24-Å resolution cryo-electron microscopy structure of PHP.eB, revealing conformational differences from other 7-mer insertion capsid variants. In PHP.eB, the 7-mer loop adopts a bent conformation, mediated by an interaction between engineered lysine and aspartate residues. Further, we identify PKD2 as the main AAV receptor (AAVR) domain recognizing both AAV9 and PHP.eB and find that the PHP.eB 7-mer partially destabilizes this interaction. Analysis of previously reported AAV structures together with our pull-down data demonstrate that the 7-mer topology determined by the lysine-aspartate interaction dictates AAVR binding strength. Our results suggest that PHP.eB's altered tropism may arise from both an additional interaction with LY6A and weakening of its AAVR interaction. Changing the insertion length, but not sequence, modifies PKD2 binding affinity, suggesting that a steric clash impedes AAVR binding. This research suggests improved library designs for future AAV selections to identify non-LY6A-dependent vectors and modulate AAVR interaction strength.

Adeno-associated viruses (AAVs) are typically single-stranded deoxyribonucleic acid (ssDNA) encapsulated within 25-nm protein capsids. Recently, tissue-specific AAV capsids (e.g. PHP.eB) have been shown to enhance brain delivery in rodents via the LY6A receptor on brain endothelial cells. Here, we create a non-invasive positron emission tomography (PET) methodology to track viruses. To provide the sensitivity required to track AAVs injected at picomolar levels, a unique multichelator construct labeled with a positron emitter (Cu-64, t 1/2 = 12.7 h) is coupled to the viral capsid. We find that brain accumulation of the PHP.eB capsid 1) exceeds that reported in any previous PET study of brain uptake of targeted therapies and 2) is correlated with optical reporter gene transduction of the brain. The PHP.eB capsid brain endothelial receptor affinity is nearly 20-fold greater than that of AAV9. The results suggest that novel PET imaging techniques can be applied to inform and optimize capsid design.

Recombinant adeno-associated viral (rAAV) capsids are increasingly used as gene delivery vectors in science and in therapeutics. However, there is room for further improvement on the efficiency and specificity to transduce the central nervous system (CNS) via non-invasive systemic delivery. We have shown in the past that marked improvement is possible using a Cre recombination-based AAV targeted evolution (CREATE) platform to identify AAV-PHP.B and AAV-PHP.eB capsids, which broadly transduce the CNS (Deverman et al, 2016; Chan et al, 2017). While CREATE was successful in identifying efficient CNS vectors in two rounds of evolution, the method was limited by its ability to identify only a few top candidates from the selection. To truly utilize the potential of a large selection design (such as a library of ~1.28 billion theoretical size using 7-mer-NNK mutagenesis strategy) across multiple different selection targets (such as endothelial cells, neurons and astrocytes), we developed Multiplexed-CREATE (M-CREATE). This method involves: (1) application of a positive selection pressure for on-target delivery and post-hoc negative selection against off-target delivery using next generation sequencing, (2) an unbiased selection design, and (3) a novel data analysis platform. M-CREATE identified distinct families of variants based on shared amino acid motifs. These include a family of PHP.B-like variants that appear to be dominant across in vivo selections for CNS transduction in Cre-transgenic adult mice. While some variants from this family showed similar tropism to PHP.B as expected, two variants (PHP.Vs) were distinct by exhibiting biased transduction to the vascular cells forming the blood-brain barrier (BBB). In addition, the new analysis platform allowed us to further mine the selection that yielded PHP.eB from its PHP.B parent (Chan et al, 2017) and therefore uncover PHP.N, that differs by 3 amino acids from PHP.B and shows greater specificity in transducing neurons over other CNS cell-types. This further demonstrates that small sequence variations within a family can yield different tropisms. Recent reports on the non-permissibility of PHP.B or PHP.eB in BALB/cJ (Hordeaux et al, 2018; Matsuzaki et al, 2019), and the subsequent identification of LY6A as the receptor underlying the improved BBB entry (Hordeaux et al, 2019; Huang et al, 2019; Batista et al, 2019) has motivated us to investigate the non-dominant families with distinct amino acid signatures in our 7-mer NNK library. We identified several non-dominant families (PHP.Cs), exhibiting better transduction of the CNS compared to parent AAV9 while also being efficient in crossing the BBB across mouse strains, including BALB/cJ. This suggests that variants with different amino acid motifs from the PHP.C-like families may have different mechanisms of BBB transmission, thereby making them promising candidates for further study toward translation of AAV vectors across strains and species.

Adeno-associated viruses (AAVs) are widely used gene delivery vectors due to their ability to transduce dividing and non-dividing cells, their long-term persistence, and low immunogenicity. However, natural AAV serotypes have a limited set of tropisms. Directed evolution has been used to engineer recombinant AAVs to target specific cell types and tissues, leveraging next generation sequencing data. The deluge of data from these deep sequencing experiments has brought about data management and analysis challenges, for which there are no current commercially available solutions. Furthermore, classical approaches to analyzing data from directed evolution heavily involves manual inspection, and often overlooks patterns present in the larger datasets. To address these challenges, we developed robust cloud-based software that provides central management for next generation sequencing data, extracts variants, performs structural modeling, and can be extended to incorporate machine learning models to make predictions for variants with specific properties. The software is composed of a set of interconnected discrete components: a modern web user interface implemented in JavaScript with React, a relational database, a distributed task queue, task workers, and a Django-based API. This architecture allows computationally intensive tasks such as alignments, structural modeling, and machine learning to scale from a single machine to hundreds of machines, with minimal configuration. The software automatically imports and manages sequencing data from several different commercial and in-house sequencing providers. When the data is imported, sequence quality metrics are automatically generated and presented to the user. Variants are extracted by performing pairwise alignments between the natural serotype and the sequencing reads. The variants are further encoded into embeddings, grouped into families, and are analyzed for prevalent sequence motifs. We use the Rosetta software libraries to perform comparative modeling simulations on selected variants. Finally, we are developing and have extension support for Pytorch-based machine learning models to generate novel variants with desirable properties as well as to select candidate variants for additional rounds of optimization and characterization. This software represents a general tool for simple, scalable, and centralized analyses of next generation sequencing data for protein engineering by directed evolution, and could be generalized for all projects with large-scale deep sequencing datasets in the future.

Recombinant adeno-associated viruses (rAAVs) are efficient gene delivery vectors via intravenous delivery; however, natural serotypes display a finite set of tropisms. To expand their utility, we evolved AAV capsids to efficiently transduce specific cell types in adult mouse brains. Building upon our Cre-recombination-based AAV targeted evolution (CREATE) platform, we developed Multiplexed-CREATE (M-CREATE) to identify variants of interest in a given selection landscape through multiple positive and negative selection criteria. M-CREATE incorporates next-generation sequencing, synthetic library generation and a dedicated analysis pipeline. We have identified capsid variants that can transduce the central nervous system broadly, exhibit bias toward vascular cells and astrocytes, target neurons with greater specificity or cross the blood–brain barrier across diverse murine strains. Collectively, the M-CREATE methodology accelerates the discovery of capsids for use in neuroscience and gene-therapy applications. M-CREATE is an in vivo screening strategy for identifying recombinant AAVs with desired tropism. The approach involves both positive and negative selection and yields vectors with diversified cell-type tropism that can cross the blood–brain barrier in adult mice across strains when delivered intravenously.

Neoantigen-specific T cells are increasingly viewed as important immunotherapy effectors, but physically isolating these rare cell populations is challenging. Here, we describe a sensitive method for the enumeration and isolation of neoantigen-specific CD8+ T cells from small samples of patient tumor or blood. The method relies on magnetic nanoparticles that present neoantigen-loaded major histocompatibility complex (MHC) tetramers at high avidity by barcoded DNA linkers. The magnetic particles provide a convenient handle to isolate the desired cell populations, and the barcoded DNA enables multiplexed analysis. The method exhibits superior recovery of antigen-specific T cell populations relative to literature approaches. We applied the method to profile neoantigen-specific T cell populations in the tumor and blood of patients with metastatic melanoma over the course of anti-PD1 checkpoint inhibitor therapy. We show that the method has value for monitoring clinical responses to cancer immunotherapy and might help guide the development of personalized mutational neoantigen-specific T cell therapies and cancer vaccines.

Integral membrane proteins (MPs) are key engineering targets due to their critical roles in regulating cell function. In engineering MPs, it can be extremely challenging to retain membrane localization capability while changing other desired properties. We have used structure-guided SCHEMA recombination to create a large set of functionally diverse chimeras from three sequence-diverse channel-rhodopsins (ChRs). We chose 218 ChR chimeras from two SCHEMA libraries and assayed them for expression and plasma membrane localization in human embryonic kidney cells. The majority of the chimeras express, with 89% of the tested chimeras outperforming the lowest-expressing parent; 12% of the tested chimeras express at even higher levels than any of the parents. A significant fraction (23%) also localize to the membrane better than the lowest-performing parent ChR. Most (93%) of these well-localizing chimeras are also functional light-gated channels. Many chimeras have stronger light-activated inward currents than the three parents, and some have unique off-kinetics and spectral properties relative to the parents. An effective method for generating protein sequence and functional diversity, SCHEMA recombination can be used to gain insights into sequence-function relationships in MPs.