Modular cytosine base editing promotes epigenomic and genomic modifications.

Prokaryotic and eukaryotic adaptive immunity differ considerably. Yet, their fundamental mechanisms of gene editing via Cas9 and activation-induced deaminase (AID), respectively, can be conveniently complimentary. Cas9 is an RNA targeted dual nuclease expressed in several bacterial species. AID is a cytosine deaminase expressed in germinal centre B cells to mediate genomic antibody diversification. AID can also mediate epigenomic reprogramming via active DNA demethylation. It is known that sequence motifs, nucleic acid structures, and associated co-factors affect AID activity. But despite repeated attempts, deciphering AID's intrinsic catalytic activities and harnessing its targeted recruitment to DNA is still intractable. Even recent cytosine base editors are unable to fully recapitulate AID's genomic and epigenomic editing properties. Here, we describe the first instance of a modular AID-based editor that recapitulates the full spectrum of genomic and epigenomic editing activity. Our 'Swiss army knife' toolbox will help better understand AID biology per se as well as improve targeted genomic and epigenomic editing.

Role of EXO1 nuclease activity in genome maintenance, the immune response and tumor suppression in Exo1D173A mice.

DNA damage response pathways rely extensively on nuclease activity to process DNA intermediates. Exonuclease 1 (EXO1) is a pleiotropic evolutionary conserved DNA exonuclease involved in various DNA repair pathways, replication, antibody diversification, and meiosis. But, whether EXO1 facilitates these DNA metabolic processes through its enzymatic or scaffolding functions remains unclear. Here, we dissect the contribution of EXO1 enzymatic versus scaffolding activity by comparing Exo1DA/DA mice expressing a proven nuclease-dead mutant form of EXO1 to entirely EXO1-deficient Exo1-/- and EXO1 wild type Exo1+/+ mice. We show that Exo1DA/DA and Exo1-/- mice are compromised in canonical DNA repair processing, suggesting that the EXO1 enzymatic role is important for error-free DNA mismatch and double-strand break repair pathways. However, in non-canonical repair pathways, EXO1 appears to have a more nuanced function. Next-generation sequencing of heavy chain V region in B cells showed the mutation spectra of Exo1DA/DA mice to be intermediate between Exo1+/+ and Exo1-/- mice, suggesting that both catalytic and scaffolding roles of EXO1 are important for somatic hypermutation. Similarly, while overall class switch recombination in Exo1DA/DA and Exo1-/- mice was comparably defective, switch junction analysis suggests that EXO1 might fulfill an additional scaffolding function downstream of class switching. In contrast to Exo1-/- mice that are infertile, meiosis progressed normally in Exo1DA/DA and Exo1+/+ cohorts, indicating that a structural but not the nuclease function of EXO1 is critical for meiosis. However, both Exo1DA/DA and Exo1-/- mice displayed similar mortality and cancer predisposition profiles. Taken together, these data demonstrate that EXO1 has both scaffolding and enzymatic functions in distinct DNA repair processes and suggest a more composite and intricate role for EXO1 in DNA metabolic processes and disease.

Role of Dot1L and H3K79 methylation in regulating somatic hypermutation of immunoglobulin genes.

Somatic hypermutation (SHM) and class-switch recombination (CSR) of the immunoglobulin (Ig) genes allow B cells to make antibodies that protect us against a wide variety of pathogens. SHM is mediated by activation-induced deaminase (AID), occurs at a million times higher frequency than other mutations in the mammalian genome, and is largely restricted to the variable (V) and switch (S) regions of Ig genes. Using the Ramos human Burkitt's lymphoma cell line, we find that H3K79me2/3 and its methyltransferase Dot1L are more abundant on the V region than on the constant (C) region, which does not undergo mutation. In primary naïve mouse B cells examined ex vivo, the H3K79me2/3 modification appears constitutively in the donor Sµ and is inducible in the recipient Sγ1 upon CSR stimulation. Knockout and inhibition of Dot1L in Ramos cells significantly reduces V region mutation and the abundance of H3K79me2/3 on the V region and is associated with a decrease of polymerase II (Pol II) and its S2 phosphorylated form at the IgH locus. Knockout of Dot1L also decreases the abundance of BRD4 and CDK9 (a subunit of the P-TEFb complex) on the V region, and this is accompanied by decreased nascent transcripts throughout the IgH gene. Treatment with JQ1 (inhibitor of BRD4) or DRB (inhibitor of CDK9) decreases SHM and the abundance of Pol II S2P at the IgH locus. Since all these factors play a role in transcription elongation, our studies reinforce the idea that the chromatin context and dynamics of transcription are critical for SHM.

The role of HIRA-dependent H3.3 deposition and its modifications in the somatic hypermutation of immunoglobulin variable regions.

The H3.3 histone variant and its chaperone HIRA are involved in active transcription, but their detailed roles in regulating somatic hypermutation (SHM) of immunoglobulin variable regions in human B cells are not yet fully understood. In this study, we show that the knockout (KO) of HIRA significantly decreased SHM and changed the mutation pattern of the variable region of the immunoglobulin heavy chain (IgH) in the human Ramos B cell line without changing the levels of activation-induced deaminase and other major proteins known to be involved in SHM. Except for H3K79me2/3 and Spt5, many factors related to active transcription, including H3.3, were substantively decreased in HIRA KO cells, and this was accompanied by decreased nascent transcription in the IgH locus. The abundance of ZMYND11 that specifically binds to H3.3K36me3 on the IgH locus was also reduced in the HIRA KO. Somewhat surprisingly, HIRA loss increased the chromatin accessibility of the IgH V region locus. Furthermore, stable expression of ectopic H3.3G34V and H3.3G34R mutants that inhibit both the trimethylation of H3.3K36 and the recruitment of ZMYND11 significantly reduced SHM in Ramos cells, while the H3.3K79M did not. Consistent with the HIRA KO, the H3.3G34V mutant also decreased the occupancy of various elongation factors and of ZMYND11 on the IgH variable and downstream switching regions. Our results reveal an unrecognized role of HIRA and the H3.3K36me3 modification in SHM and extend our knowledge of how transcription-associated chromatin structure and accessibility contribute to SHM in human B cells.

Mutational pressure by host APOBEC3s more strongly affects genes expressed early in the lytic phase of herpes simplex virus-1 (HSV-1) and human polyomavirus (HPyV) infection.

Herpes-Simplex Virus 1 (HSV-1) infects most humans when they are young, sometimes with fatal consequences. Gene expression occurs in a temporal order upon lytic HSV-1 infection: immediate early (IE) genes are expressed, then early (E) genes, followed by late (L) genes. During this infection cycle, the HSV-1 genome has the potential for exposure to APOBEC3 (A3) proteins, a family of cytidine deaminases that cause C>U mutations on single-stranded DNA (ssDNA), often resulting in a C>T transition. We developed a computational model for the mutational pressure of A3 on the lytic cycle of HSV-1 to determine which viral kinetic gene class is most vulnerable to A3 mutations. Using in silico stochastic methods, we simulated the infectious cycle under varying intensities of A3 mutational pressure. We found that the IE and E genes are more vulnerable to A3 than L genes. We validated this model by analyzing the A3 evolutionary footprints in 25 HSV-1 isolates. We find that IE and E genes have evolved to underrepresent A3 hotspot motifs more so than L genes, consistent with greater selection pressure on IE and E genes. We extend this model to two-step infections, such as those of polyomavirus, and find that the same pattern holds for over 25 human Polyomavirus (HPyVs) genomes. Genes expressed earlier during infection are more vulnerable to mutations than those expressed later.

STAT3 imparts BRCAness by impairing homologous recombination repair in Epstein-Barr virus-transformed B lymphocytes.

Epstein-Barr virus (EBV) causes lymphomas and epithelial cell cancers. Though generally silent in B lymphocytes, this widely prevalent virus can cause endemic Burkitt lymphoma and post-transplant lymphoproliferative disorders/lymphomas in immunocompromised hosts. By learning how EBV breaches barriers to cell proliferation, we hope to undermine those strategies to treat EBV lymphomas and potentially other cancers. We had previously found that EBV, through activation of cellular STAT3 prevents phosphorylation of Chk1, and thereby, suppresses activation of the intra-S phase cell-cycle checkpoint, a potent barrier to oncogene-driven proliferation. This observation prompted us to examine the consequences on DNA repair since homologous recombination repair, the most error-free form, requires phosphoChk1. We now report that the defect in Chk1 phosphorylation also curtails RAD51 nucleation, and thereby, homologous recombination repair of DNA double strand breaks. The resulting reliance on error-prone microhomology-mediated end-joining (MMEJ) repair makes EBV-transformed cells susceptible to PARP inhibition and simultaneous accrual of genome-wide deletions and insertions resulting from synthesis-dependent MMEJ. Analysis of transcriptomic and drug susceptibility data from hundreds of cancer lines reveals a STAT3-dependent gene-set predictive of susceptibility of cancers to synthetic lethal PARP inhibition. These findings i) demonstrate how the tumor virus EBV re-shapes cellular DNA repair, ii) provide the first genome-wide evidence for insertions resulting from MMEJ in human cells, and iii) expand the range of cancers (EBV-related and -unrelated) that are likely to respond to synthetic lethal inhibitors given the high prevalence of cancers with constitutively active STAT3.

A Bayesian model based computational analysis of the relationship between bisulfite accessible single-stranded DNA in chromatin and somatic hypermutation of immunoglobulin genes.

The B cells in our body generate protective antibodies by introducing somatic hypermutations (SHM) into the variable region of immunoglobulin genes (IgVs). The mutations are generated by activation induced deaminase (AID) that converts cytosine to uracil in single stranded DNA (ssDNA) generated during transcription. Attempts have been made to correlate SHM with ssDNA using bisulfite to chemically convert cytosines that are accessible in the intact chromatin of mutating B cells. These studies have been complicated by using different definitions of "bisulfite accessible regions" (BARs). Recently, deep-sequencing has provided much larger datasets of such regions but computational methods are needed to enable this analysis. Here we leveraged the deep-sequencing approach with unique molecular identifiers and developed a novel Hidden Markov Model based Bayesian Segmentation algorithm to characterize the ssDNA regions in the IGHV4-34 gene of the human Ramos B cell line. Combining hierarchical clustering and our new Bayesian model, we identified recurrent BARs in certain subregions of both top and bottom strands of this gene. Using this new system, the average size of BARs is about 15 bp. We also identified potential G-quadruplex DNA structures in this gene and found that the BARs co-locate with G-quadruplex structures in the opposite strand. Using various correlation analyses, there is not a direct site-to-site relationship between the bisulfite accessible ssDNA and all sites of SHM but most of the highly AID mutated sites are within 15 bp of a BAR. In summary, we developed a novel platform to study single stranded DNA in chromatin at a base pair resolution that reveals potential relationships among BARs, SHM and G-quadruplexes. This platform could be applied to genome wide studies in the future.

PD-1/PD-L1 immune checkpoint and p53 loss facilitate tumor progression in activated B-cell diffuse large B-cell lymphomas.

Refractory or relapsed diffuse large B-cell lymphoma (DLBCL) often associates with the activated B-cell-like (ABC) subtype and genetic alterations that drive constitutive NF-κB activation and impair B-cell terminal differentiation. Here, we show that DNA damage response by p53 is a central mechanism suppressing the pathogenic cooperation of IKK2ca-enforced canonical NF-κB and impaired differentiation resulting from Blimp1 loss in ABC-DLBCL lymphomagenesis. We provide evidences that the interplay between these genetic alterations and the tumor microenvironment select for additional molecular addictions that promote lymphoma progression, including aberrant coexpression of FOXP1 and the B-cell mutagenic enzyme activation-induced deaminase, and immune evasion through major histocompatibility complex class II downregulation, PD-L1 upregulation, and T-cell exhaustion. Consistently, PD-1 blockade cooperated with anti-CD20-mediated B-cell cytotoxicity, promoting extended T-cell reactivation and antitumor specificity that improved long-term overall survival in mice. Our data support a pathogenic cooperation among NF-κB-driven prosurvival, genetic instability, and immune evasion mechanisms in DLBCL and provide preclinical proof of concept for including PD-1/PD-L1 blockade in combinatorial immunotherapy for ABC-DLBCL.

A-to-I editing of Malacoherpesviridae RNAs supports the antiviral role of ADAR1 in mollusks.

BACKGROUND: Adenosine deaminase enzymes of the ADAR family are conserved in metazoans. They convert adenine into inosine in dsRNAs and thus alter both structural properties and the coding potential of their substrates. Acting on exogenous dsRNAs, ADAR1 exerts a pro- or anti-viral role in vertebrates and Drosophila. RESULTS: We traced 4 ADAR homologs in 14 lophotrochozoan genomes and we classified them into ADAD, ADAR1 or ADAR2, based on phylogenetic and structural analyses of the enzymatic domain. Using RNA-seq and quantitative real time PCR we demonstrated the upregulation of one ADAR1 homolog in the bivalve Crassostrea gigas and in the gastropod Haliotis diversicolor supertexta during Ostreid herpesvirus-1 or Haliotid herpesvirus-1 infection. Accordingly, we demonstrated an extensive ADAR-mediated editing of viral RNAs. Single nucleotide variation (SNV) profiles obtained by pairing RNA- and DNA-seq data from the viral infected individuals resulted to be mostly compatible with ADAR-mediated A-to-I editing (up to 97%). SNVs occurred at low frequency in genomic hotspots, denoted by the overlapping of viral genes encoded on opposite DNA strands. The SNV sites and their upstream neighbor nucleotide indicated the targeting of selected adenosines. The analysis of viral sequences suggested that, under the pressure of the ADAR editing, the two Malacoherpesviridae genomes have evolved to reduce the number of deamination targets. CONCLUSIONS: We report, for the first time, evidence of an extensive editing of Malacoherpesviridae RNAs attributable to host ADAR1 enzymes. The analysis of base neighbor preferences, structural features and expression profiles of molluscan ADAR1 supports the conservation of the enzyme function among metazoans and further suggested that ADAR1 exerts an antiviral role in mollusks.

Correction to: The cytidine deaminase under-representation reporter (CDUR) as a tool to study evolution of sequences under deaminase mutational pressure.

Following publication of the original article [1], the authors reported that Figs. 1 and 3 were interchanged. The original article has been corrected.

The cytidine deaminase under-representation reporter (CDUR) as a tool to study evolution of sequences under deaminase mutational pressure.

BACKGROUND: Activation induced deaminase (AID) and apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like 3 (APOBEC3) are deaminases that mutate C to U on single-stranded DNA (ssDNA). AID is expressed primarily in germinal center B-cells, where it facilitates affinity maturation and class-switch recombination. APOBEC3 are a family of anti-viral proteins that act as part of the intrinsic immune response. In both cases, there are particular sequence motifs, also known as "mutation motifs", to which these deaminases prefer to bind and mutate. RESULTS: We present a program, the cytidine deaminase under-representation reporter (CDUR) designed to statistically determine whether a given sequence has an under/over-representation of these mutation motifs. CDUR shows consitency with other studies of mutation motifs, as we show by analyzing sequences from the adeno-associated virus 2 (AAV2) and human papillomavirus (HPV). CONCLUSION: Using various shuffling mechanisms to generate different null model distributions, we can tailor CDUR to correct for metrics such as GC-content, dinucleotide frequency, and codon bias.

The preferred nucleotide contexts of the AID/APOBEC cytidine deaminases have differential effects when mutating retrotransposon and virus sequences compared to host genes.

The AID / APOBEC genes are a family of cytidine deaminases that have evolved in vertebrates, and particularly mammals, to mutate RNA and DNA at distinct preferred nucleotide contexts (or "hotspots") on foreign genomes such as viruses and retrotransposons. These enzymes play a pivotal role in intrinsic immunity defense mechanisms, often deleteriously mutating invading retroviruses or retrotransposons and, in the case of AID, changing antibody sequences to drive affinity maturation. We investigate the strength of various hotspots on their known biological targets by evaluating the potential impact of mutations on the DNA coding sequences of these targets, and compare these results to hypothetical hotspots that did not evolve. We find that the existing AID / APOBEC hotspots have a large impact on retrotransposons and non-mammalian viruses while having a much smaller effect on vital mammalian genes, suggesting co-evolution with AID / APOBECs may have had an impact on the genomes of the viruses we analyzed. We determine that GC content appears to be a significant, but not sole, factor in resistance to deaminase activity. We discuss possible mechanisms AID and APOBEC viral targets have adopted to escape the impacts of deamination activity, including changing the GC content of the genome.

Overlapping hotspots in CDRs are critical sites for V region diversification.

Activation-induced deaminase (AID) mediates the somatic hypermutation (SHM) of Ig variable (V) regions that is required for the affinity maturation of the antibody response. An intensive analysis of a published database of somatic hypermutations that arose in the IGHV3-23*01 human V region expressed in vivo by human memory B cells revealed that the focus of mutations in complementary determining region (CDR)1 and CDR2 coincided with a combination of overlapping AGCT hotspots, the absence of AID cold spots, and an abundance of polymerase eta hotspots. If the overlapping hotspots in the CDR1 or CDR2 did not undergo mutation, the frequency of mutations throughout the V region was reduced. To model this result, we examined the mutation of the human IGHV3-23*01 biochemically and in the endogenous heavy chain locus of Ramos B cells. Deep sequencing revealed that IGHV3-23*01 in Ramos cells accumulates AID-induced mutations primarily in the AGCT in CDR2, which was also the most frequent site of mutation in vivo. Replacing the overlapping hotspots in CDR1 and CDR2 with neutral or cold motifs resulted in a reduction in mutations within the modified motifs and, to some degree, throughout the V region. In addition, some of the overlapping hotspots in the CDRs were at sites in which replacement mutations could change the structure of the CDR loops. Our analysis suggests that the local sequence environment of the V region, and especially of the CDR1 and CDR2, is highly evolved to recruit mutations to key residues in the CDRs of the IgV region.

Potential for evolution of complex defense strategies in a multi-scale model of virus-host coevolution.

BACKGROUND: Host resistance and viral pathogenicity are determined by molecular interactions that are part of the evolutionary arms race between viruses and their hosts. Viruses are obligate intracellular parasites and entry to the host cell is the first step of any virus infection. Commonly, viruses enter host cells by binding cell surface receptors. We adopt a computational modeling approach to study the evolution of the first infection step, where we consider two possible levels of resistance mechanism: at the level of the binding interaction between the host receptor and a virus binding protein, and at the level of receptor protein expression where we use a standard gene regulatory network model. At the population level we adopted the Susceptible-Infected-Susceptible (SIS) model. We used our multi-scale model to understand what conditions might determine the balance between use of resistance mechanisms at the two different levels. RESULTS: We explored a range of different conditions (model parameters) that affect host evolutionary dynamics and, in particular, the balance between the use of different resistance mechanisms. These conditions include the complexity of the receptor binding protein-protein interaction, selection pressure on the host population (pathogenicity), and the number of expressed cell-surface receptors. In particular, we found that as the receptor binding complexity (understood as the number of amino acids involved in the interaction between the virus entry protein and the host receptor) increases, viruses tend to become specialists and target one specific receptor. At the same time, on the host side, the potential for resistance shifts from the changes at the level of receptor binding (protein-protein) interaction towards changes at the level of gene regulation, suggesting a mechanism for increased biological complexity. CONCLUSIONS: Host resistance and viral pathogenicity depend on quite different evolutionary conditions. Viruses may evolve cell entry strategies that use small receptor binding regions, represented by low complexity binding in our model. Our modeling results suggest that if the virus adopts a strategy based on binding to low complexity sites on the host receptor, the host will select a defense strategy at the protein (receptor) level, rather than at the level of the regulatory network - a virus-host strategy that appears to have been selected most often in nature.

Antagonistic Coevolution Drives Whack-a-Mole Sensitivity in Gene Regulatory Networks.

Robustness, defined as tolerance to perturbations such as mutations and environmental fluctuations, is pervasive in biological systems. However, robustness often coexists with its counterpart, evolvability--the ability of perturbations to generate new phenotypes. Previous models of gene regulatory network evolution have shown that robustness evolves under stabilizing selection, but it is unclear how robustness and evolvability will emerge in common coevolutionary scenarios. We consider a two-species model of coevolution involving one host and one parasite population. By using two interacting species, key model parameters that determine the fitness landscapes become emergent properties of the model, avoiding the need to impose these parameters externally. In our study, parasites are modeled on species such as cuckoos where mimicry of the host phenotype confers high fitness to the parasite but lower fitness to the host. Here, frequent phenotype changes are favored as each population continually adapts to the other population. Sensitivity evolves at the network level such that point mutations can induce large phenotype changes. Crucially, the sensitive points of the network are broadly distributed throughout the network and continually relocate. Each time sensitive points in the network are mutated, new ones appear to take their place. We have therefore named this phenomenon "whack-a-mole" sensitivity, after a popular fun park game. We predict that this type of sensitivity will evolve under conditions of strong directional selection, an observation that helps interpret existing experimental evidence, for example, during the emergence of bacterial antibiotic resistance.

Epigenetics decouples mutational from environmental robustness. Did it also facilitate multicellularity?

The evolution of ever increasing complex life forms has required innovations at the molecular level in order to overcome existing barriers. For example, evolving processes for cell differentiation, such as epigenetic mechanisms, facilitated the transition to multicellularity. At the same time, studies using gene regulatory network models, and corroborated in single-celled model organisms, have shown that mutational robustness and environmental robustness are correlated. Such correlation may constitute a barrier to the evolution of multicellularity since cell differentiation requires sensitivity to cues in the internal environment during development. To investigate how this barrier might be overcome, we used a gene regulatory network model which includes epigenetic control based on the mechanism of histone modification via Polycomb Group Proteins, which evolved in tandem with the transition to multicellularity. Incorporating the Polycomb mechanism allowed decoupling of mutational and environmental robustness, thus allowing the system to be simultaneously robust to mutations while increasing sensitivity to the environment. In turn, this decoupling facilitated cell differentiation which we tested by evaluating the capacity of the system for producing novel output states in response to altered initial conditions. In the absence of the Polycomb mechanism, the system was frequently incapable of adding new states, whereas with the Polycomb mechanism successful addition of new states was nearly certain. The Polycomb mechanism, which dynamically reshapes the network structure during development as a function of expression dynamics, decouples mutational and environmental robustness, thus providing a necessary step in the evolution of multicellularity.

A core erythroid transcriptional network is repressed by a master regulator of myelo-lymphoid differentiation.

Two mechanisms that play important roles in cell fate decisions are control of a "core transcriptional network" and repression of alternative transcriptional programs by antagonizing transcription factors. Whether these two mechanisms operate together is not known. Here we report that GATA-1, SCL, and Klf1 form an erythroid core transcriptional network by co-occupying >300 genes. Importantly, we find that PU.1, a negative regulator of terminal erythroid differentiation, is a highly integrated component of this network. GATA-1, SCL, and Klf1 act to promote, whereas PU.1 represses expression of many of the core network genes. PU.1 also represses the genes encoding GATA-1, SCL, Klf1, and important GATA-1 cofactors. Conversely, in addition to repressing PU.1 expression, GATA-1 also binds to and represses >100 PU.1 myelo-lymphoid gene targets in erythroid progenitors. Mathematical modeling further supports that this dual mechanism of repressing both the opposing upstream activator and its downstream targets provides a synergistic, robust mechanism for lineage specification. Taken together, these results amalgamate two key developmental principles, namely, regulation of a core transcriptional network and repression of an alternative transcriptional program, thereby enhancing our understanding of the mechanisms that establish cellular identity.

IGHV-unmutated and IGHV-mutated chronic lymphocytic leukemia cells produce activation-induced deaminase protein with a full range of biologic functions.

Clonal evolution occurs during the course of chronic lymphocytic leukemia (CLL) and activation-induced deaminase (AID) could influence this process. However, this possibility has been questioned in CLL because the number of circulating AID mRNA(+) cells is exceedingly low; synthesis of AID protein by blood CLL cells has not been demonstrated; the full range of AID functions is lacking in unmutated CLL (U-CLL), and no prospective analysis linking AID expression and disease severity has been reported. The results of the present study show that circulating CLL cells and those within secondary lymphoid tissues can make AID mRNA and protein. This production is related to cell division because more AID mRNA was detected in recently divided cells and AID protein was limited to the dividing fraction and was up-regulated on induction of cell division. AID protein was functional because AID(+) dividing cells exhibited more double-stranded DNA breaks, IGH class switching, and new IGHV-D-J mutations. Each of these actions was documented in U-CLL and mutated CLL (M-CLL). Furthermore, AID protein was associated with worse patient outcome and adverse cytogenetics. We conclude that the production of fully functional AID protein by U-CLL and M-CLL cells could be involved in clonal evolution of the disease.

The underlying molecular and network level mechanisms in the evolution of robustness in gene regulatory networks.

Gene regulatory networks show robustness to perturbations. Previous works identified robustness as an emergent property of gene network evolution but the underlying molecular mechanisms are poorly understood. We used a multi-tier modeling approach that integrates molecular sequence and structure information with network architecture and population dynamics. Structural models of transcription factor-DNA complexes are used to estimate relative binding specificities. In this model, mutations in the DNA cause changes on two levels: (a) at the sequence level in individual binding sites (modulating binding specificity), and (b) at the network level (creating and destroying binding sites). We used this model to dissect the underlying mechanisms responsible for the evolution of robustness in gene regulatory networks. Results suggest that in sparse architectures (represented by short promoters), a mixture of local-sequence and network-architecture level changes are exploited. At the local-sequence level, robustness evolves by decreasing the probabilities of both the destruction of existent and generation of new binding sites. Meanwhile, in highly interconnected architectures (represented by long promoters), robustness evolves almost entirely via network level changes, deleting and creating binding sites that modify the network architecture.

Interpretable deep learning reveals the role of an E-box motif in suppressing somatic hypermutation of AGCT motifs within human immunoglobulin variable regions.

Introduction: Somatic hypermutation (SHM) of immunoglobulin variable (V) regions by activation induced deaminase (AID) is essential for robust, long-term humoral immunity against pathogen and vaccine antigens. AID mutates cytosines preferentially within WRCH motifs (where W=A or T, R=A or G and H=A, C or T). However, it has been consistently observed that the mutability of WRCH motifs varies substantially, with large variations in mutation frequency even between multiple occurrences of the same motif within a single V region. This has led to the notion that the immediate sequence context of WRCH motifs contributes to mutability. Recent studies have highlighted the potential role of local DNA sequence features in promoting mutagenesis of AGCT, a commonly mutated WRCH motif. Intriguingly, AGCT motifs closer to 5' ends of V regions, within the framework 1 (FW1) sub-region1, mutate less frequently, suggesting an SHM-suppressing sequence context. Methods: Here, we systematically examined the basis of AGCT positional biases in human SHM datasets with DeepSHM, a machine-learning model designed to predict SHM patterns. This was combined with integrated gradients, an interpretability method, to interrogate the basis of DeepSHM predictions. Results: DeepSHM predicted the observed positional differences in mutation frequencies at AGCT motifs with high accuracy. For the conserved, lowly mutating AGCT motifs in FW1, integrated gradients predicted a large negative contribution of 5'C and 3'G flanking residues, suggesting that a CAGCTG context in this location was suppressive for SHM. CAGCTG is the recognition motif for E-box transcription factors, including E2A, which has been implicated in SHM. Indeed, we found a strong, inverse relationship between E-box motif fidelity and mutation frequency. Moreover, E2A was found to associate with the V region locale in two human B cell lines. Finally, analysis of human SHM datasets revealed that naturally occurring mutations in the 3'G flanking residues, which effectively ablate the E-box motif, were associated with a significantly increased rate of AGCT mutation. Discussion: Our results suggest an antagonistic relationship between mutation frequency and the binding of E-box factors like E2A at specific AGCT motif contexts and, therefore, highlight a new, suppressive mechanism regulating local SHM patterns in human V regions.