Explainable machine learning for profiling the immunological synapse and functional characterization of therapeutic antibodies.

Therapeutic antibodies are widely used to treat severe diseases. Most of them alter immune cells and act within the immunological synapse; an essential cell-to-cell interaction to direct the humoral immune response. Although many antibody designs are generated and evaluated, a high-throughput tool for systematic antibody characterization and prediction of function is lacking. Here, we introduce the first comprehensive open-source framework, scifAI (single-cell imaging flow cytometry AI), for preprocessing, feature engineering, and explainable, predictive machine learning on imaging flow cytometry (IFC) data. Additionally, we generate the largest publicly available IFC dataset of the human immunological synapse containing over 2.8 million images. Using scifAI, we analyze class frequency and morphological changes under different immune stimulation. T cell cytokine production across multiple donors and therapeutic antibodies is quantitatively predicted in vitro, linking morphological features with function and demonstrating the potential to significantly impact antibody design. scifAI is universally applicable to IFC data. Given its modular architecture, it is straightforward to incorporate into existing workflows and analysis pipelines, e.g., for rapid antibody screening and functional characterization.

Defining the RBPome of primary T helper cells to elucidate higher-order Roquin-mediated mRNA regulation.

Post-transcriptional gene regulation in T cells is dynamic and complex as targeted transcripts respond to various factors. This is evident for the Icos mRNA encoding an essential costimulatory receptor that is regulated by several RNA-binding proteins (RBP), including Roquin-1 and Roquin-2. Here, we identify a core RBPome of 798 mouse and 801 human T cell proteins by utilizing global RNA interactome capture (RNA-IC) and orthogonal organic phase separation (OOPS). The RBPome includes Stat1, Stat4 and Vav1 proteins suggesting unexpected functions for these transcription factors and signal transducers. Based on proximity to Roquin-1, we select ~50 RBPs for testing coregulation of Roquin-1/2 targets by induced expression in wild-type or Roquin-1/2-deficient T cells. Besides Roquin-independent contributions from Rbms1 and Cpeb4 we also show Roquin-1/2-dependent and target-specific coregulation of Icos by Celf1 and Igf2bp3. Connecting the cellular RBPome in a post-transcriptional context, we find contributions from multiple RBPs to the prototypic regulation of mRNA targets by individual trans-acting factors.

Salt generates antiinflammatory Th17 cells but amplifies pathogenicity in proinflammatory cytokine microenvironments.

Th cells integrate signals from their microenvironment to acquire distinct specialization programs for efficient clearance of diverse pathogens or for immunotolerance. Ionic signals have recently been demonstrated to affect T cell polarization and function. Sodium chloride (NaCl) was proposed to accumulate in peripheral tissues upon dietary intake and to promote autoimmunity via the Th17 cell axis. Here, we demonstrate that high-NaCl conditions induced a stable, pathogen-specific, antiinflammatory Th17 cell fate in human T cells in vitro. The p38/MAPK pathway, involving NFAT5 and SGK1, regulated FoxP3 and IL-17A expression in high-NaCl conditions. The NaCl-induced acquisition of an antiinflammatory Th17 cell fate was confirmed in vivo in an experimental autoimmune encephalomyelitis (EAE) mouse model, which demonstrated strongly reduced disease symptoms upon transfer of T cells polarized in high-NaCl conditions. However, NaCl was coopted to promote murine and human Th17 cell pathogenicity, if T cell stimulation occurred in a proinflammatory and TGF-ß-low cytokine microenvironment. Taken together, our findings reveal a context-dependent, dichotomous role for NaCl in shaping Th17 cell pathogenicity. NaCl might therefore prove beneficial for the treatment of chronic inflammatory diseases in combination with cytokine-blocking drugs.

Immune homeostasis and regulation of the interferon pathway require myeloid-derived Regnase-3.

The RNase Regnase-1 is a master RNA regulator in macrophages and T cells that degrades cellular and viral RNA upon NF-κB signaling. The roles of its family members, however, remain largely unknown. Here, we analyzed Regnase-3-deficient mice, which develop hypertrophic lymph nodes. We used various mice with immune cell-specific deletions of Regnase-3 to demonstrate that Regnase-3 acts specifically within myeloid cells. Regnase-3 deficiency systemically increased IFN signaling, which increased the proportion of immature B and innate immune cells, and suppressed follicle and germinal center formation. Expression analysis revealed that Regnase-3 and Regnase-1 share protein degradation pathways. Unlike Regnase-1, Regnase-3 expression is high specifically in macrophages and is transcriptionally controlled by IFN signaling. Although direct targets in macrophages remain unknown, Regnase-3 can bind, degrade, and regulate mRNAs, such as Zc3h12a (Regnase-1), in vitro. These data indicate that Regnase-3, like Regnase-1, is an RNase essential for immune homeostasis but has diverged as key regulator in the IFN pathway in macrophages.

Alternative Splice Forms of CYLD Mediate Ubiquitination of SMAD7 to Prevent TGFB Signaling and Promote Colitis.

BACKGROUND & AIMS: The CYLD lysine 63 deubiquitinase gene (CYLD) encodes tumor suppressor protein that is mutated in familial cylindromatosus, and variants have been associated with Crohn disease (CD). Splice forms of CYLD that lack exons 7 and 8 regulate transcription factors and functions of immune cells. We examined the expression of splice forms of CYLD in colon tissues from patients with CD and their effects in mice. METHODS: We performed immunohistochemical analyses of colon tissues from patients with untreated CD and patients without inflammatory bowel diseases (controls). We obtained mice that expressed splice forms of CYLD (sCYLD mice) without or with SMAD7 (sCYLD/SMAD7 mice) from transgenes and CYLD-knockout mice (with or without transgenic expression of SMAD7) and performed endoscopic analyses. Colitis was induced in Rag1-/- mice by transfer of CD4+ CD62L+ T cells from C57/Bl6 or transgenic mice. T cells were isolated from mice and analyzed by flow cytometry and quantitative real-time polymerase chain reaction and intestinal tissues were analyzed by histology and immunohistochemistry. CYLD forms were expressed in mouse embryonic fibroblasts, primary T cells, and HEK293T cells, which were analyzed by immunoblot, mobility shift, and immunoprecipitation assays. RESULTS: The colonic lamina propria from patients with CD was infiltrated by T cells and had higher levels of sCYLD (but not full-length CYLD) and SMAD7 than tissues from controls. Incubation of mouse embryonic fibroblasts and T cells with transforming growth factor ß increased their production of sCYLD and decreased full-length CYLD. Transgenic expression of sCYLD and SMAD7 in T cells prevented the differentiation of regulatory T cells and T-helper type 17 cells and increased the differentiation of T-helper type 1 cells. The same effects were observed in colon tissues from sCYLD/SMAD7 mice but not in those from CYLD-knockout SMAD7 mice. The sCYLD mice had significant increases in the numbers of T-helper type 1 cells and CD44high CD62Llow memory-effector CD4+ T cells in the spleen and mesenteric lymph nodes compared with wild-type mice; sCYLD/SMAD7 mice had even larger increases. The sCYLD/SMAD7 mice spontaneously developed severe colitis, with infiltration of the colon by dendritic cells, neutrophils, macrophages, and CD4+ T cells and increased levels of Ifng, Il6, Il12a, Il23a, and Tnf mRNAs. Co-transfer of regulatory T cells from wild-type, but not from sCYLD/SMAD7, mice prevented the induction of colitis in Rag1-/- mice by CD4+ T cells. We found increased levels of poly-ubiquitinated SMAD7 in sCYLD CD4+ T cells. CYLD formed a nuclear complex with SMAD3, whereas sCYLD recruited SMAD7 to the nucleus, which inhibited the expression of genes regulated by SMAD3 and SMAD4. We found that sCYLD mediated lysine 63-linked ubiquitination of SMAD7. The sCYLD-SMAD7 complex inhibited transforming growth factor ß signaling in CD4+ T cells. CONCLUSIONS: Levels of the spliced form of CYLD are increased in colon tissues from patients with CD. sCYLD mediates ubiquitination and nuclear translocation of SMAD7 and thereby decreases transforming growth factor ß signaling in T cells. This prevents immune regulatory mechanisms and leads to colitis in mice.

Real-time Analysis of Transcription Factor Binding, Transcription, Translation, and Turnover to Display Global Events During Cellular Activation.

Upon activation, cells rapidly change their functional programs and, thereby, their gene expression profile. Massive changes in gene expression occur, for example, during cellular differentiation, morphogenesis, and functional stimulation (such as activation of immune cells), or after exposure to drugs and other factors from the local environment. Depending on the stimulus and cell type, these changes occur rapidly and at any possible level of gene regulation. Displaying all molecular processes of a responding cell to a certain type of stimulus/drug is one of the hardest tasks in molecular biology. Here, we describe a protocol that enables the simultaneous analysis of multiple layers of gene regulation. We compare, in particular, transcription factor binding (Chromatin-immunoprecipitation-sequencing (ChIP-seq)), de novo transcription (4-thiouridine-sequencing (4sU-seq)), mRNA processing, and turnover as well as translation (ribosome profiling). By combining these methods, it is possible to display a detailed and genome-wide course of action. Sequencing newly transcribed RNA is especially recommended when analyzing rapidly adapting or changing systems, since this depicts the transcriptional activity of all genes during the time of 4sU exposure (irrespective of whether they are up- or downregulated). The combinatorial use of total RNA-seq and ribosome profiling additionally allows the calculation of RNA turnover and translation rates. Bioinformatic analysis of high-throughput sequencing results allows for many means for analysis and interpretation of the data. The generated data also enables tracking co-transcriptional and alternative splicing, just to mention a few possible outcomes. The combined approach described here can be applied for different model organisms or cell types, including primary cells. Furthermore, we provide detailed protocols for each method used, including quality controls, and discuss potential problems and pitfalls.

Immune Responses - Transcriptional and Post-Transcriptional Networks Pass the Baton.

Diverse gene regulatory mechanisms impact on immune homeostasis, and a new model now emerges as fundamental in light of recent genome-wide studies. In this picture, transcriptional networks drive functional changes during immune activation, whereas autoregulatory feedback loops of post-transcriptional programs ensure the original cell lineage identity and subsequent immune resolution.

Splicing in immune cells-mechanistic insights and emerging topics.

Differential splicing of mRNAs not only enables regulation of gene expression levels, but also ensures a high degree of gene-product diversity. The extent to which splicing of mRNAs is utilized as a mechanism in immune cells has become evident within the last few years. Still, only a few of these mechanisms have been well studied. In this review, we discuss some of the best-understood mechanisms, for instance the differential splicing of CD45 in T cells, as well as immunoglobulin genes in B cells. Beyond that we provide general mechanistic insights on how, when and where this process takes place and discuss the current knowledge regarding these topics in immune cells. We also highlight some of the reported links to immune-related diseases, genome-wide sequencing studies that revealed thousands of differentially spliced transcripts, as well as splicing studies on immune cells that remain mechanistically not fully understood. We thereby display potential emerging topics for future studies centered on splicing mechanisms in immune cells.

Elevated levels of Bcl-3 inhibits Treg development and function resulting in spontaneous colitis.

Bcl-3 is an atypical NF-κB family member that regulates NF-κB-dependent gene expression in effector T cells, but a cell-intrinsic function in regulatory T (Treg) cells and colitis is not clear. Here we show that Bcl-3 expression levels in colonic T cells correlate with disease manifestation in patients with inflammatory bowel disease. Mice with T-cell-specific overexpression of Bcl-3 develop severe colitis that can be attributed to defective Treg cell development and function, leading to the infiltration of immune cells such as pro-inflammatory γδT cells, but not αß T cells. In Treg cells, Bcl-3 associates directly with NF-κB p50 to inhibit DNA binding of p50/p50 and p50/p65 NF-κB dimers, thereby regulating NF-κB-mediated gene expression. This study thus reveals intrinsic functions of Bcl-3 in Treg cells, identifies Bcl-3 as a potential prognostic marker for colitis and illustrates the mechanism by which Bcl-3 regulates NF-κB activity in Tregs to prevent colitis.

Rapid Genome-wide Recruitment of RNA Polymerase II Drives Transcription, Splicing, and Translation Events during T Cell Responses.

Activation of immune cells results in rapid functional changes, but how such fast changes are accomplished remains enigmatic. By combining time courses of 4sU-seq, RNA-seq, ribosome profiling (RP), and RNA polymerase II (RNA Pol II) ChIP-seq during T cell activation, we illustrate genome-wide temporal dynamics for ∼10,000 genes. This approach reveals not only immediate-early and posttranscriptionally regulated genes but also coupled changes in transcription and translation for >90% of genes. Recruitment, rather than release of paused RNA Pol II, primarily mediates transcriptional changes. This coincides with a genome-wide temporary slowdown in cotranscriptional splicing, even for polyadenylated mRNAs that are localized at the chromatin. Subsequent splicing optimization correlates with increasing Ser-2 phosphorylation of the RNA Pol II carboxy-terminal domain (CTD) and activation of the positive transcription elongation factor (pTEFb). Thus, rapid de novo recruitment of RNA Pol II dictates the course of events during T cell activation, particularly transcription, splicing, and consequently translation.

Alternatively activated macrophages do not synthesize catecholamines or contribute to adipose tissue adaptive thermogenesis.

Adaptive thermogenesis is the process of heat generation in response to cold stimulation. It is under the control of the sympathetic nervous system, whose chief effector is the catecholamine norepinephrine (NE). NE enhances thermogenesis through ß3-adrenergic receptors to activate brown adipose tissue and by 'browning' white adipose tissue. Recent studies have reported that alternative activation of macrophages in response to interleukin (IL)-4 stimulation induces the expression of tyrosine hydroxylase (TH), a key enzyme in the catecholamine synthesis pathway, and that this activation provides an alternative source of locally produced catecholamines during the thermogenic process. Here we report that the deletion of Th in hematopoietic cells of adult mice neither alters energy expenditure upon cold exposure nor reduces browning in inguinal adipose tissue. Bone marrow-derived macrophages did not release NE in response to stimulation with IL-4, and conditioned media from IL-4-stimulated macrophages failed to induce expression of thermogenic genes, such as uncoupling protein 1 (Ucp1), in adipocytes cultured with the conditioned media. Furthermore, chronic treatment with IL-4 failed to increase energy expenditure in wild-type, Ucp1-/- and interleukin-4 receptor-α double-negative (Il4ra-/-) mice. In agreement with these findings, adipose-tissue-resident macrophages did not express TH. Thus, we conclude that alternatively activated macrophages do not synthesize relevant amounts of catecholamines, and hence, are not likely to have a direct role in adipocyte metabolism or adaptive thermogenesis.

The molecular choreography of IRF4 and IRF8 with immune system partners.

The transcription factors IRF4 and IRF8 represent immune-specific members of the interferon regulatory family. They play major roles in controlling the development and functioning of innate and adaptive cells. Genes encoding these factors appear to have been coopted by the immune system via gene duplication and divergence of regulatory and protein coding sequences to enable the acquisition of unique molecular properties and functions. Unlike other members of the IRF family, IRF4 and IRF8 do not activate transcription of Type 1 interferon genes or positively regulate interferon-induced gene expression. Instead, they bind to unusual composite Ets-IRF or AP-1-IRF elements with specific Ets or AP-1 family transcription factors, respectively, and regulate the expression of diverse sets of immune response genes in innate as well as adaptive cells.

Compensatory dendritic cell development mediated by BATF-IRF interactions.

The AP1 transcription factor Batf3 is required for homeostatic development of CD8α(+) classical dendritic cells that prime CD8 T-cell responses against intracellular pathogens. Here we identify an alternative, Batf3-independent pathway in mice for CD8α(+) dendritic cell development operating during infection with intracellular pathogens and mediated by the cytokines interleukin (IL)-12 and interferon-γ. This alternative pathway results from molecular compensation for Batf3 provided by the related AP1 factors Batf, which also functions in T and B cells, and Batf2 induced by cytokines in response to infection. Reciprocally, physiological compensation between Batf and Batf3 also occurs in T cells for expression of IL-10 and CTLA4. Compensation among BATF factors is based on the shared capacity of their leucine zipper domains to interact with non-AP1 factors such as IRF4 and IRF8 to mediate cooperative gene activation. Conceivably, manipulating this alternative pathway of dendritic cell development could be of value in augmenting immune responses to vaccines.

A genomic regulatory element that directs assembly and function of immune-specific AP-1-IRF complexes.

Interferon regulatory factor 4 (IRF4) and IRF8 regulate B, T, macrophage, and dendritic cell differentiation. They are recruited to cis-regulatory Ets-IRF composite elements by PU.1 or Spi-B. How these IRFs target genes in most T cells is enigmatic given the absence of specific Ets partners. Chromatin immunoprecipitation sequencing in T helper 17 (T(H)17) cells reveals that IRF4 targets sequences enriched for activating protein 1 (AP-1)-IRF composite elements (AICEs) that are co-bound by BATF, an AP-1 factor required for T(H)17, B, and dendritic cell differentiation. IRF4 and BATF bind cooperatively to structurally divergent AICEs to promote gene activation and T(H)17 differentiation. The AICE motif directs assembly of IRF4 or IRF8 with BATF heterodimers and is also used in T(H)2, B, and dendritic cells. This genomic regulatory element and cognate factors appear to have evolved to integrate diverse immunomodulatory signals.

Dicer-dependent and -independent Argonaute2 protein interaction networks in mammalian cells.

Argonaute (Ago) proteins interact with small regulatory RNAs such as microRNAs (miRNAs) and facilitate gene-silencing processes. miRNAs guide Ago proteins to specific mRNAs leading to translational silencing or mRNA decay. In order to understand the mechanistic details of miRNA function, it is important to characterize Ago protein interactors. Although several proteomic studies have been performed, it is not clear how the Ago interactome changes on miRNA or mRNA binding. Here, we report the analysis of Ago protein interactions in miRNA-containing and miRNA-depleted cells. Using stable isotope labeling in cell culture in conjunction with Dicer knock out mouse embryonic fibroblasts, we identify proteins that interact with Ago2 in the presence or the absence of Dicer. In contrast to our current view, we find that Ago-mRNA interactions can also take place in the absence of miRNAs. Our proteomics approach provides a rich resource for further functional studies on the cellular roles of Ago proteins.

Roquin binds inducible costimulator mRNA and effectors of mRNA decay to induce microRNA-independent post-transcriptional repression.

The molecular mechanism by which roquin controls the expression of inducible costimulator (ICOS) to prevent autoimmunity remains unsolved. Here we show that in helper T cells, roquin localized to processing (P) bodies and downregulated ICOS expression. The repression was dependent on the RNA helicase Rck, and roquin interacted with Rck and the enhancer of decapping Edc4, which act together in mRNA decapping. Sequences in roquin that confer P-body localization were essential for roquin-mediated ICOS repression. However, this process did not require microRNAs or the RNA-induced silencing complex (RISC). Instead, roquin bound ICOS mRNA directly, showing an intrinsic preference for a previously unrecognized sequence in the 3' untranslated region (3' UTR). Our results support a model in which roquin controls ICOS expression through binding to the 3' UTR of ICOS mRNA and by interacting with proteins that confer post-transcriptional repression.

Mouse Eri1 interacts with the ribosome and catalyzes 5.8S rRNA processing.

Eri1 is a 3'-to-5' exoribonuclease conserved from fission yeast to humans. Here we show that Eri1 associates with ribosomes and ribosomal RNA (rRNA). Ribosomes from Eri1-deficient mice contain 5.8S rRNA that is aberrantly extended at its 3' end, and Eri1, but not a catalytically inactive mutant, converts this abnormal 5.8S rRNA to the wild-type form in vitro and in cells. In human and murine cells, Eri1 localizes to the cytoplasm and nucleus, with enrichment in the nucleolus, the site of preribosome biogenesis. RNA binding residues in the Eri1 SAP and linker domains promote stable association with rRNA and thereby facilitate 5.8S rRNA 3' end processing. Taken together, our findings indicate that Eri1 catalyzes the final trimming step in 5.8S rRNA processing, functionally and spatially connecting this regulator of RNAi with the basal translation machinery.

Mass spectrometry reveals the missing links in the assembly pathway of the bacterial 20 S proteasome.

The 20 S proteasome is an essential proteolytic particle, responsible for degrading short-lived and abnormal intracellular proteins. The 700-kDa assembly is comprised of 14 alpha-type and 14 beta-type subunits, which form a cylindrical architecture composed of four stacked heptameric rings (alpha7beta7beta7alpha7). The formation of the 20 S proteasome is a complex process that involves a cascade of folding, assembly, and processing events. To date, the understanding of the assembly pathway is incomplete due to the experimental challenges of capturing short-lived intermediates. In this study, we have applied a real-time mass spectrometry approach to capture transient species along the assembly pathway of the 20 S proteasome from Rhodococcus erythropolis. In the course of assembly, we observed formation of an early alpha/beta-heterodimer as well as an unprocessed half-proteasome particle. Formation of mature holoproteasomes occurred in concert with the disappearance of half-proteasomes. We also analyzed the beta-subunits before and during assembly and reveal that those with longer propeptides are incorporated into half- and full proteasomes more rapidly than those that are heavily truncated. To characterize the preholoproteasome, formed by docking of two unprocessed half-proteasomes and not observed during assembly of wild type subunits, we trapped this intermediate using a beta-subunit mutational variant. In summary, this study provides evidence for transient intermediates in the assembly pathway and reveals detailed insight into the cleavage sites of the propeptide.