Microbiome-based disease prediction with multimodal variational information bottlenecks.

Scientific research is shedding light on the interaction of the gut microbiome with the human host and on its role in human health. Existing machine learning methods have shown great potential in discriminating healthy from diseased microbiome states. Most of them leverage shotgun metagenomic sequencing to extract gut microbial species-relative abundances or strain-level markers. Each of these gut microbial profiling modalities showed diagnostic potential when tested separately; however, no existing approach combines them in a single predictive framework. Here, we propose the Multimodal Variational Information Bottleneck (MVIB), a novel deep learning model capable of learning a joint representation of multiple heterogeneous data modalities. MVIB achieves competitive classification performance while being faster than existing methods. Additionally, MVIB offers interpretable results. Our model adopts an information theoretic interpretation of deep neural networks and computes a joint stochastic encoding of different input data modalities. We use MVIB to predict whether human hosts are affected by a certain disease by jointly analysing gut microbial species-relative abundances and strain-level markers. MVIB is evaluated on human gut metagenomic samples from 11 publicly available disease cohorts covering 6 different diseases. We achieve high performance (0.80 < ROC AUC < 0.95) on 5 cohorts and at least medium performance on the remaining ones. We adopt a saliency technique to interpret the output of MVIB and identify the most relevant microbial species and strain-level markers to the model's predictions. We also perform cross-study generalisation experiments, where we train and test MVIB on different cohorts of the same disease, and overall we achieve comparable results to the baseline approach, i.e. the Random Forest. Further, we evaluate our model by adding metabolomic data derived from mass spectrometry as a third input modality. Our method is scalable with respect to input data modalities and has an average training time of < 1.4 seconds. The source code and the datasets used in this work are publicly available.

Vaccination of Macaques with DNA Followed by Adenoviral Vectors Encoding Simian Immunodeficiency Virus (SIV) Gag Alone Delays Infection by Repeated Mucosal Challenge with SIV.

Vaccines aimed at inducing T cell responses to protect against human immunodeficiency virus (HIV) infection have been under development for more than 15 years. Replication-defective adenovirus (rAd) vaccine vectors are at the forefront of this work and have been tested extensively in the simian immunodeficiency virus (SIV) challenge macaque model. Vaccination with rAd vectors coding for SIV Gag or other nonenvelope proteins induces T cell responses that control virus load but disappointingly is unsuccessful so far in preventing infection, and attention has turned to inducing antibodies to the envelope. However, here we report that Mauritian cynomolgus macaques (MCM), Macaca fascicularis, vaccinated with unmodified SIV gag alone in a DNA prime followed by an rAd boost exhibit increased protection from infection by repeated intrarectal challenge with low-dose SIVmac251. There was no evidence of infection followed by eradication. A significant correlation was observed between cytokine expression by CD4 T cells and delayed infection. Vaccination with gag fused to the ubiquitin gene or fragmented, designed to increase CD8 magnitude and breadth, did not confer resistance to challenge or enhance immunity. On infection, a significant reduction in peak virus load was observed in all vaccinated animals, including those vaccinated with modified gag These findings suggest that a nonpersistent viral vector vaccine coding for internal virus proteins may be able to protect against HIV type 1 (HIV-1) infection. The mechanisms are probably distinct from those of antibody-mediated virus neutralization or cytotoxic CD8 cell killing of virus-infected cells and may be mediated in part by CD4 T cells.IMPORTANCE The simian immunodeficiency virus (SIV) macaque model represents the best animal model for testing new human immunodeficiency virus type 1 (HIV-1) vaccines. Previous studies employing replication-defective adenovirus (rAd) vectors that transiently express SIV internal proteins induced T cell responses that controlled virus load but did not protect against virus challenge. However, we show for the first time that SIV gag delivered in a DNA prime followed by a boost with an rAd vector confers resistance to SIV intrarectal challenge. Other partially successful SIV/HIV-1 protective vaccines induce antibody to the envelope and neutralize the virus or mediate antibody-dependent cytotoxicity. Induction of CD8 T cells which do not prevent initial infection but eradicate infected cells before infection becomes established has also shown some success. In contrast, the vaccine described here mediates resistance by a different mechanism from that described above, which may reflect CD4 T cell activity. This could indicate an alternative approach for HIV-1 vaccine development.

Pathway discovery using transcriptomic profiles in adult-onset severe asthma.

BACKGROUND: Adult-onset severe asthma is characterized by highly symptomatic disease despite high-intensity asthma treatments. Understanding of the underlying pathways of this heterogeneous disease is needed for the development of targeted treatments. Gene set variation analysis is a statistical technique used to identify gene profiles in heterogeneous samples. OBJECTIVE: We sought to identify gene profiles associated with adult-onset severe asthma. METHODS: This was a cross-sectional, observational study in which adult patients with adult-onset of asthma (defined as starting at age ≥18 years) as compared with childhood-onset severe asthma (<18 years) were selected from the U-BIOPRED cohort. Gene expression was assessed on the total RNA of induced sputum (n = 83), nasal brushings (n = 41), and endobronchial brushings (n = 65) and biopsies (n = 47) (Affymetrix HT HG-U133+ PM). Gene set variation analysis was used to identify differentially enriched predefined gene signatures of leukocyte lineage, inflammatory and induced lung injury pathways. RESULTS: Significant differentially enriched gene signatures in patients with adult-onset as compared with childhood-onset severe asthma were identified in nasal brushings (5 signatures), sputum (3 signatures), and endobronchial brushings (6 signatures). Signatures associated with eosinophilic airway inflammation, mast cells, and group 3 innate lymphoid cells were more enriched in adult-onset severe asthma, whereas signatures associated with induced lung injury were less enriched in adult-onset severe asthma. CONCLUSIONS: Adult-onset severe asthma is characterized by inflammatory pathways involving eosinophils, mast cells, and group 3 innate lymphoid cells. These pathways could represent useful targets for the treatment of adult-onset severe asthma.

Transcriptomic gene signatures associated with persistent airflow limitation in patients with severe asthma.

A proportion of severe asthma patients suffers from persistent airflow limitation (PAL), often associated with more symptoms and exacerbations. Little is known about the underlying mechanisms. Here, our aim was to discover unexplored potential mechanisms using Gene Set Variation Analysis (GSVA), a sensitive technique that can detect underlying pathways in heterogeneous samples.Severe asthma patients from the U-BIOPRED cohort with PAL (post-bronchodilator forced expiratory volume in 1 s/forced vital capacity ratio below the lower limit of normal) were compared with those without PAL. Gene expression was assessed on the total RNA of sputum cells, nasal brushings, and endobronchial brushings and biopsies. GSVA was applied to identify differentially enriched predefined gene signatures based on all available gene expression publications and data on airways disease.Differentially enriched gene signatures were identified in nasal brushings (n=1), sputum (n=9), bronchial brushings (n=1) and bronchial biopsies (n=4) that were associated with response to inhaled steroids, eosinophils, interleukin-13, interferon-α, specific CD4+ T-cells and airway remodelling.PAL in severe asthma has distinguishable underlying gene networks that are associated with treatment, inflammatory pathways and airway remodelling. These findings point towards targets for the therapy of PAL in severe asthma.

Fusion of ubiquitin to HIV gag impairs human monocyte-derived dendritic cell maturation and reduces ability to induce gag T cell responses.

The efficient induction of CD8 T cell immunity is dependent on the processing and presentation of antigen on MHC class I molecules by professional antigen presenting cells (APC). To develop an improved T cell vaccine for HIV we investigated whether fusing the ubiquitin gene to the N terminus of the HIV gag gene enhanced targeting to the proteasome resulting in better CD8 T cell responses. Human monocyte derived dendritic cells (moDC), transduced with adenovirus vectors carrying either ubiquitinated or non-ubiquitinated gag transgene constructs, were co-cultured with autologous naïve T cells and T cell responses were measured after several weekly cycles of stimulation. Despite targeting of the ubiquitin gag transgene protein to the proteasome, ubiquitination did not increase CD8 T cell immune responses and in some cases diminished responses to gag peptides. There were no marked differences in cytokines produced from ubiquitinated and non-ubiquitinated gag stimulated cultures or in the expression of inhibitory molecules on expanded T cells. However, the ability of moDC transduced with ubiquitinated gag gene to upregulate co-stimulatory molecules was reduced, whilst no difference in moDC maturation was observed with a control ubiquitinated and non-ubiquitinated MART gene. Furthermore moDC transduced with ubiquitinated gag produced more IL-10 than transduction with unmodified gag. Thus failure of gag ubiquitination to enhance CD8 responses may be caused by suppression of moDC maturation. These results indicate that when designing a successful vaccine strategy to target a particular cell population, attention must also be given to the effect of the vaccine on APCs.

Distinct conformations of the chemokine receptor CCR4 with implications for its targeting in allergy.

CC chemokine receptor 4 (CCR4) is expressed by Th2 and regulatory T cells and directs their migration along gradients of the chemokines CCL17 and CCL22. Both chemokines and receptor are upregulated in allergic disease, making CCR4 a therapeutic target for the treatment of allergy. We set out to assess the mechanisms underlying a previous report that CCL22 is a dominant ligand of CCR4, which may have implications for its therapeutic targeting. Human T cells expressing endogenous CCR4 and transfectants engineered to express CCR4 were assessed for receptor function, using assays of calcium release, chemotaxis, receptor endocytosis, and ligand binding. Despite the two ligands having equal potency in calcium flux and chemotaxis assays, CCL22 showed dominance in both receptor endocytosis assays and heterologous competitive binding assays. Using two different CCR4-specific Abs, we showed that CCR4 exists in at least two distinct conformations, which are differentially activated by ligand. A major population is activated by both CCL17 and CCL22, whereas a minor population is activated only by CCL22. Mutation of a single C-terminal residue K310 within a putative CCR4 antagonist binding site ablated activation of CCR4 by CCL17, but not by CCL22, despite having no effect on the binding of either ligand. We conclude that CCL17 and CCL22 are conformationally selective ligands of CCR4 and interact with the receptor by substantially different mechanisms. This finding suggests that the selective blockade of CCR4 in allergy may be feasible when one CCR4 ligand dominates, allowing the inhibition of Th2 signaling via one ligand while sparing regulatory T cell recruitment via another.

Fragmentation of SIV-gag vaccine induces broader T cell responses.

BACKGROUND: High mutation rates of human immunodeficiency virus (HIV) allows escape from T cell recognition preventing development of effective T cell vaccines. Vaccines that induce diverse T cell immune responses would help overcome this problem. Using SIV gag as a model vaccine, we investigated two approaches to increase the breadth of the CD8 T cell response. Namely, fusion of vaccine genes to ubiquitin to target the proteasome and increase levels of MHC class I peptide complexes and gene fragmentation to overcome competition between epitopes for presentation and recognition. METHODOLOGY/PRINCIPAL FINDINGS: three vaccines were compared: full-length unmodified SIV-mac239 gag, full-length gag fused at the N-terminus to ubiquitin and 7 gag fragments of equal size spanning the whole of gag with ubiquitin-fused to the N-terminus of each fragment. Genes were cloned into a replication defective adenovirus vector and immunogenicity assessed in an in vitro human priming system. The breadth of the CD8 T cell response, defined by the number of distinct epitopes, was assessed by IFN-γ-ELISPOT and memory phenotype and cytokine production evaluated by flow cytometry. We observed an increase of two- to six-fold in the number of epitopes recognised in the ubiquitin-fused fragments compared to the ubiquitin-fused full-length gag. In contrast, although proteasomal targeting was achieved, there was a marked reduction in the number of epitopes recognised in the ubiquitin-fused full-length gag compared to the full-length unmodified gene, but there were no differences in the number of epitope responses induced by non-ubiquitinated full-length gag and the ubiquitin-fused mini genes. Fragmentation and ubiquitination did not affect T cell memory differentiation and polyfunctionality, though most responses were directed against the Ad5 vector. CONCLUSION/SIGNIFICANCE: Fragmentation but not fusion with ubiquitin increases the breadth of the CD8 T vaccine response against SIV-mac239 gag. Thus gene fragmentation of HIV vaccines may maximise responses.

Adenovirus vector vaccination induces expansion of memory CD4 T cells with a mucosal homing phenotype that are readily susceptible to HIV-1.

In the recently halted HIV type 1 (HIV-1) vaccine STEP trial, individuals that were seropositive for adenovirus serotype 5 (Ad5) showed increased rates of HIV-1 infection on vaccination with an Ad5 vaccine. We propose that this was due to activation and expansion of Ad5-specific mucosal-homing memory CD4 T cells. To test this hypothesis, Ad5 and Ad11 antibody titers were measured in 20 healthy volunteers. Dendritic cells (DCs) from these individuals were pulsed with replication defective Ad5 or Ad11 and co-cultured with autologous lymphocytes. Cytokine profiles, proliferative capacity, mucosal migration potential, and susceptibility to HIV infection of the adenovirus-stimulated memory CD4 T cells were measured. Stimulation of T cells from healthy Ad5-seropositive but Ad11-seronegative individuals with Ad5, or serologically distinct Ad11 vectors induced preferential expansion of adenovirus memory CD4 T cells expressing alpha(4)beta(7) integrins and CCR9, indicating a mucosal-homing phenotype. CD4 T-cell proliferation and IFN-gamma production in response to Ad stimulation correlated with Ad5 antibody titers. However, Ad5 serostatus did not correlate with total cytokine production upon challenge with Ad5 or Ad11. Expanded Ad5 and Ad11 memory CD4 T cells showed an increase in CCR5 expression and higher susceptibility to infection by R5 tropic HIV-1. This suggests that adenoviral-based vaccination against HIV-1 in individuals with preexisting immunity against Ad5 results in preferential expansion of HIV-susceptible activated CD4 T cells that home to mucosal tissues, increases the number of virus targets, and leads to a higher susceptibility to HIV acquisition.

The chemokine receptor CXCR3 is degraded following internalization and is replenished at the cell surface by de novo synthesis of receptor.

The chemokine receptor CXCR3 is expressed on the surface of both resting and activated T lymphocytes. We describe in this study the endocytosis of CXCR3 using T lymphocytes and CXCR3 transfectants. Chemokine-induced CXCR3 down-regulation occurred in a rapid, dose-dependent manner, with CXCL11 the most potent and efficacious ligand. Endocytosis was mediated in part by arrestins, but appeared to occur independently of clathrin and caveolae. In contrast to other chemokine receptors, which are largely recycled to the cell surface within an hour, cell surface replenishment of CXCR3 occurred over several hours and was dependent upon mRNA transcription, de novo protein synthesis, and transport through the endoplasmic reticulum and Golgi. Confocal microscopy and Western blotting confirmed the fate of endocytosed CXCR3 to be degradation, mediated in part by lysosomes and proteosomes. Site-directed mutagenesis of the CXCR3 C terminus revealed that internalization and degradation were independent of phosphorylation, ubiquitination, or a conserved LL motif. CXCR3 was found to be efficiently internalized in the absence of ligand, a process involving a YXXL motif at the extreme of the C terminus. Although freshly isolated T lymphocytes expressed moderate cell surface levels of CXCR3, they were only responsive to CXCL11 with CXCL9 and CXCL10 only having significant activity on activated T lymphocytes. Thus, the activities of CXCR3 are tightly controlled following mRNA translation. Because CXCR3(+) cells are themselves a source of IFN-gamma, which potently induces the expression of CXCR3 ligands, such tight regulation of CXCR3 may serve as a control to avoid the unnecessary amplification of activated T lymphocyte recruitment.

CXCL4-induced migration of activated T lymphocytes is mediated by the chemokine receptor CXCR3.

The chemokine CXCL4/platelet factor-4 is released by activated platelets in micromolar concentrations and is a chemoattractant for leukocytes via an unidentified receptor. Recently, a variant of the human chemokine receptor CXCR3 (CXCR3-B) was described, which transduced apoptotic but not chemotactic signals in microvascular endothelial cells following exposure to high concentrations of CXCL4. Here, we show that CXCL4 can induce intracellular calcium release and the migration of activated human T lymphocytes. CXCL4-induced chemotaxis of T lymphocytes was inhibited by a CXCR3 antagonist and pretreatment of cells with pertussis toxin (PTX), suggestive of CXCR3-mediated G-protein signaling via Galphai-sensitive subunits. Specific binding by T lymphocytes of the CXCR3 ligand CXCL10 was not effectively competed by CXCL4, suggesting that the two are allotopic ligands. We subsequently used expression systems to dissect the potential roles of each CXCR3 isoform in mediating CXCL4 function. Transient expression of the CXCR3-A and CXCR3-B isoforms in the murine pre-B cell L1.2 produced cells that migrated in response to CXCL4 in a manner sensitive to PTX and a CXCR3 antagonist. Binding of radiolabeled CXCL4 to L1.2 CXCR3 transfectants was of low affinity and appeared to be mediated chiefly by glycosaminoglycans (GAGs), as no specific CXCL4 binding was observed in GAG-deficient 745-Chinese hamster ovary cells stably expressing CXCR3. We suggest that following platelet activation, the CXCR3/CXCL4 axis may play a role in T lymphocyte recruitment and the subsequent amplification of inflammation observed in diseases such as atherosclerosis. In such a setting, antagonism of the CXCR3/CXCL4 axis may represent a useful, therapeutic intervention.

Plasma membrane budding as an alternative release mechanism of the extracellular enveloped form of vaccinia virus from HeLa cells.

In HeLa cells the assembly of modified vaccinia virus Ankara (MVA), an attenuated vaccinia virus (VV) strain, is blocked. No intracellular mature viruses (IMVs) are made and instead, immature viruses accumulate, some of which undergo condensation and are released from the cell. The condensed particles may undergo wrapping by membranes of the trans-Golgi network and fusion with the plasma membrane prior to their release (M. W. Carroll and B. Moss, Virology 238:198-211, 1997). The present study shows by electron microscopy (EM), however, that the dense particles made in HeLa cells are also released by a budding process at the plasma membrane. By labeling the plasma membrane with antibodies to B5R, a membrane protein of the extracellular enveloped virus, we show that budding occurs at sites that concentrate this protein. EM quantitation revealed that the cell surface around a budding profile was as strongly labeled with anti-B5R antibody as were the extracellular particles, whereas the remainder of the plasma membrane was significantly less labeled. To test whether budding was a characteristic of MVA infection, HeLa cells were infected with the replication competent VV strains Western Reserve strain (WR) and International Health Department strain-J (IHD-J) and also prepared for EM. EM analyses, surprisingly, revealed for both virus strains IMVs that evidently budded at the cell surface at sites that were significantly labeled with anti-B5R. EM also indicated that budding of MVA dense particles was more efficient than budding of IMVs from WR- or IHD-J-infected cells. This was confirmed by semipurifying [(35)S]methionine-labeled dense particles or extracellular enveloped virus (EEVs) from the culture supernatant of MVA- or IHD-J-infected HeLa cells, respectively, showing that threefold more labeled dense particles were secreted than EEVs. Finally, although the released MVA dense particles contain some DNA, they are not infectious, as assessed by plaque assays.

Comparison of virus production in chicken embryo fibroblasts infected with the WR, IHD-J and MVA strains of vaccinia virus: IHD-J is most efficient in trans-Golgi network wrapping and extracellular enveloped virus release.

Modified vaccinia virus Ankara (MVA) is an attenuated strain derived from vaccinia virus (VV) Ankara that grows efficiently in primary chicken embryo fibroblasts (CEFs) and baby hamster kidney cells only. MVA produces significantly more of the enveloped forms of VV in infected CEFs compared with VV strain Copenhagen. In the present study, production of the different infectious forms of VV was compared in CEFs infected with MVA or with two well-characterized replication-competent VV strains, WR and IHD-J. In a time-course experiment, the infectivity associated with the extracellular enveloped virus (EEV), the cell-associated enveloped virus (CEV) and intracellular mature and enveloped viruses was determined. Further, the production of the different viral forms was quantified by electron microscopy (EM). The data collectively indicate that IHD-J is most efficient in producing all of the trans-Golgi network-wrapped forms and releases the highest titres of EEVs into the extracellular medium, with WR being least efficient. MVA initially replicated with faster kinetics, resulting in more intracellular virus and CEVs between 8 and 24 h post-infection (p.i.). As assessed by EM, the faster growth kinetics of MVA resulted in 3.5-fold more CEVs at the cell surface at 24 h p.i., compared with both WR and IHD-J. Accordingly, we found that despite the presence of two in-frame deletions in the A36R gene of MVA, this virus was able to make actin tails in CEFs.