Tensor-based insights into systems immunity and infectious disease.

Profiling immune responses across several dimensions, including time, patients, molecular features, and tissue sites, can deepen our understanding of immunity as an integrated system. These studies require new analytical approaches to realize their full potential. We highlight recent applications of tensor methods and discuss several future opportunities.

Design of cell-type-specific hyperstable IL-4 mimetics via modular de novo scaffolds.

The interleukin-4 (IL-4) cytokine plays a critical role in modulating immune homeostasis. Although there is great interest in harnessing this cytokine as a therapeutic in natural or engineered formats, the clinical potential of native IL-4 is limited by its instability and pleiotropic actions. Here, we design IL-4 cytokine mimetics (denoted Neo-4) based on a de novo engineered IL-2 mimetic scaffold and demonstrate that these cytokines can recapitulate physiological functions of IL-4 in cellular and animal models. In contrast with natural IL-4, Neo-4 is hyperstable and signals exclusively through the type I IL-4 receptor complex, providing previously inaccessible insights into differential IL-4 signaling through type I versus type II receptors. Because of their hyperstability, our computationally designed mimetics can directly incorporate into sophisticated biomaterials that require heat processing, such as three-dimensional-printed scaffolds. Neo-4 should be broadly useful for interrogating IL-4 biology, and the design workflow will inform targeted cytokine therapeutic development.

A stimulus-contingent positive feedback loop enables IFN-ß dose-dependent activation of pro-inflammatory genes.

Type I interferons (IFN) induce powerful antiviral and innate immune responses via the transcription factor, IFN-stimulated gene factor (ISGF3). However, in some pathological contexts, type I IFNs are responsible for exacerbating inflammation. Here, we show that a high dose of IFN-ß also activates an inflammatory gene expression program in contrast to IFN-λ3, a type III IFN, which elicits only the common antiviral gene program. We show that the inflammatory gene program depends on a second, potentiated phase in ISGF3 activation. Iterating between mathematical modeling and experimental analysis, we show that the ISGF3 activation network may engage a positive feedback loop with its subunits IRF9 and STAT2. This network motif mediates stimulus-specific ISGF3 dynamics that are dependent on ligand, dose, and duration of exposure, and when engaged activates the inflammatory gene expression program. Our results reveal a previously underappreciated dynamical control of the JAK-STAT/IRF signaling network that may produce distinct biological responses and suggest that studies of type I IFN dysregulation, and in turn therapeutic remedies, may focus on feedback regulators within it.

Disulfide-HMGB1 signals through TLR4 and TLR9 to induce inflammatory macrophages capable of innate-adaptive crosstalk in human liver transplantation.

Ischemia-reperfusion injury (IRI) during orthotopic liver transplantation (OLT) contributes to graft rejection and poor clinical outcomes. The disulfide form of high mobility group box 1 (diS-HMGB1), an intracellular protein released during OLT-IRI, induces pro-inflammatory macrophages. How diS-HMGB1 differentiates human monocytes into macrophages capable of activating adaptive immunity remains unknown. We investigated if diS-HMGB1 binds toll-like receptor (TLR) 4 and TLR9 to differentiate monocytes into pro-inflammatory macrophages that activate adaptive immunity and promote graft injury and dysfunction. Assessment of 106 clinical liver tissue and longitudinal blood samples revealed that OLT recipients were more likely to experience IRI and graft dysfunction with increased diS-HMGB1 released during reperfusion. Increased diS-HMGB1 concentration also correlated with TLR4/TLR9 activation, polarization of monocytes into pro-inflammatory macrophages, and production of anti-donor antibodies. In vitro, healthy volunteer monocytes stimulated with purified diS-HMGB1 had increased inflammatory cytokine secretion, antigen presentation machinery, and reactive oxygen species production. TLR4 inhibition primarily impeded cytokine/chemokine and costimulatory molecule programs, whereas TLR9 inhibition decreased HLA-DR and reactive oxygen species production. diS-HMGB1-polarized macrophages also showed increased capacity to present antigens and activate T memory cells. In murine OLT, diS-HMGB1 treatment potentiated ischemia-reperfusion-mediated hepatocellular injury, accompanied by increased serum alanine transaminase levels. This translational study identifies the diS-HMGB1/TLR4/TLR9 axis as potential therapeutic targets in OLT-IRI recipients.

Tensor-structured decomposition improves systems serology analysis.

Systems serology provides a broad view of humoral immunity by profiling both the antigen-binding and Fc properties of antibodies. These studies contain structured biophysical profiling across disease-relevant antigen targets, alongside additional measurements made for single antigens or in an antigen-generic manner. Identifying patterns in these measurements helps guide vaccine and therapeutic antibody development, improve our understanding of diseases, and discover conserved regulatory mechanisms. Here, we report that coupled matrix-tensor factorization (CMTF) can reduce these data into consistent patterns by recognizing the intrinsic structure of these data. We use measurements from two previous studies of HIV- and SARS-CoV-2-infected subjects as examples. CMTF outperforms standard methods like principal components analysis in the extent of data reduction while maintaining equivalent prediction of immune functional responses and disease status. Under CMTF, model interpretation improves through effective data reduction, separation of the Fc and antigen-binding effects, and recognition of consistent patterns across individual measurements. Data reduction also helps make prediction models more replicable. Therefore, we propose that CMTF is an effective general strategy for data exploration in systems serology.

Systems approaches to uncovering the contribution of environment-mediated drug resistance.

Cancer drug response is heavily influenced by the extracellular matrix (ECM) environment. Despite a clear appreciation that the ECM influences cancer drug response and progression, a unified view of how, where, and when environment-mediated drug resistance contributes to cancer progression has not coalesced. Here, we survey some specific ways in which the ECM contributes to cancer resistance with a focus on how materials development can coincide with systems biology approaches to better understand and perturb this contribution. We argue that part of the reason that environment-mediated resistance remains a perplexing problem is our lack of a wholistic view of the entire range of environments and their impacts on cell behavior. We cover a series of recent experimental and computational tools that will aid exploration of ECM reactions space, and how they might be synergistically integrated.

Human induced pluripotent stem cell-derived glial cells and neural progenitors display divergent responses to Zika and dengue infections.

Maternal Zika virus (ZIKV) infection during pregnancy is recognized as the cause of an epidemic of microcephaly and other neurological anomalies in human fetuses. It remains unclear how ZIKV accesses the highly vulnerable population of neural progenitors of the fetal central nervous system (CNS), and which cell types of the CNS may be viral reservoirs. In contrast, the related dengue virus (DENV) does not elicit teratogenicity. To model viral interaction with cells of the fetal CNS in vitro, we investigated the tropism of ZIKV and DENV for different induced pluripotent stem cell-derived human cells, with a particular focus on microglia-like cells. We show that ZIKV infected isogenic neural progenitors, astrocytes, and microglia-like cells (pMGLs), but was only cytotoxic to neural progenitors. Infected glial cells propagated ZIKV and maintained ZIKV load over time, leading to viral spread to susceptible cells. DENV triggered stronger immune responses and could be cleared by neural and glial cells more efficiently. pMGLs, when cocultured with neural spheroids, invaded the tissue and, when infected with ZIKV, initiated neural infection. Since microglia derive from primitive macrophages originating in proximity to the maternal vasculature, they may act as a viral reservoir for ZIKV and establish infection of the fetal brain. Infection of immature neural stem cells by invading microglia may occur in the early stages of pregnancy, before angiogenesis in the brain rudiments. Our data are also consistent with ZIKV and DENV affecting the integrity of the blood-brain barrier, thus allowing infection of the brain later in life.

Long-Term maintenance of anomia treatment effects in primary progressive aphasia.

This study examined the maintenance of anomia treatment effects in primary progressive aphasia (PPA). Following baseline testing, a phonological treatment and an orthographic treatment were administered over the course of six months. The treatment stimuli consisted of nouns that were consistently named correctly at baseline (Prophylaxis items) and/or nouns that were consistently named incorrectly at baseline (Remediation items). Naming accuracy was measured at baseline, and it was measured at 1 month, 8 months, and 15 months post-treatment. The change in naming accuracy from baseline to each post-treatment evaluation was calculated within each treatment condition, and within a matched untrained condition. The change in naming accuracy was then compared between the three conditions. The results of these analyses indicate that phonological and orthographic treatments are both effective in the Prophylaxis and Remediation of anomia in all three variants of PPA. For Prophylaxis items, some of the effects of each treatment can persist for as long as 15 months post-treatment. These long-term treatment effects were more robust in the orthographic treatment condition and for participants with the semantic variant of PPA.

Prophylaxis and remediation of anomia in the semantic and logopenic variants of primary progressive aphasia.

This study evaluated the efficacy of phonological and orthographic treatments for anomia in the semantic and logopenic variants of primary progressive aphasia (svPPA and lvPPA, respectively). Both treatments were administered for 6 months. The treatment stimuli consisted of nouns that were consistently named correctly at baseline (prophylaxis items) and/or nouns that were consistently named incorrectly at baseline (remediation items). Oral naming accuracy was measured for trained and untrained picture exemplars, as well as matched items from an untrained condition (UC). Written naming and scene description tasks were also conducted. For all tasks, the change in naming accuracy from baseline to 1 month post-treatment was compared between the UC and each treatment condition. These comparisons indicated that both treatments were effective in the remediation and prophylaxis of anomia in both variants. Furthermore, generalisation to untrained exemplars occurred in both subtypes, whereas item generalisation occurred in lvPPA, and task generalisation was present in svPPA.

ADAM-10 and -17 regulate endometriotic cell migration via concerted ligand and receptor shedding feedback on kinase signaling.

A Disintegrin and Metalloproteinases (ADAMs) are the principal enzymes for shedding receptor tyrosine kinase (RTK) ectodomains and ligands from the cell surface. Multiple layers of activity regulation, feedback, and catalytic promiscuity impede our understanding of context-dependent ADAM "sheddase" function and our ability to predictably target that function in disease. This study uses combined measurement and computational modeling to examine how various growth factor environments influence sheddase activity and cell migration in the invasive disease of endometriosis. We find that ADAM-10 and -17 dynamically integrate numerous signaling pathways to direct cell motility. Data-driven modeling reveals that induced cell migration is a quantitative function of positive feedback through EGF ligand release and negative feedback through RTK shedding. Although sheddase inhibition prevents autocrine ligand shedding and resultant EGF receptor transactivation, it also leads to an accumulation of phosphorylated receptors (HER2, HER4, and MET) on the cell surface, which subsequently enhances Jnk/p38 signaling. Jnk/p38 inhibition reduces cell migration by blocking sheddase activity while additionally preventing the compensatory signaling from accumulated RTKs. In contrast, Mek inhibition reduces ADAM-10 and -17 activities but fails to inhibit compensatory signaling from accumulated RTKs, which actually enhances cell motility in some contexts. Thus, here we present a sheddase-based mechanism of rapidly acquired resistance to Mek inhibition through reduced RTK shedding that can be overcome with rationally directed combination inhibitor treatment. We investigate the clinical relevance of these findings using targeted proteomics of peritoneal fluid from endometriosis patients and find growth-factor-driven ADAM-10 activity and MET shedding are jointly dysregulated with disease.

Censored Least Squares for Imputing Missing Values in PARAFAC Tensor Factorization.

Tensor factorization is a dimensionality reduction method applied to multidimensional arrays. These methods are useful for identifying patterns within a variety of biomedical datasets due to their ability to preserve the organizational structure of experiments and therefore aid in generating meaningful insights. However, missing data in the datasets being analyzed can impose challenges. Tensor factorization can be performed with some level of missing data and reconstruct a complete tensor. However, while tensor methods may impute these missing values, the choice of fitting algorithm may influence the fidelity of these imputations. Previous approaches, based on alternating least squares with prefilled values or direct optimization, suffer from introduced bias or slow computational performance. In this study, we propose that censored least squares can better handle missing values with data structured in tensor form. We ran censored least squares on four different biological datasets and compared its performance against alternating least squares with prefilled values and direct optimization. We used the error of imputation and the ability to infer masked values to benchmark their missing data performance. Censored least squares appeared best suited for the analysis of high-dimensional biological data by accuracy and convergence metrics across several studies.

Baseline Conceptual-Semantic Impairment Predicts Longitudinal Treatment Effects for Anomia in Primary Progressive Aphasia and Alzheimer's Disease.

Background: An individual's diagnostic subtype may fail to predict the efficacy of a given type of treatment for anomia. Classification by conceptual-semantic impairment may be more informative. Aims: This study examined the effects of conceptual-semantic impairment and diagnostic subtype on anomia treatment effects in primary progressive aphasia (PPA) and Alzheimer's disease (AD). Methods & Procedures: At baseline, the picture and word versions of the Pyramids and Palm Trees and Kissing and Dancing tests were used to measure conceptual-semantic processing. Based on norming that was conducted with unimpaired older adults, participants were classified as being impaired on both the picture and word versions (i.e., modality-general conceptual-semantic impairment), the picture version (Objects or Actions) only (i.e., visual-conceptual impairment), the word version (Nouns or Verbs) only (i.e., lexical-semantic impairment), or neither the picture nor the word version (i.e., no impairment). Following baseline testing, a lexical treatment and a semantic treatment were administered to all participants. The treatment stimuli consisted of nouns and verbs that were consistently named correctly at baseline (Prophylaxis items) and/or nouns and verbs that were consistently named incorrectly at baseline (Remediation items). Naming accuracy was measured at baseline, and it was measured at three, seven, eleven, fourteen, eighteen, and twenty-one months. Outcomes & Results: Compared to baseline naming performance, lexical and semantic treatments both improved naming accuracy for treated Remediation nouns and verbs. For Prophylaxis items, lexical treatment was effective for both nouns and verbs, and semantic treatment was effective for verbs, but the pattern of results was different for nouns -- the effect of semantic treatment was initially nonsignificant or marginally significant, but it was significant beginning at 11 Months, suggesting that the effects of prophylactic semantic treatment may become more apparent as the disorder progresses. Furthermore, the interaction between baseline Conceptual-Semantic Impairment and the Treatment Condition (Lexical vs. Semantic) was significant for verb Prophylaxis items at 3 and 18 Months, and it was significant for noun Prophylaxis items at 14 and 18 Months. Conclusions: The pattern of results suggested that individuals who have modality-general conceptual-semantic impairment at baseline are more likely to benefit from lexical treatment, while individuals who have unimpaired conceptual-semantic processing at baseline are more likely to benefit from semantic treatment as the disorder progresses. In contrast to conceptual-semantic impairment, diagnostic subtype did not typically predict the treatment effects.

Tensor modeling of MRSA bacteremia cytokine and transcriptional patterns reveals coordinated, outcome-associated immunological programs.

Methicillin-resistant Staphylococcus aureus (MRSA) bacteremia is a common and life-threatening infection that imposes up to 30% mortality even when appropriate therapy is used. Despite in vitro efficacy determined by minimum inhibitory concentration breakpoints, antibiotics often fail to resolve these infections in vivo, resulting in persistent MRSA bacteremia. Recently, several genetic, epigenetic, and proteomic correlates of persistent outcomes have been identified. However, the extent to which single variables or their composite patterns operate as independent predictors of outcome or reflect shared underlying mechanisms of persistence is unknown. To explore this question, we employed a tensor-based integration of host transcriptional and cytokine datasets across a well-characterized cohort of patients with persistent or resolving MRSA bacteremia outcomes. This method yielded high correlative accuracy with outcomes and immunologic signatures united by transcriptomic and cytokine datasets. Results reveal that patients with persistent MRSA bacteremia (PB) exhibit signals of granulocyte dysfunction, suppressed antigen presentation, and deviated lymphocyte polarization. In contrast, patients with resolving bacteremia (RB) heterogeneously exhibit correlates of robust antigen-presenting cell trafficking and enhanced neutrophil maturation corresponding to appropriate T lymphocyte polarization and B lymphocyte response. These results suggest that transcriptional and cytokine correlates of PB vs. RB outcomes are complex and may not be disclosed by conventional modeling. In this respect, a tensor-based integration approach may help to reveal consensus molecular and cellular mechanisms and their biological interpretation.

Signaling network state predicts twist-mediated effects on breast cell migration across diverse growth factor contexts.

Epithelial-mesenchymal transition (EMT), whether in developmental morphogenesis or malignant transformation, prominently involves modified cell motility behavior. Although major advances have transpired in understanding the molecular pathways regulating the process of EMT induction per se by certain environmental stimuli, an important outstanding question is how the activities of signaling pathways governing motility yield the diverse movement behaviors characteristic of pre-induction versus postinduction states across a broad landscape of growth factor contexts. For the particular case of EMT induction in human mammary cells by ectopic expression of the transcription factor Twist, we found the migration responses to a panel of growth factors (EGF, HRG, IGF, HGF) dramatically disparate between confluent pre-Twist epithelial cells and sparsely distributed post-Twist mesenchymal cells-but that a computational model quantitatively integrating multiple key signaling node activities could nonetheless account for this full range of behavior. Moreover, motility in both conditions was successfully predicted a priori for an additional growth factor (PDGF) treatment. Although this signaling network state model could comprehend motility behavior globally, modulation of the network interactions underlying the altered pathway activities was identified by ascertaining differences in quantitative topological influences among the nodes between the two conditions.

Mixed IgG Fc immune complexes exhibit blended binding profiles and refine FcR affinity estimates.

Immunoglobulin (Ig)G antibodies coordinate immune effector responses by selectively binding to target antigens and then interacting with various effector cells via the Fcγ receptors. The Fc domain of IgG can promote or inhibit distinct effector responses across several different immune cell types through variation based on subclass and Fc domain glycosylation. Extensive characterization of these interactions has revealed how the inclusion of certain Fc subclasses or glycans results in distinct immune responses. During an immune response, however, IgG is produced with mixtures of Fc domain properties, so antigen-IgG immune complexes are likely to almost always be comprised of a combination of Fc forms. Whether and how this mixed composition influences immune effector responses has not been examined. Here, we measured Fcγ receptor binding to immune complexes of mixed Fc domain composition. We found that the binding properties of the mixed-composition immune complexes fell along a continuum between those of the corresponding pure cases. Binding quantitatively matched a mechanistic binding model, except for several low-affinity interactions mostly involving IgG2. We found that the affinities of these interactions are different than previously reported, and that the binding model could be used to provide refined estimates of these affinities. Finally, we demonstrated that the binding model can predict effector-cell elicited platelet depletion in humanized mice, with the model inferring the relevant effector cell populations. Contrary to the previous view in which IgG2 poorly engages with effector populations, we observe appreciable binding through avidity, but insufficient amounts to observe immune effector responses. Overall, this work demonstrates a quantitative framework for reasoning about effector response regulation arising from IgG of mixed Fc composition. Summary points: The binding behavior of mixed Fc immune complexes is a blend of the binding properties for each constituent IgG species.An equilibrium, multivalent binding model can be generalized to incorporate immune complexes of mixed Fc composition.Particularly for low-affinity IgG-Fcγ receptor interactions, immune complexes provide better estimates of affinities.The FcγR binding model predicts effector-elicited cell clearance in humanized mice.

Multivalent, asymmetric IL-2-Fc fusions show enhanced selectivity for regulatory T cells.

The cytokine interleukin-2 (IL-2) has the potential to treat autoimmune disease but is limited by its modest specificity toward immunosuppressive regulatory T (Treg) cells. IL-2 receptors consist of combinations of α, ß, and γ chains of variable affinity and cell specificity. Engineering IL-2 to treat autoimmunity has primarily focused on retaining binding to the relatively Treg-selective, high-affinity receptor while reducing binding to the less selective, low-affinity receptor. However, we found that refining the designs to focus on targeting the high-affinity receptor through avidity effects is key to optimizing Treg selectivity. We profiled the dynamics and dose dependency of signaling responses in primary human immune cells induced by engineered fusions composed of either wild-type IL-2 or mutant forms with altered affinity, valency, and fusion to the antibody Fc region for stability. Treg selectivity and signaling response variations were explained by a model of multivalent binding and dimer-enhanced avidity-a combined measure of the strength, number, and conformation of interaction sites-from which we designed tetravalent IL-2-Fc fusions that had greater Treg selectivity in culture than do current designs. Biasing avidity toward IL2Rα with an asymmetrical multivalent design consisting of one α/ß chain-binding and one α chain-binding mutant further enhanced Treg selectivity. Comparative analysis revealed that IL2Rα was the optimal cell surface target for Treg selectivity, indicating that avidity for IL2Rα may be the optimal route to producing IL-2 variants that selectively target Tregs.

Mixed IgG Fc immune complexes exhibit blended binding profiles and refine FcR affinity estimates.

Immunoglobulin G (IgG) antibodies coordinate immune effector responses by interacting with effector cells via fragment crystallizable γ (Fcγ) receptors. The IgG Fc domain directs effector responses through subclass and glycosylation variation. Although each Fc variant has been extensively characterized in isolation, during immune responses, IgG is almost always produced in Fc mixtures. How this influences effector responses has not been examined. Here, we measure Fcγ receptor binding to mixed Fc immune complexes. Binding of these mixtures falls along a continuum between pure cases and quantitatively matches a mechanistic model, except for several low-affinity interactions mostly involving IgG2. We find that the binding model provides refined estimates of their affinities. Finally, we demonstrate that the model predicts effector cell-elicited platelet depletion in humanized mice. Contrary to previous views, IgG2 exhibits appreciable binding through avidity, though it is insufficient to induce effector responses. Overall, this work demonstrates a quantitative framework for modeling mixed IgG Fc-effector cell regulation.

Live Cell Lineage Tracing of Dormant Cancer Cells.

Breast cancer is a leading cause of global cancer-related deaths, and metastasis is the overwhelming culprit of poor patient prognosis. The most nefarious aspect of metastasis is dormancy, a prolonged period between primary tumor resection and relapse. Current therapies are insufficient at killing dormant cells; thus, they can remain quiescent in the body for decades until eventually undergoing a phenotypic switch, resulting in metastases that are more adaptable and drug resistant. Unfortunately, dormancy has few in vitro models, largely because lab-derived cell lines are highly proliferative. Existing models address tumor dormancy, not cellular dormancy, because tracking individual cells is technically challenging. To combat this problem, a live cell lineage approach to find and track individual dormant cells, distinguishing them from proliferative and dying cells over multiple days, is adapted. This approach is applied across a range of different in vitro microenvironments. This approach reveals that the proportion of cells that exhibit long-term quiescence is regulated by both cell intrinsic and extrinsic factors, with the most dormant cells found in 3D collagen gels. This paper envisions that this approach will prove useful to biologists and bioengineers in the dormancy community to identify, quantify, and study dormant tumor cells.

Analysis and modeling of cancer drug responses using cell cycle phase-specific rate effects.

Identifying effective therapeutic treatment strategies is a major challenge to improving outcomes for patients with breast cancer. To gain a comprehensive understanding of how clinically relevant anti-cancer agents modulate cell cycle progression, here we use genetically engineered breast cancer cell lines to track drug-induced changes in cell number and cell cycle phase to reveal drug-specific cell cycle effects that vary across time. We use a linear chain trick (LCT) computational model, which faithfully captures drug-induced dynamic responses, correctly infers drug effects, and reproduces influences on specific cell cycle phases. We use the LCT model to predict the effects of unseen drug combinations and confirm these in independent validation experiments. Our integrated experimental and modeling approach opens avenues to assess drug responses, predict effective drug combinations, and identify optimal drug sequencing strategies.

Addressing the genetic/nongenetic duality in cancer with systems biology.

The dogma that cancer is a genetic disease is being questioned. Recent findings suggest that genetic/nongenetic duality is necessary for cancer progression. A think tank organized by the Shraman Foundation's Institute for Theoretical Biology compiled key challenges and opportunities that theoreticians, experimentalists, and clinicians can explore from a systems biology perspective to provide a better understanding of the disease as well as help discover new treatment options and therapeutic strategies.