Making human immune systems more interpretable through systems immunology

Abstract

Systems immunology is the study of cell–cell interactions in the immune system, their regulatory functions, and the emergent properties of immune responses.We posit that, despite large interindividual variation, cell composition is nonrandom, and understanding its rules will make human immune systems more predictable in the future.A growing number of computational tools exists for the inference of cell–cell relationships, but no ‘gold standard’ methods exist, and statistical power is often limiting.Minimizing the technical sources of variance by optimal experimental design and sampling procedures allows for more biological ‘signals’ to be discovered. A systems approach is necessary to comprehend the complexity of an organism’s immune system since there is an underlying shared mechanism of cell–cell relationships. Better inferences of cell–cell interactions are essential to increase our understanding of immune responses for various health conditions. A well-planned longitudinal sampling of whole blood is vital in translating in vitro findings to in vivo biology, allowing the systematic tracking of immune system responses over time, and facilitating the development of more accurate models with improved predictability for immune system behaviors. Imagine standing in front of the famous water lilies painting by Monet. Gazing from afar, you admire the carefully chosen colors and the beautiful scenery. However, as you move closer, expecting to see more detail, the patterns dissolve into a disorganized collection of brush strokes, and it becomes difficult to distinguish the water lilies that were so clear from a distance. Just as Monet’s paintings are composed of countless small strokes that combine to form a beautiful painting, the immune system is composed of many individual cells, proteins, and molecules that represent a complex and intricate network of interacting agents. Now, if you are an aspiring artist, you must stand close to the painting and study the individual brush strokes to understand the painting, and how the strokes combine to make up the spectacular network of color. But it is just as important to take a step back, view the whole picture, and think to yourself ‘what does this image represent?’, examining all the components in the painting to truly understand the artist’s motive and reasoning. Like a painting that consists of multiple elements, the immune system consists of many different components and cell populations with specialized functions. Achieving balance in these different populations is key to a healthy immune-system function and the overall health of an individual; indeed, these cell populations activate and inhibit each other, compete for growth factors and ligands, move in and out of compartmentalized organs with different regulatory logic and, while doing so, must maintain their regulatory balance and tune their thresholds for activation to remain tolerant to surrounding cells and tissues, yet responsive to relative changes in potentially harmful processes and targets [1.Parkin J. Cohen B. An overview of the immune system.Lancet. 2001; 357: 1777-1789Abstract Full Text Full Text PDF PubMed Scopus (791) Google Scholar, 2.Brodin P. et al.NK cell education: not an on-off switch but a tunable rheostat.Trends Immunol. 2009; 4: 143-149Abstract Full Text Full Text PDF Scopus (188) Google Scholar, 3.Grossman Z. Paul W.E. Dynamic tuning of lymphocytes: physiological basis, mechanisms, and function.Annu. Rev. Immunol. 2015; 1: 677-713Crossref Scopus (63) Google Scholar]. The immune system is a compelling and apt subject for a systems biology approach to research, as it is a distributed, multiagent system with emergent properties. It is a complex adaptive system (CAS), self-organized, and regulated by dynamical interactions of individual cells with a capacity to respond to environmental changes both collectively and individually [4.Mitleton-Kelly E. Complex Systems and Evolutionary Perspectives on Organisations: the Application of Complexity Theory to Organisations. Pergamon, 2003Google Scholar, 5.Holland J.H. Signals and Boundaries: Building Blocks for Complex Adaptive Systems. MIT Press, 2014Google Scholar, 6.Ahmed E. Hashish A. On modelling the immune system as a complex system.Theory Biosci. 2006; 124: 413-418Crossref PubMed Scopus (22) Google Scholar]. Therefore, to understand the higher-order functions of this complex system, comprehensive analyses involving all cell populations are simultaneously required. Only then can we fully understand the consequences of activating one or several cell populations on the system, just like taking a step back to see the water lilies more clearly. In this opinion article, we introduce the concept of systems immunology and discuss important factors that must be considered, both methodological and analytical, for the successful application of this strategy. We also present examples of the clinical impact that systems immunology can have, our general hypothesis being that only a holistic approach can provide a full understanding of immune system function in health and disease. The fact that the immune system is structured as a CAS does not mean that the system lacks order or structure among its various components. There are certainly logical regulatory mechanisms at play as well, although most of these remain unknown. From an evolutionary perspective, our survival in an environment full of microbes has depended on a potent defense system, and microbes have played a dominant role in shaping it [7.Quintana-Murci L. et al.Immunology in natura: clinical, epidemiological and evolutionary genetics of infectious diseases.Nat. Immunol. 2007; 8: 1165-1171Crossref PubMed Scopus (144) Google Scholar]. Recent advances in single-cell biology have equipped us with the tools to perform systems immunology analyses capturing all the immune-cell populations present within a tissue sample and some functional properties of each cell, thereby allowing us to examine relationships among immune-cell populations at scale for the first time (Figure 1A, Key figure ). When seeing correlated changes among two cell populations, these could either indicate shared or coordinated responses or a shared influence by a third influencing cell population (Figure 1B). Such possible relationships are impossible to disentangle without further mechanistic experiments, although longitudinal monitoring of blood, when possible, permits more causal modeling of cell–cell relationships from time series data [8.Lakshmikanth T. et al.Human immune system variation during 1 year.Cell Rep. 2020; 32107923Abstract Full Text Full Text PDF PubMed Scopus (18) Google Scholar]. Our laboratory’s internal data illustrate the nonrandom and reproducible correlation structure among cell types. For illustration purposes: in preliminary work (which remains to be validated), we analyzed 1644 blood samples from healthy individuals using a 56-parameter mass cytometry panel; we noticed strong correlations between the frequencies of multiple cell types in circulation (Figure 1C) (Forlin et al., unpublished). For example, there is a strong inverse correlation between neutrophil abundance and the abundance of natural killer (NK) cells and B cells, and conversely, strong positive correlations between plasmacytoid dendritic cells (pDCs) and both eosinophils and monocytes (Figure 1C). This finding suggests that shared regulatory mechanisms exist among these various cell types and among individuals, preventing a defense system in complete chaos. This finding is particularly interesting because the immune system is unique in each individual, shaped by both heritable and nonheritable factors (such as the environment and upbringing, microbiota, and age) [9.Carr E.J. et al.The cellular composition of the human immune system is shaped by age and cohabitation.Nat. Immunol. 2016; 4: 461-468Crossref Scopus (188) Google Scholar, 10.Reichert T. et al.Lymphocyte subset reference ranges in adult Caucasians.Clin. Immunol. Immunopathol. 1991; 2: 190-208Crossref Scopus (289) Google Scholar, 11.Tollerud D.J. et al.The influence of age, race, and gender on peripheral blood mononuclear-cell subsets in healthy nonsmokers.J. Clin. Immunol. 1989; 3: 214-222Crossref Scopus (142) Google Scholar, 12.Hill D.L. et al.Immune system development varies according to age, location, and anemia in African children.Sci. Transl. Med. 2020; 12eaaw9522Crossref Scopus (40) Google Scholar, 13.Schluter J. et al.The gut microbiota is associated with immune cell dynamics in humans.Nature. 2020; 588: 303-307Crossref PubMed Scopus (184) Google Scholar]. Consequently, a considerable amount of variability in immunological profiles across individuals exists. Many characteristics of the immune system – such as cell population frequencies, cytokine responses, and serum proteins – are impacted by nonheritable influences such as age and cytomegalovirus seropositivity [14.The Milieu Intérieur Consortium et al.Natural variation in the parameters of innate immune cells is preferentially driven by genetic factors.Nat. Immunol. 2018; 3: 302-314Google Scholar,15.Brodin P. et al.Variation in the human immune system is largely driven by non-heritable influences.Cell. 2015; 1–2: 37-47Abstract Full Text Full Text PDF Scopus (641) Google Scholar]. Some parameters become more variable with age – such as NK and regulatory T cell (Treg) cell frequencies, as well as interleukin (IL)-15 and IL-17 serum protein concentrations – suggesting a cumulative influence of environmental exposures [15.Brodin P. et al.Variation in the human immune system is largely driven by non-heritable influences.Cell. 2015; 1–2: 37-47Abstract Full Text Full Text PDF Scopus (641) Google Scholar]. Other factors display a strong heritable influence, such as the matching of phenotype KIR/LIR (killer inhibitory receptor/leukocyte Ig-like receptor) molecules with the major histocompatibility complex class I alleles in NK cells, CD39 expression on Tregs, and expression of CD32 on myeloid dendritic cells [16.Mangino M. et al.Innate and adaptive immune traits are differentially affected by genetic and environmental factors.Nat. Commun. 2017; 8: 13850Crossref PubMed Scopus (77) Google Scholar,17.Roederer M. et al.The genetic architecture of the human immune system: a bioresource for autoimmunity and disease pathogenesis.Cell. 2015; 2: 387-403Abstract Full Text Full Text PDF Scopus (194) Google Scholar]. While there is variation in immune-cell composition among individuals [9.Carr E.J. et al.The cellular composition of the human immune system is shaped by age and cohabitation.Nat. Immunol. 2016; 4: 461-468Crossref Scopus (188) Google Scholar,15.Brodin P. et al.Variation in the human immune system is largely driven by non-heritable influences.Cell. 2015; 1–2: 37-47Abstract Full Text Full Text PDF Scopus (641) Google Scholar,18.Tsang J.S. et al.Global analyses of human immune variation reveal baseline predictors of postvaccination responses.Cell. 2014; 2: 499-513Abstract Full Text Full Text PDF Scopus (295) Google Scholar], the variation of cell frequencies within an individual is mostly stable over a time-scale of months, in the absence of infections or other perturbations [8.Lakshmikanth T. et al.Human immune system variation during 1 year.Cell Rep. 2020; 32107923Abstract Full Text Full Text PDF PubMed Scopus (18) Google Scholar,19.Brodin P. Davis M.M. Human immune system variation.Nat. Rev. Immunol. 2017; 1: 21-29Crossref Scopus (344) Google Scholar]. One of the major goals of systems immunology is to increase the predictability of the immune system, and thereby contribute to the advancement of personalized medicine, vaccine development, and other clinical therapies. It can also provide a better understanding of the pathology of various immune-related diseases, such as cancer, as well as autoimmune and infectious diseases [20.Germain R.N. Schwartzberg P.L. The human condition: an immunological perspective.Nat. Immunol. 2011; 5: 369-372Crossref Scopus (16) Google Scholar,21.Germain R.N. Will systems biology deliver its promise and contribute to the development of new or improved vaccines?: what really constitutes the study of “systems biology” and how might such an approach facilitate vaccine design.Cold Spring Harb. Perspect. Biol. 2018; 10a033308Crossref PubMed Scopus (11) Google Scholar]. Numerous advances have been made towards this goal in recent years. For example, one study [22.Kaczorowski K.J. et al.Continuous immunotypes describe human immune variation and predict diverse responses.Proc. Natl. Acad. Sci. U. S. A. 2017; 114: E6097-E6106Crossref PubMed Scopus (71) Google Scholar] showed that by using the baseline composition of immune cells present in preparations of human peripheral blood mononuclear cells (PBMCs), three latent vectors describing the variance in immune-cell composition were predictive of functional responses – measured as phosphorylation of signal transducer and activator of transcription (STAT)1, STAT3, and STAT5 – to a variety of cytokines – IL-2, IL-6, IL-7, IL-10, IL-21, interferon (IFN)-α, and IFN-γ – exogenously to in has also been to an immune that can predict responses and has been made in using human baseline populations – such as stable B cell to predict concentrations as an that does not on J.S. et al.Global analyses of human immune variation reveal baseline predictors of postvaccination responses.Cell. 2014; 2: 499-513Abstract Full Text Full Text PDF Scopus (295) Google et of to across multiple and in diverse populations shared 2015; Full Text Full Text PDF Scopus Google et immune activation shared for vaccine in healthy individuals and disease in with Med. 2020; 4: Scopus Google Scholar]. The of such findings are that stable immune can be that are predictive of responses to and other immune In one study S. et innate immune predictive of responses to Immunol. Scopus Google Scholar], three were among to individuals with an showed serum responses 1 This to the different and B further that monocytes and myeloid dendritic cells were sources of this when with single-cell S. et innate immune predictive of responses to Immunol. Scopus Google Scholar]. other work et of human and single-cell vaccine are Google Scholar] showed that perturbations in various cell types between and myeloid dendritic cells and monocytes in blood these were to be two of the key of the postvaccination of as well as monocytes and associated T In the of responses towards or one et immune activation shared for vaccine in healthy individuals and disease in with Med. 2020; 4: Scopus Google Scholar] a of and that stable and correlated with the of cells, predictive of an in human blood, measured as J.S. et al.Global analyses of human immune variation reveal baseline predictors of postvaccination responses.Cell. 2014; 2: 499-513Abstract Full Text Full Text PDF Scopus (295) Google Scholar]. This that the B cell composition of different subsets is more or responsive to of the of the B This to as an of the immune and predictive of responses to either one of the among healthy individuals, the of and suggesting a for personalized on immune The also correlated with the of in a of with when during of or no disease This that not be as predictive but that mechanisms between might a predictive power to a of disease or immune system characteristics between individuals, the mechanisms underlying cell–cell interactions and relationships to be to shared mechanisms the of or that are shared across can predict the of the immune system and to the system as a is one of the of systems The mechanisms underlying these shared immunological properties remain The immune system is a system with some of are general for a of and The and technical in single-cell data have to an in cell–cell there is a growing number of computational tools these S. et and of cell–cell using Commun. PubMed Scopus Google Scholar, M. et cell–cell from expression of 2020; 4: Scopus Google Scholar, et modeling by to 2020; 2: Scopus Google Scholar, et from longitudinal expression as approach for 2016; Scopus Google Scholar, T. et of single-cell Full Text Full Text PDF PubMed Scopus Google Scholar], of cell–cell interactions from data more This is essential for of we see what we see at a for each of the computational tools is that on of interactions and a of cell–cell between different subsets of cells on the expression of a and a However, the for the and the methods to cell–cell among different S. et and of cell–cell using Commun. PubMed Scopus Google Scholar, M. et cell–cell from expression of 2020; 4: Scopus Google Scholar, et modeling by to 2020; 2: Scopus Google Scholar, et from longitudinal expression as approach for 2016; Scopus Google Scholar, T. et of single-cell Full Text Full Text PDF PubMed Scopus Google Scholar] as no exists. While the methods to technical the of both and is to impact the et of methods and for cell–cell inference from single-cell Commun. 1: Scopus Google Scholar]. Therefore, as of these interactions to be with but can be to systems to provide a understanding of cell–cell interactions is necessary and be of to systems between two or more cell types a cell et of cell with Natl. Acad. Sci. U. S. A. PubMed Scopus Google et design of a stable 2018; 4: Full Text Full Text PDF Scopus Google Scholar], and in a system, there are multiple This in a study et design of a stable 2018; 4: Full Text Full Text PDF Scopus Google Scholar] in vitro cellular systems were to growth between and in the two cell types with to the population This population also to such as 1 growth B or despite the being their capacity is often the with in vivo the were to of between two cell types in an and system et design of a stable 2018; 4: Full Text Full Text PDF Scopus Google Scholar]. have also a unique the to for interactions between J. et for the human immune PubMed Scopus Google Scholar]. that of human immune are that protein other This more in proteins for such as and its and A while the several and other From the also a that, while being could relative frequencies at human immune cells The accurate to be with data on the of vivo cell–cell interactions among human peripheral mononuclear et al.Global of the of Biol. 2017; PubMed Scopus (40) Google Scholar], for the of a collection of also that immune activation the of cellular interactions in an and interactions in a the for responses when J. et for the human immune PubMed Scopus Google Scholar]. from a recent et cellular to in PubMed Scopus Google Scholar] that cellular is not only by and cell (such as growth but also by cellular to be at as important as the for how cells when network models of the immune system, it is important to the and of each of a (such as growth the from a subset of cells can also the cellular such as are in an or It is to a such as these et design of a stable 2018; 4: Full Text Full Text PDF Scopus Google J. et for the human immune PubMed Scopus Google et cellular to in PubMed Scopus Google Scholar] could of interactions might and what will when a cell is in a a design of all can be remains to be within this and models are often further into the design of cell–cell relationships to predict the of cells, or subsets of cells, will be of for a understanding of immune functions. A key of the modeling is that the of a to in vivo biology on the underlying data as in the Therefore, it is vital to the of the data to a and to that it is for the application and A is the statistical is often when of made the samples It is essential to technical sources of variation in the data to the of capturing biological of these statistical power Our recent have that the of is a major of technical variation when cell by mass cytometry to whole blood as well as cytokine responses measured by P. et for blood – in human Full Text Full Text PDF Scopus Google Scholar]. preparations also to the of the cell population in blood with important regulatory and functions during infections and conditions. and are during can impact the variability in cell composition and concentrations of proteins many cell–cell regulatory relationships. Consequently, functional responses to in whole blood be more reproducible and physiological P. et for blood – in human Full Text Full Text PDF Scopus Google Scholar, et whole blood of immune responses in Immunol. 2017; PubMed Scopus Google Scholar, et of T cells in whole blood Immunol. 2006; PubMed Scopus (40) Google Scholar]. in because populations exist only of the – in either or – we posit that whole blood is the most tissue for blood analyses and a more of the in vivo blood analyses are not of and immune-cell interactions with cells, or other immune cells in et for a Full Text Full Text PDF Scopus Google Scholar]. this of recent in (such as for modeling adaptive immune responses important promise for systems immunology in vitro and can mechanistic in human cells et human adaptive immune responses with Med. 1: Scopus Google Scholar]. longitudinal sampling of or longitudinal analyses of can variability over time and distinguish technical sources of variation and from biological of and in to of biological – and in – with possible impact on immune these are from sampling also inferences to be made or protein expression over time and their possible causal relationships and However, a key when such inferences from longitudinal samples is to the technical sources of variation, such that changes in cell population or functional are with of factors to to technical variation the of for sample collection and and to as many to from It is particularly important to combine all longitudinal samples from a a subject in the that are in each experimental such as or be all at for each such sources of technical variation might be an for what types of inference methods or modeling are to be must be at the experimental design and to sample to for technical variation, statistical power in and for clear biological A is to samples first and then to all the data are often to and in clear from complex In we that we have the of taking a step back and the immune system as a methodological to biological data for such as longitudinal We have also a of the many recent advances and possible to make the immune system more systems such as on cell–cell interactions and modeling the cell–cell population systems immunology like a we that by doing so many of the surrounding immune function might be This can have an important clinical impact on vaccine development and infectious diseases, autoimmune cancer, and and of provide the most for understanding immune-cell and inference methods are optimal for modeling cell–cell and how data is are cell–cell regulatory mechanisms across individual populations of different and and of provide the most for understanding immune-cell and inference methods are optimal for modeling cell–cell and how data is are cell–cell regulatory mechanisms across individual populations of different and sources are to to and to is a of of of and and

Topics & Concepts