Search

Macrophages sense pathogens and orchestrate specific immune responses. Stimulus specificity is thought to be achieved through combinatorial and dynamical coding by signaling pathways. While NFκB dynamics are known to encode stimulus information, dynamical coding in other signaling pathways and their combinatorial coordination remain unclear. Here, we established live-cell microscopy to investigate how NFκB and p38 dynamics interface in stimulated macrophages. Information theory and machine learning revealed that p38 dynamics distinguish cytokine TNF from pathogen-associated molecular patterns and high doses from low, but contributed little to information-rich NFκB dynamics when both pathways are considered. This suggests that immune response genes benefit from decoding immune signaling dynamics or combinatorics, but not both. We found that the heterogeneity of the two pathways is surprisingly uncorrelated. Mathematical modeling revealed potential sources of uncorrelated heterogeneity in the branched pathway network topology and predicted it to drive gene expression variability. Indeed, genes dependent on both p38 and NFκB showed high scRNAseq variability and bimodality. These results identify combinatorial signaling as a mechanism to restrict NFκB-AND-p38-responsive inflammatory cytokine expression to few cells.

We developed an on-slide decellularization approach to generate acellular extracellular matrix (ECM) myoscaffolds that can be repopulated with various cell types to interrogate cell-ECM interactions. Using this platform, we investigated whether fibrotic ECM scarring affected human skeletal muscle progenitor cell (SMPC) functions that are essential for myoregeneration. SMPCs exhibited robust adhesion, motility, and differentiation on healthy muscle-derived myoscaffolds. All SPMC interactions with fibrotic myoscaffolds from dystrophic muscle were severely blunted including reduced motility rate and migration. Furthermore, SMPCs were unable to remodel laminin dense fibrotic scars within diseased myoscaffolds. Proteomics and structural analysis revealed that excessive collagen deposition alone is not pathological, and can be compensatory, as revealed by overexpression of sarcospan and its associated ECM receptors in dystrophic muscle. Our in vivo data also supported that ECM remodeling is important for SMPC engraftment and that fibrotic scars may represent one barrier to efficient cell therapy.

Quantifying the dependency between mRNA abundance and downstream cellular phenotypes is a fundamental open problem in biology. Advances in multimodal single-cell measurement technologies provide an opportunity to apply new computational frameworks to dissect the contribution of individual genes and gene combinations to a given phenotype. Using an information theory approach, we analyzed multimodal data of the expression of 83 genes in the Ca2+ signaling network and the dynamic Ca2+ response in the same cell. We found that the overall expression levels of these 83 genes explain approximately 60% of Ca2+ signal entropy. The average contribution of each single gene was 17%, revealing a large degree of redundancy between genes. Using different heuristics, we estimated the dependency between the size of a gene set and its information content, revealing that on average, a set of 53 genes contains 54% of the information about Ca2+ signaling. Our results provide the first direct quantification of information content about complex cellular phenotype that exists in mRNA abundance measurements.

MotivationSingle-cell RNA sequencing (scRNA-seq) captures whole transcriptome information of individual cells. While scRNA-seq measures thousands of genes, researchers are often interested in only dozens to hundreds of genes for a closer study. Then, a question is how to select those informative genes from scRNA-seq data. Moreover, single-cell targeted gene profiling technologies are gaining popularity for their low costs, high sensitivity and extra (e.g. spatial) information; however, they typically can only measure up to a few hundred genes. Then another challenging question is how to select genes for targeted gene profiling based on existing scRNA-seq data.ResultsHere, we develop the single-cell Projective Non-negative Matrix Factorization (scPNMF) method to select informative genes from scRNA-seq data in an unsupervised way. Compared with existing gene selection methods, scPNMF has two advantages. First, its selected informative genes can better distinguish cell types. Second, it enables the alignment of new targeted gene profiling data with reference data in a low-dimensional space to facilitate the prediction of cell types in the new data. Technically, scPNMF modifies the PNMF algorithm for gene selection by changing the initialization and adding a basis selection step, which selects informative bases to distinguish cell types. We demonstrate that scPNMF outperforms the state-of-the-art gene selection methods on diverse scRNA-seq datasets. Moreover, we show that scPNMF can guide the design of targeted gene profiling experiments and the cell-type annotation on targeted gene profiling data.Availability and implementationThe R package is open-access and available at https://github.com/JSB-UCLA/scPNMF. The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997.Supplementary informationSupplementary data are available at Bioinformatics online.

RNA hybridization-based spatial transcriptomics provides unparalleled detection sensitivity. However, inaccuracies in segmentation of image volumes into cells cause misassignment of mRNAs which is a major source of errors. Here, we develop JSTA, a computational framework for joint cell segmentation and cell type annotation that utilizes prior knowledge of cell type-specific gene expression. Simulation results show that leveraging existing cell type taxonomy increases RNA assignment accuracy by more than 45%. Using JSTA, we were able to classify cells in the mouse hippocampus into 133 (sub)types revealing the spatial organization of CA1, CA3, and Sst neuron subtypes. Analysis of within cell subtype spatial differential gene expression of 80 candidate genes identified 63 with statistically significant spatial differential gene expression across 61 (sub)types. Overall, our work demonstrates that known cell type expression patterns can be leveraged to improve the accuracy of RNA hybridization-based spatial transcriptomics while providing highly granular cell (sub)type information. The large number of newly discovered spatial gene expression patterns substantiates the need for accurate spatial transcriptomic measurements that can provide information beyond cell (sub)type labels.

Rapid death of infected cells is an important antiviral strategy. However, fast decisions that are based on limited evidence can be erroneous and cause unnecessary cell death and subsequent tissue damage. How cells optimize their death decision making strategy to maximize both speed and accuracy is unclear. Here, we show that exposure to TNF, which is secreted by macrophages during viral infection, causes cells to change their decision strategy from "slow and accurate" to "fast and error-prone". Mathematical modeling combined with experiments in cell culture and whole organ culture show that the regulation of the cell death decision strategy is critical to prevent HSV-1 spread. These findings demonstrate that immune regulation of cellular cognitive processes dynamically changes a tissues' tolerance for self-damage, which is required to protect against viral spread.

Cellular responses to environmental changes are encoded in the complex temporal patterns of signaling proteins. However, quantifying the accumulation of information over time to direct cellular decision-making remains an unsolved challenge. This is, in part, due to the combinatorial explosion of possible configurations that need to be evaluated for information in time-course measurements. Here, we develop a quantitative framework, based on inferred trajectory probabilities, to calculate the mutual information encoded in signaling dynamics while accounting for cell-cell variability. We use it to understand NFκB transcriptional dynamics in response to different immune threats, and reveal that some threats are distinguished faster than others. Our analyses also suggest specific temporal phases during which information distinguishing threats becomes available to immune response genes; one specific phase could be mapped to the functionality of the IκBα negative feedback circuit. The framework is generally applicable to single-cell time series measurements, and enables understanding how temporal regulatory codes transmit information over time.

Balancing cell death is essential to maintain healthy tissue homeostasis and prevent disease. Tumor necrosis factor (TNF) not only activates nuclear factor κB (NFκB), which coordinates the cellular response to inflammation, but may also trigger necroptosis, a pro-inflammatory form of cell death. Whether TNF-induced NFκB affects the fate decision to undergo TNF-induced necroptosis is unclear. Live-cell microscopy and model-aided analysis of death kinetics identified a molecular circuit that interprets TNF-induced NFκB/RelA dynamics to control necroptosis decisions. Inducible expression of TNFAIP3/A20 forms an incoherent feedforward loop to interfere with the RIPK3-containing necrosome complex and protect a fraction of cells from transient, but not long-term TNF exposure. Furthermore, dysregulated NFκB dynamics often associated with disease diminish TNF-induced necroptosis. Our results suggest that TNF's dual roles in either coordinating cellular responses to inflammation, or further amplifying inflammation are determined by a dynamic NFκB-A20-RIPK3 circuit, that could be targeted to treat inflammation and cancer.

Gene expression variability, differences in the number of mRNA per cell across a population of cells, is ubiquitous across diverse organisms with broad impacts on cellular phenotypes. The role of chromatin in regulating average gene expression has been extensively studied. However, what aspects of the chromatin contribute to gene expression variability is still underexplored. Here we addressed this problem by leveraging chromatin diversity and using a systematic investigation of randomly integrated expression reporters to identify what aspects of chromatin microenvironment contribute to gene expression variability. Using DNA barcoding and split-pool decoding, we created a large library of isogenic reporter clones and identified reporter integration sites in a massive and parallel manner. By mapping our measurements of reporter expression at different genomic loci with multiple epigenetic profiles including the enrichment of transcription factors and the distance to different chromatin states, we identified new factors that impact the regulation of gene expression distributions.

The dynamic signal encoding paradigm suggests that information flows from the extracellular environment into specific signaling patterns (encoding) that are then read by downstream effectors to control cellular behavior. Previous work empirically quantified the information content of dynamic signaling patterns. However, whether this information can be faithfully transmitted to the gene expression level is unclear. Here we used NFkB signaling as a model to understand the accuracy of information transmission from signaling dynamics into gene expression. Using a detailed mathematical model, we simulated realistic NFkB signaling patterns with different degrees of variability. The NFkB patterns were used as an input to a simple gene expression model. Analysis of information transmission between ligand and NFkB and ligand and gene expression allows us to determine information loss in transmission between receptors to dynamic signaling patterns and between signaling dynamics to gene expression. Information loss could occur due to biochemical noise or due to a lack of specificity. We found that noise-free gene expression has very little information loss suggesting that gene expression can preserve specificity in NFkB patterns. As expected, the addition of noise to the gene expression model results in information loss. Interestingly, this effect can be mitigated by a specific choice of parameters that can substantially reduce information loss due to biochemical noise during gene expression. Overall our results show that the cellular capacity for information transmission from dynamic signaling patterns to gene expression can be high enough to preserve ligand specificity and thereby the accuracy of cellular response to environmental cues.

Therapeutic targeting of epigenetic modulators offers a novel approach to the treatment of multiple diseases. The cellular consequences of chemical compounds that target epigenetic regulators (epi-drugs) are complex. Epi-drugs affect global cellular phenotypes and cause local changes to gene expression due to alteration of a gene chromatin environment. Despite increasing use in the clinic, the mechanisms responsible for cellular changes are unclear. Specifically, to what degree the effects are a result of cell-wide changes or disease related locus specific effects is unknown. Here we developed a platform to systematically and simultaneously investigate the sensitivity of epi-drugs at hundreds of genomic locations by combining DNA barcoding, unique split-pool encoding, and single cell expression measurements. Internal controls are used to isolate locus specific effects separately from any global consequences these drugs have. Using this platform we discovered wide-spread loci specific sensitivities to epi-drugs for three distinct epi-drugs that target histone deacetylase, DNA methylation and bromodomain proteins. By leveraging ENCODE data on chromatin modification, we identified features of chromatin environments that are most likely to be affected by epi-drugs. The measurements of loci specific epi-drugs sensitivities will pave the way to the development of targeted therapy for personalized medicine.

During early development, extrinsic triggers prompt pluripotent cells to begin the process of differentiation. When and how human embryonic stem cells (hESCs) irreversibly commit to differentiation is a fundamental yet unanswered question. By combining single-cell imaging, genomic approaches, and mathematical modeling, we find that hESCs commit to exiting pluripotency unexpectedly early. We show that bone morphogenetic protein 4 (BMP4), an important differentiation trigger, induces a subset of early genes to mirror the sustained, bistable dynamics of upstream signaling. Induction of one of these genes, GATA3, drives differentiation in the absence of BMP4. Conversely, GATA3 knockout delays differentiation and prevents fast commitment to differentiation. We show that positive feedback at the level of the GATA3-BMP4 axis induces fast, irreversible commitment to differentiation. We propose that early commitment may be a feature of BMP-driven fate choices and that interlinked feedback is the molecular basis for an irreversible transition from pluripotency to differentiation.

Gene expression variability in mammalian systems plays an important role in physiological and pathophysiological conditions. This variability can come from differential regulation related to cell state (extrinsic) and allele-specific transcriptional bursting (intrinsic). Yet, the relative contribution of these two distinct sources is unknown. Here, we exploit the qualitative difference in the patterns of covariance between these two sources to quantify their relative contributions to expression variance in mammalian cells. Using multiplexed error robust RNA fluorescent in situ hybridization (MERFISH), we measured the multivariate gene expression distribution of 150 genes related to Ca2+ signaling coupled with the dynamic Ca2+ response of live cells to ATP. We show that after controlling for cellular phenotypic states such as size, cell cycle stage, and Ca2+ response to ATP, the remaining variability is effectively at the Poisson limit for most genes. These findings demonstrate that the majority of expression variability results from cell state differences and that the contribution of transcriptional bursting is relatively minimal.

Cell growth is controlled by a lysosomal signalling complex containing Rag small GTPases and mammalian target of rapamycin complex 1 (mTORC1) kinase. Here, we carried out a microscopy-based genome-wide human short interfering RNA screen and discovered a lysosome-localized G protein-coupled receptor (GPCR)-like protein, GPR137B, that interacts with Rag GTPases, increases Rag localization and activity, and thereby regulates mTORC1 translocation and activity. High GPR137B expression can recruit and activate mTORC1 in the absence of amino acids. Furthermore, GPR137B also regulates the dissociation of activated Rag from lysosomes, suggesting that GPR137B controls a cycle of Rag activation and dissociation from lysosomes. GPR137B-knockout cells exhibited defective autophagy and an expanded lysosome compartment, similar to Rag-knockout cells. Like zebrafish RagA mutants, GPR137B-mutant zebrafish had upregulated TFEB target gene expression and an expanded lysosome compartment in microglia. Thus, GPR137B is a GPCR-like lysosomal regulatory protein that controls dynamic Rag and mTORC1 localization and activity as well as lysosome morphology.

The robustness of biological systems is often depicted as a key system-level emergent property that allows uniform phenotypes in fluctuating environments. Yet, analysis of single-cell signaling responses identified multiple examples of cellular responses with high degrees of heterogeneity. Here we discuss the implications of the observed lack of response accuracy in the context of new observations coming from single-cell approaches. Single-cell approaches provide a new way to measure the abundance of thousands of molecular species in a single-cell. Repeatedly, analysis of cell distributions identifies clusters within these distributions where cells can be grouped into specific cell states. If cells in a population occupy distinct cell states, the observed variable response could in fact be accurate for each cell conditioned on its own internal state. In this view, the observed lack of accuracy, i.e. response heterogeneity, could in fact be beneficial and a potentially regulated feature of cell state variability. Therefore, to truly determine whether the observed response heterogeneity is a desired property or a physical limitation, future analysis of signaling heterogeneity must take into account the internal states of cells in the population.

During early infection, the HIV virus makes a key decision between two states: lytic and lysogenic fate. Deterministic bistability requires combination of positive feedback and ultrasensitivity. Although HIV circuity includes positive feedback activation of the Tat transactivator, it lacks ultrasensitivity. How does the HIV circuit allow for multiple fates without ultrasensitivity? A new article suggests that HIV bistability is a result of a transient threshold that allows the kinetic trapping of the inactive state. Interestingly, the model shows that the transient threshold is a result of a single molecule threshold that occurs when the promoter toggles between inactive and active states.

Damage-associated molecular patterns (DAMPs) are critical mediators of information concerning tissue damage from damaged cells to neighboring healthy cells. ATP acts as an effective DAMP when released into extracellular space from damaged cells. Extracellular ATP receptors monitor tissue damage and activate a Ca2+ wave in the surrounding healthy cells. How the Ca2+ wave propagates through cells after a wound is unclear. Ca2+ wave activation can occur extracellularly via external receptors or intracellularly through GAP junctions. Three potential mechanisms to propagate the Ca2+ wave are source and sink, amplifying wave, and release and diffusion. Both source and sink and amplifying wave regulate ATP levels using hydrolysis or secretion, respectively, whereas release and diffusion relies on dilution. Here we systematically test these hypotheses using a microfluidics assay to mechanically wound an epithelial monolayer in combination with direct manipulation of ATP hydrolysis and release. We show that a release and diffusion model sufficiently explains Ca2+-wave propagation after an epithelial wound. A release and diffusion model combines the benefits of fast activation at short length scales with a self-limiting response to prevent unnecessary inflammatory responses harmful to the organism.

Protein engineering is a powerful tool in drug design and therapeutics, where disulphide bridges are commonly introduced to stabilize proteins. However, these bonds also introduce covalent loops, which are often neglected. These loops may entrap the protein backbone on opposite sides, leading to a "knotted" topology, forming a so-called Pierced Lasso (PL). In this elegant system, the "knot" is held together with a single disulphide bridge where part of the polypeptide chain is threaded through. The size and position of these covalent loops can be manipulated through protein design in vitro, whereas nature uses polymorphism to switch the PL topology. The PL protein leptin shows genetic modification of an N-terminal residue, adding a third cysteine to the same sequence. In an effort to understand the mechanism of threading of these diverse topologies, we designed three loop variants to mimic the polymorphic sequence. This adds elegance to the system under study, as it allows the generation of three possible covalent loops; they are the original wild-type C-terminal loop protein, the fully circularized unthreaded protein, and the N-terminal loop protein, responsible for different lasso topologies. The size of the loop changes the threading mechanism from a slipknotting to a plugging mechanism, with increasing loop size. Interestingly, the ground state of the native protein structure is largely unaffected, but biological assays show that the activity is maximized by properly controlled dynamics in the threaded state. A threaded topology with proper conformational dynamics is important for receptor interaction and activation of the signaling pathways in vivo.

The heterogeneity in mammalian cells signaling response is largely a result of pre-existing cell-to-cell variability. It is unknown whether cell-to-cell variability rises from biochemical stochastic fluctuations or distinct cellular states. Here, we utilize calcium response to adenosine trisphosphate as a model for investigating the structure of heterogeneity within a population of cells and analyze whether distinct cellular response states coexist. We use a functional definition of cellular state that is based on a mechanistic dynamical systems model of calcium signaling. Using Bayesian parameter inference, we obtain high confidence parameter value distributions for several hundred cells, each fitted individually. Clustering the inferred parameter distributions revealed three major distinct cellular states within the population. The existence of distinct cellular states raises the possibility that the observed variability in response is a result of structured heterogeneity between cells. The inferred parameter distribution predicts, and experiments confirm that variability in IP3R response explains the majority of calcium heterogeneity. Our work shows how mechanistic models and single-cell parameter fitting can uncover hidden population structure and demonstrate the need for parameter inference at the single-cell level.

Signal transduction, or how cells interpret and react to external events, is a fundamental aspect of cellular function. Traditional study of signal transduction pathways involves mapping cellular signaling pathways at the population level. However, population-averaged readouts do not adequately illuminate the complex dynamics and heterogeneous responses found at the single-cell level. Recent technological advances that observe cellular response, computationally model signaling pathways, and experimentally manipulate cells now enable studying signal transduction at the single-cell level. These studies will enable deeper insights into the dynamic nature of signaling networks.

Collective behavior emerges from interactions among group members who often vary in their behavior. The presence of just one or a few keystone individuals, such as leaders or tutors, may have a large effect on collective outcomes. These individuals can catalyze behavioral changes in other group members, thus altering group composition and collective behavior. The influence of keystone individuals on group function may lead to trade-offs between ecological situations, because the behavioral composition they facilitate may be suitable in one situation but not another. We use computer simulations to examine various mechanisms that allow keystone individuals to exert their influence on group members. We further discuss a trade-off between two potentially conflicting collective outcomes, cooperative prey attack and disease dynamics. Our simulations match empirical data from a social spider system and produce testable predictions for the causes and consequences of the influence of keystone individuals on group composition and collective outcomes. We find that a group's behavioral composition can be impacted by the keystone individual through changes to interaction patterns or behavioral persistence over time. Group behavioral composition and the mechanisms that drive the distribution of phenotypes influence collective outcomes and lead to trade-offs between disease dynamics and cooperative prey attack.

Population averaging due to paracrine communication can arbitrarily reduce cellular response variability. Yet, variability is ubiquitously observed, suggesting limits to paracrine averaging. It remains unclear whether and how biological systems may be affected by such limits of paracrine signaling. To address this question, we quantify the signal and noise of Ca(2+) and ERK spatial gradients in response to an in vitro wound within a novel microfluidics-based device. We find that while paracrine communication reduces gradient noise, it also reduces the gradient magnitude. Accordingly we predict the existence of a maximum gradient signal to noise ratio. Direct in vitro measurement of paracrine communication verifies these predictions and reveals that cells utilize optimal levels of paracrine signaling to maximize the accuracy of gradient-based positional information. Our results demonstrate the limits of population averaging and show the inherent tradeoff in utilizing paracrine communication to regulate cellular response fidelity.

The innate immune response is largely initiated by pathogen-responsive activation of the transcription factor IRF3. Among other target genes, IRF3 controls the expression of IFN-β, which triggers the activation of the transcription factor ISGF3 via the IFNAR. IRF3 and ISGF3 have been reported to control many of the same target genes and together, control the antimicrobial innate-immune program; however, their respective contributions and specificities remain unclear. Here, we used genomic technologies to characterize their specificity in terms of their physical DNA-binding and genetic function. With the use of ChiP-seq and transcriptomic measurements in WT versus ifnar(-/-) versus ifnar(-/-)irf3(-/-) macrophages responding to intracellular dsRNA, we confirmed the known ISGF3 DNA-binding motif and further specified a distinct IRF3 consensus sequence. The functional specificity of IRF3 is particularly pronounced in cytokine/chemokine regulation; yet, even in the control of IFN-β, that specificity is not absolute. By mathematically modeling IFN-β production within an abstracted tissue layer, we find that IRF3 versus ISGF3 specificity may be critical to limiting IFN-β production and ISGF3 activation, temporally and spatially, but that partial overlap in their specificity is tolerable and may enhance the effectiveness of the innate-immune response.

Stochasticity inherent to biochemical reactions (intrinsic noise) and variability in cellular states (extrinsic noise) degrade information transmitted through signaling networks. We analyzed the ability of temporal signal modulation--that is, dynamics--to reduce noise-induced information loss. In the extracellular signal-regulated kinase (ERK), calcium (Ca(2+)), and nuclear factor kappa-B (NF-κB) pathways, response dynamics resulted in significantly greater information transmission capacities compared to nondynamic responses. Theoretical analysis demonstrated that signaling dynamics has a key role in overcoming extrinsic noise. Experimental measurements of information transmission in the ERK network under varying signal-to-noise levels confirmed our predictions and showed that signaling dynamics mitigate, and can potentially eliminate, extrinsic noise-induced information loss. By curbing the information-degrading effects of cell-to-cell variability, dynamic responses substantially increase the accuracy of biochemical signaling networks.

Stochasticity inherent to biochemical reactions (intrinsic noise) and variability in cellular states (extrinsic noise) degrade information transmitted through signaling networks. We analyzed the ability of temporal signal modulation--that is, dynamics--to reduce noise-induced information loss. In the extracellular signal-regulated kinase (ERK), calcium (Ca(2+)), and nuclear factor kappa-B (NF-κB) pathways, response dynamics resulted in significantly greater information transmission capacities compared to nondynamic responses. Theoretical analysis demonstrated that signaling dynamics has a key role in overcoming extrinsic noise. Experimental measurements of information transmission in the ERK network under varying signal-to-noise levels confirmed our predictions and showed that signaling dynamics mitigate, and can potentially eliminate, extrinsic noise-induced information loss. By curbing the information-degrading effects of cell-to-cell variability, dynamic responses substantially increase the accuracy of biochemical signaling networks.

The social amoeba Dictyostelium discoideum decodes an oscillatory signal to measure time.

The social amoeba Dictyostelium discoideum decodes an oscillatory signal to measure time.

We tested the classical hypothesis that astral, prometaphase bipolar mitotic spindles are maintained by balanced outward and inward forces exerted on spindle poles by kinesin-5 and -14 using modeling of in vitro and in vivo data from Drosophila melanogaster embryos. Throughout prometaphase, puncta of both motors aligned on interpolar microtubules (MTs [ipMTs]), and motor perturbation changed spindle length, as predicted. Competitive motility of purified kinesin-5 and -14 was well described by a stochastic, opposing power stroke model incorporating motor kinetics and load-dependent detachment. Motor parameters from this model were applied to a new stochastic force-balance model for prometaphase spindles, providing a good fit to data from embryos. Maintenance of virtual spindles required dynamic ipMTs and a narrow range of kinesin-5 to kinesin-14 ratios matching that found in embryos. Functional perturbation and modeling suggest that this range can be extended significantly by a disassembling lamin-B envelope that surrounds the prometaphase spindle and augments the finely tuned, antagonistic kinesin force balance to maintain robust prometaphase spindles as MTs assemble and chromosomes are pushed to the equator.

The mitotic spindle is a complex macromolecular machine that coordinates accurate chromosome segregation. The spindle accomplishes its function using forces generated by microtubules (MTs) and multiple molecular motors, but how these forces are integrated remains unclear, since the temporal activation profiles and the mechanical characteristics of the relevant motors are largely unknown. Here, we developed a computational search algorithm that uses experimental measurements to 'reverse engineer' molecular mechanical machines. Our algorithm uses measurements of length time series for wild-type and experimentally perturbed spindles to identify mechanistic models for coordination of the mitotic force generators in Drosophila embryo spindles. The search eliminated thousands of possible models and identified six distinct strategies for MT-motor integration that agree with available data. Many features of these six predicted strategies are conserved, including a persistent kinesin-5-driven sliding filament mechanism combined with the anaphase B-specific inhibition of a kinesin-13 MT depolymerase on spindle poles. Such conserved features allow predictions of force-velocity characteristics and activation-deactivation profiles of key mitotic motors. Identified differences among the six predicted strategies regarding the mechanisms of prometaphase and anaphase spindle elongation suggest future experiments.