A Multidatabase ExTRaction PipEline (METRE) for facile cross validation in critical care research.

Transforming raw EHR data into machine learning model-ready inputs requires considerable effort. One widely used EHR database is Medical Information Mart for Intensive Care (MIMIC). Prior work on MIMIC-III cannot query the updated and improved MIMIC-IV version. Besides, the need to use multicenter datasets further highlights the challenge of EHR data extraction. Therefore, we developed an extraction pipeline that works on both MIMIC-IV and eICU Collaborative Research Database and allows for model cross validation using these 2 databases. Under the default choices, the pipeline extracted 38,766 and 126,448 ICU records for MIMIC-IV and eICU, respectively. Using the extracted time-dependent variables, we compared the Area Under the Curve (AUC) performance with ​​prior works on clinically relevant tasks such as in-hospital mortality prediction. METRE achieved comparable performance with AUC 0.723-0.888 across all tasks with MIMIC-IV. Additionally, when we evaluated the model directly on MIMIC-IV data using a model trained on eICU, we observed that the AUC change can be as small as +0.019 or -0.015. Our open-source pipeline transforms MIMIC-IV and eICU into structured data frames and allows researchers to perform model training and testing using data collected from different institutions, which is of critical importance for model deployment under clinical contexts. The code used to extract the data and perform training is available here: https://github.com/weiliao97/METRE.

Microfluidics in structured multimaterial fibers.

Traditional fabrication techniques for microfluidic devices utilize a planar chip format that possesses limited control over the geometry of and materials placement around microchannel cross-sections. This imposes restrictions on the design of flow fields and external forces (electric, magnetic, piezoelectric, etc.) that can be imposed onto fluids and particles. Here we report a method of fabricating microfluidic channels with complex cross-sections. A scaled-up version of a microchannel is dimensionally reduced through a thermal drawing process, enabling the fabrication of meters-long microfluidic fibers with nonrectangular cross-sectional shapes, such as crosses, five-pointed stars, and crescents. In addition, by codrawing compatible materials, conductive domains can be integrated at arbitrary locations along channel walls. We validate this technology by studying unexplored regimes in hydrodynamic flow and by designing a high-throughput cell separation device. By enabling these degrees of freedom in microfluidic device design, fiber microfluidics provides a method to create microchannel designs that are inaccessible using planar techniques.

Longitudinal multiparameter assay of lymphocyte interactions from onset by microfluidic cell pairing and culture.

Resolving how the early signaling events initiated by cell-cell interactions are transduced into diverse functional outcomes necessitates correlated measurements at various stages. Typical approaches that rely on bulk cocultures and population-wide correlations, however, only reveal these relationships broadly at the population level, not within each individual cell. Here, we present a microfluidics-based cell-cell interaction assay that enables longitudinal investigation of lymphocyte interactions at the single-cell level through microfluidic cell pairing, on-chip culture, and multiparameter assays, and allows recovery of desired cell pairs by micromanipulation for off-chip culture and analyses. Well-defined initiation of interactions enables probing cellular responses from the very onset, permitting single-cell correlation analyses between early signaling dynamics and later-stage functional outcomes within same cells. We demonstrate the utility of this microfluidic assay with natural killer cells interacting with tumor cells, and our findings suggest a possible role for the strength of early calcium signaling in selective coordination of subsequent cytotoxicity and IFN-gamma production. Collectively, our experiments demonstrate that this new approach is well-suited for resolving the relationships between complex immune responses within each individual cell.

Designable 3D Microshapes Fabricated at the Intersection of Structured Flow and Optical Fields.

3D structures with complex geometric features at the microscale, such as microparticles and microfibers, have promising applications in biomedical engineering, self-assembly, and photonics. Fabrication of complex 3D microshapes at scale poses a unique challenge; high-resolution methods such as two-photon-polymerization have print speeds too low for high-throughput production, while top-down approaches for bulk processing using microfabricated template molds have limited control of microstructure geometries over multiple axes. Here, a method for microshape fabrication is presented that combines a thermally drawn transparent fiber template with a masked UV-photopolymerization approach to enable biaxial control of microshape fabrication. Using this approach, high-resolution production of complex microshapes not producible using alternative methods is demonstrated, such as octahedrons, dreidels, and axially asymmetric fibers, at throughputs as high as 825 structures/minute. Finally, the fiber template is functionalized with conductive electrodes to enable hierarchical subparticle localization using dielectrophoretic forces.

Multiplexed Cell-Based Sensors for Assessing the Impact of Engineered Systems and Methods on Cell Health.

Bioinstrumentation engineers have long been creating platforms to study cell health and disease. It becomes necessary to ensure that such cell-probing tools do not themselves harm cells through complex stressors resulting from their design or operational conditions. Here, we present multiplexed cell-based sensors to simultaneously quantify stress induced by diverse mechanisms such as shear stress, DNA damage, and heat shock. Our sensors do not require additional reagents and can be conveniently quantified by flow cytometry and real-time imaging. Successful adaptation of our sensors by external users enabled systematic assessment of multiple flow sorters, alongside their operational parameters using the same cells and preparation. Our results provide insight into "gentle" and stressful sorting parameters that had not been quantified previously. Overall, this work presents a facile and quantitative approach to investigate multifactorial cell-stress emergent from diverse bioinstrumentation, which can be utilized to discover design and operation conditions ideal for cell health.

Attenuation of extrinsic signaling reveals the importance of matrix remodeling on maintenance of embryonic stem cell self-renewal.

The role of extrinsic factors in maintaining self-renewal of embryonic stem cells (ESCs) has been extensively studied since the cells' isolation, but the necessity for cell-secreted factors in self-renewal has remained undefined to date. Although it is generally accepted that addition of leukemia inhibitory factor (LIF) together with either serum or bone morphogenetic protein 4 (BMP4) is sufficient to maintain mouse ESCs (mESCs) in a self-renewing state, this does not preclude the possibility that autocrine factors are also required. Here we make use of a microfluidic perfusion device that is able to globally diminish diffusible autocrine signaling by applying continuous media flow to deplete cell-secreted factors. We demonstrate mESC culture for several days under continuous microfluidic perfusion and show that cell-secreted factors are removed and can be recovered downstream. We find that perturbing cell-secreted signaling causes mESCs to exit their stable self-renewing state in defined conditions that normally support self-renewal and to exhibit properties characteristic of epiblast cells. This state change is not due to the presence of the known autocrine differentiation inducer fibroblast growth factor 4, but, remarkably, it can be prevented by global remodeling of the extracellular matrix (ECM). We also find that cell-secreted matrix remodeling proteins are removed under perfusion and that inhibition of extracellular matrix remodeling causes mESCs to differentiate. Taken together, our data indicate that LIF and BMP4 are not sufficient to maintain self-renewal and that cell-secreted factors are necessary to continuously remodel the ECM and thereby prevent differentiation, revealing a previously undescribed level of mESC regulation through the use of microfluidic perfusion technology.

Image-based single-cell sorting via dual-photopolymerized microwell arrays.

We present a simple and cost-effective single-cell sorting method using two sequential photopolymerization steps that enables sorting based upon imaged phenotypes. The first photopolymerization step uses a thiolene-based resin with minimal autofluorescence to create an array of microwells to capture cells. The second photopolymerization uses (poly)ethylene glycol diacrylate to encapsulate undesired cells in a hydrogel, allowing for retrieval of the desired cell population using simple washing. We quantitatively characterize the method using fluorescently labeled cells and then applied the method to isolate cells based upon imaged fluorescence localization. The method is readily transferrable to other laboratories and will provide a facile route to sorting of cells based on imaged phenotypes.

Cell-based biosensor to report DNA damage in micro- and nanosystems.

Understanding how newly engineered micro- and nanoscale materials and systems that interact with cells impact cell physiology is crucial for the development and ultimate adoption of such technologies. Reports regarding the genotoxic impact of forces applied to cells in such systems that can both directly or indirectly damage DNA emphasize the need for developing facile methods to assess how materials and technologies affect cell physiology. To address this need we have developed a TurboRFP-based DNA damage reporter cell line in NIH-3T3 cells that fluoresce to report genotoxic stress caused by a wide variety of agents, from chemical genotoxic agents to UV-C radiation. Our biosensor was successfully implemented in reporting the genotoxic impact of nanomaterials, demonstrating the ability to assess size dependent geno- and cyto-toxicity. The biosensor cells can be assayed in a high throughput, noninvasive manner, with no need for overly sophisticated equipment or additional reagents. We believe that this open-source biosensor is an important resource for the community of micro- and nanomaterials and systems designers and users who wish to evaluate the impact of systems and materials on cell physiology.

Matrix remodeling maintains embryonic stem cell self-renewal by activating Stat3.

While a variety of natural and synthetic matrices have been used to influence embryonic stem cell (ESC) self-renewal or differentiation, and ESCs also deposit a rich matrix of their own, the mechanisms behind how extracellular matrix affects cell fate are largely unexplored. The ESC matrix is continuously remodeled by matrix metalloproteinases (MMPs), a process that we find is enhanced by the presence of mouse embryonic fibroblast feeders in a paracrine manner. Matrix remodeling by MMPs aids in the self-renewal of ESCs, as inhibition of MMPs inhibits the ability of ESCs to self-renew. We also find that addition of the interstitial collagenase MMP1 is sufficient to maintain long-term leukemia inhibitory factor (LIF)-independent mouse ESC (mESC) self-renewal in a dose-dependent manner. This remarkable ability is due to the presence of endogenously produced self-renewal-inducing signals, including the LIF-family ligand ciliary neurotrophic factor, that are normally trapped within the ECM and become exposed upon MMP-induced matrix remodeling to signal through JAK and Stat3. These results uncover a new role for feeder cells in maintaining self-renewal and show that mESCs normally produce sufficient levels of autocrine-acting pro-self-renewal ligands.

Rapid, low-cost fabrication of electronic microfluidics via inkjet-printing and xurography (MINX).

Microfluidics has shown great promise for point-of-care assays due to unique chemical and physical advantages that occur at the micron scale. Furthermore, integration of electrodes into microfluidic systems provides additional capabilities for assay operation and electronic readout. However, while these systems are abundant in biological and biomedical research settings, translation of microfluidic devices with embedded electrodes are limited. In part, this is due to the reliance on expensive, inaccessible, and laborious microfabrication techniques. Although innovative prior work has simplified microfluidic fabrication or inexpensively patterned electrodes, low-cost, accessible, and robust methods to incorporate all these elements are lacking. Here, we present MINX, a low-cost <1 USD and rapid (∼minutes) fabrication technique to manufacture microfluidic device with embedded electrodes. We characterize the structures created using MINX, and then demonstrate the utility of the approach by using MINX to implement an electrochemical bead-based biomarker detection assay. We show that the MINX technique enables the scalable, inexpensive fabrication of microfluidic devices with electronic sensors using widely accessible desktop machines and low-cost materials.

Accelerating the optimization of vertical flow assay performance guided by a rational systematic model-based approach.

Rapid diagnostic tests (RDTs) have shown to be instrumental in healthcare and disease control. However, they have been plagued by many inefficiencies in the laborious empirical development and optimization process for the attainment of clinically relevant sensitivity. While various studies have sought to model paper-based RDTs, most have relied on continuum-based models that are not necessarily applicable to all operation regimes, and have solely focused on predicting the specific interactions between the antigen and binders. It is also unclear how the model predictions may be utilized for optimizing assay performance. Here, we propose a streamlined and simplified model-based framework, only relying on calibration with a minimal experimental dataset, for the acceleration of assay optimization. We show that our models are capable of recapitulating experimental data across different formats and antigen-binder-matrix combinations. By predicting signals due to both specific and background interactions, our facile approach enables the estimation of several pertinent assay performance metrics such as limit-of-detection, sensitivity, signal-to-noise ratio and difference. We believe that our proposed workflow would be a valuable addition to the toolset of any assay developer, regardless of the amount of resources they have in their arsenal, and aid assay optimization at any stage in their assay development process.

Microfluidic control of cell pairing and fusion.

Cell fusion has been used for many different purposes, including generation of hybridomas and reprogramming of somatic cells. The fusion step is the key event in initiation of these procedures. Standard fusion techniques, however, provide poor and random cell contact, leading to low yields. We present here a microfluidic device to trap and properly pair thousands of cells. Using this device, we paired different cell types, including fibroblasts, mouse embryonic stem cells and myeloma cells, achieving pairing efficiencies up to 70%. The device is compatible with both chemical and electrical fusion protocols. We observed that electrical fusion was more efficient than chemical fusion, with membrane reorganization efficiencies of up to 89%. We achieved greater than 50% properly paired and fused cells over the entire device, fivefold greater than with a commercial electrofusion chamber and observed reprogramming in hybrids between mouse embryonic stem cells and mouse embryonic fibroblasts.

Fluid shear stress primes mouse embryonic stem cells for differentiation in a self-renewing environment via heparan sulfate proteoglycans transduction.

Shear stress is a ubiquitous environmental cue experienced by stem cells when they are being differentiated or expanded in perfusion cultures. However, its role in modulating self-renewing stem cell phenotypes is unclear, since shear is usually only studied in the context of cardiovascular differentiation. We used a multiplex microfluidic array, which overcomes the limitations of macroperfusion systems in shear application throughput and precision, to initiate a comprehensive, quantitative study of shear effects on self-renewing mouse embryonic stem cells (mESCs), where shear stresses varying by >1000 times (0.016-16 dyn/cm(2)) are applied simultaneously. When compared with static controls in the presence or absence of a saturated soluble environment (i.e., mESC-conditioned medium), we ascertained that flow-induced shear stress specifically up-regulates the epiblast marker Fgf5. Epiblast-state transition in mESCs involves heparan sulfate proteoglycans (HSPGs), which have also been shown to transduce shear stress in endothelial cells. By disrupting (with sulfation inhibitors and heparinase) and partially reconstituting (with heparin) HSPG function, we show that mESCs also mechanically sense shear stress via HSPGs to modulate Fgf5 expression. This study demonstrates that self-renewing mESCs possess the molecular machinery to sense shear stress and provides quantitative shear application benchmarks for future scalable stem cell culture systems.

A sample-to-answer electrochemical biosensor system for biomarker detection.

Biomarker detection is critical for the diagnosis and treatment of numerous diseases. Typically, target biomarkers in blood samples are measured through tests conducted at centralized laboratories. Testing at central laboratories increases wait times for results, in turn increasing healthcare costs and negatively impacting patient outcomes. Alternatively, point-of-care platforms enable the rapid measurement of biomarkers, expand testing location capabilities and mitigate manual processing steps through integration and automation. However, many of these systems focus on sample detection rather than the equally important sample preparation. Here we present a fully integrated and automated sample-to-answer electrochemical biosensing platform which incorporates each aspect of the biomarker testing workflow from blood collection to sample preparation to assay operation and readout. The system combines a commercial microneedle blood sampling device with membrane-based plasma filtration upstream of a bead-based electrochemical immunoassay. We characterize the high separation efficiency (>99%) and low non-specific binding of the whole blood-to-plasma filtration membrane under a range of operating conditions. We demonstrate a full sample-to-answer workflow through the analysis of interlukin-6-spiked blood samples.

Fully Automated, Sample-to-Answer Leukocyte Functional Assessment Platform for Continuous Sepsis Monitoring via Microliters of Blood.

We report a fully automated, sample-to-answer, and label-free leukocyte activation analysis platform for monitoring immune responses in sepsis, by integrating the multidimensional double spiral (MDDS) and isodielectric separation (IDS) subplatforms. The integrated platform can provide rapid and fully automated identification of clinically diagnosed sepsis patients from only 50 µL of peripheral blood volume within 25 min. Many critical innovations were implemented in direct interconnection between the two subplatforms, such as intermediate sample storage and sample transfer, addressing flow rate mismatch (from mL/min to µL/min), and integration of a ridge array for upstream cell focusing in the IDS subplatform. The ridge array in the IDS subplatform can prevent the distortion of electrical profiling due to the residual red blood cells even after the MDDS process. We showed that the integrated platform can separate leukocytes (up to >99.9% red blood cell removal) in the MDDS subplatform and automatically transfer them to the downstream ridge-integrated IDS subplatform for their activation analysis without any apparent ex vivo cell activation and any human intervention. We also demonstrated that the integrated platform can identify differences between leukocytes from human sepsis and healthy subjects significantly (p = 0.0024, 95% confidence interval) by looking into differences in the intrinsic electrical properties of leukocytes. The integrated platform could enable monitoring of host leukocyte function daily or even hourly as a bedside assessment tool, which is currently a critical yet unmet need for managing many critical care patients.

Inflammation resolution circuits are uncoupled in acute sepsis and correlate with clinical severity.

Sepsis is a critical illness characterized by dysregulated inflammatory responses lacking counter-regulation. Specialized proresolving mediators are agonists for antiinflammation and for promoting resolution, and they are protective in preclinical sepsis models. Here, in human sepsis, we mapped resolution circuits for the specialized proresolving mediators resolvin D1 and resolvin D2 in peripheral blood neutrophils and monocytes, their regulation of leukocyte activation and function ex vivo, and their relationships to measures of clinical severity. Neutrophils and monocytes were isolated from healthy subjects and patients with sepsis by inertial microfluidics and resolvin D1 and resolvin D2 receptor expression determined by flow cytometry. The impact of these resolvins on leukocyte activation was determined by isodielectric separation and leukocyte function by stimulated phagolysosome formation. Leukocyte proresolving receptor expression was significantly higher in sepsis. In nanomolar concentrations, resolvin D1 and resolvin D2 partially reversed sepsis-induced changes in leukocyte activation and function. Principal component analyses of leukocyte resolvin receptor expression and responses differentiated sepsis from health and were associated with measures of sepsis severity. These findings indicate that resolvin D1 and resolvin D2 signaling for antiinflammation and resolution are uncoupled from leukocyte activation in early sepsis and suggest that indicators of diminished resolution signaling correlate with clinical disease severity.

Surface-patterned electrode bioreactor for electrical stimulation.

We present a microscale cell culture system with an interdigitated microarray of excimer-laser-ablated indium tin oxide electrodes for electrical stimulation of cultured cells. The system has been characterized in a range of geometeries and stimulation regimes via electrochemical impedance spectroscopy and used to culture primary cardiomyocytes and human adipose derived stem cells. Over 6 days of culture with electrical stimulation (2 ms duration, 1 Hz, 180 microm wide electrodes with 200 microm spacing), both cell types exhibited enhanced proliferation, elongation and alignment, and adipose derived stem cells exhibited higher numbers of Connexin-43-composed gap junctions.

An integrated model for bead-based immunoassays.

Bead-based immunoassays have shown great promise for rapid and sensitive protein quantification. However, there still lacks holistic understanding of assay performance that can inform assay design and optimization. In this paper, we present an integrated mathematical model for surface coverage bead-based assays. This model examines the building blocks of surface coverage assays, including heterogeneous binding of analyte molecules on bead or sensor surfaces, attachment of bead labels to sensor surfaces, and generation of electrochemical current by bead labels. To demonstrate and validate this model, we analyze a semi-homogeneous bead-based electronic enzyme-linked immunosorbent assay and find that experimental results agree with various model predictions. We show that the model can provide design guidance for choice of various assay parameters including bead size, bead number, antibody affinity and assay time, and provide a perspective to reconcile the performance of various implementations of surface coverage assays.

Plastic masters-rigid templates for soft lithography.

We demonstrate a simple process for the fabrication of rigid plastic master molds for soft lithography directly from (poly)dimethysiloxane devices. Plastics masters (PMs) provide a cost-effective alternative to silicon-based masters and can be easily replicated without the need for cleanroom facilities. We have successfully demonstrated the use of plastics micromolding to generate both single and dual-layer plastic structures, and have characterized the fidelity of the molding process. Using the PM fabrication technique, world-to-chip connections can be integrated directly into the master enabling devices with robust, well-aligned fluidic ports directly after molding. PMs provide an easy technique for the fabrication of microfluidic devices and a simple route for the scaling-up of fabrication of robust masters for soft lithography.

Separation of particles by pulsed dielectrophoresis.

In this paper, we introduce a dielectrophoresis (DEP)-based separation method that allows for tunable multiplex separation of particles. In traditional DEP separations where the field is applied continuously, size-based separation of particles uses the cubic dependence of the DEP force on particle radius, causing large particles to be retained while small particles are released. Here we show that by pulsing the DEP force in time, we are able to reverse the order of separation (eluting the large particles while retaining the small ones), and even extract mid-size particles from a heterogeneous population in one step. The operation is reminiscent of prior dielectrophoretic ratchets which made use of DEP and Brownian motion, but we have applied the asymmetric forces in time rather than in a spatial arrangement of electrodes, thus simplifying the system. We present an analytical model to study the dynamic behavior of particles under pulsed DEP and to understand the different modes of separation. Results from the model and the experimental observations are shown to be in agreement.