Single-Cell Phenotypic Analysis and Digital Molecular Detection Linkable by a Hydrogel Bead-Based Platform

Cell heterogeneity, such as antibiotic heteroresistance and cancer cell heterogeneity, has been increasingly observed. To probe the underlying molecular mechanisms in the dynamically changing heterogeneous cells, a high throughput platform is urgently needed to establish single cell genotype-phenotype correlations. Herein, we report a platform combining single-cell viability phenotypic analysis with digital molecular detection for bacterial cells. The platform utilizes polyethylene glycol hydrogel that cross-links through a thiol-Michael addition, which is biocompatible, fast, and spontaneous. To generate uniform nanoliter-sized hydrogel beads (Gelbeads), we developed a convenient and disposable device made of needles and microcentrifuge tubes. Gelbead-based single cell viability and molecular detection assays were established. Enhanced thermal stability and uncompromised efficiency were achieved for digital polymerase chain reaction (PCR) and digital loop-mediated isothermal amplification (LAMP) within the Gelbeads. Reagent exchange for in situ PCR following viability phenotypic analyses was demonstrated. The combined analyses may address the genotypic differences between cellular subpopulations exhibiting distinct phenotypes. The platform promises unique perspectives in mechanism elucidation of environment-evolution interaction that may be extended to other cell types for medical research.


S-2
Supplementary Tables   Table S1. LAMP mix PCR mix TSB media  Table S1. PEG hydrogel crosslinking characterization in bulk for LAMP mix, PCR mix, and TSB media. Sol-Gel transition time was experimentally determined (See Online methods). The pH values were supplied by the manufactures. The pore sizes were theoretically estimated for our gel concentration by scaling from experimentally measured mesh sizes, assuming a simplified hydrogel architecture. 1 S-3 Table S2. Table S2. Sequences of primers and probe for PCR and LAMP assays. PCR primers and probes target a region in gene STY0201 specific for S. Typhi for an amplicon size of 131 bp. 2 LAMP primers target a 196 bp region within the S. Typhi specific gene STY1607. 3 The target regions for PCR and LAMP were both found to occur at one copy per cell, by searching the sequences within the complete genome of S. Typhi strain CT18 (Accession no. NC_003198) using Basic Local Alignment Search Tool (BLAST) 4 . S-14

Supplementary Note 1: Characterization of PEG hydrogel crosslinking
To identify the reasonable time frame for Gelbead generation, PEG gelation time in our targeted reaction matrices by bulk phase sol-gel transition experiments (online methods) was first measured.
Ideally the compartmentalization process should be completed before the sol-gel transition starting, after which further crosslinking would considerably alter fluid properties such as viscosity and surface tension. For the three types of reaction matrix examined, results showed that the sol-gel transition start time spanned from 4.5 min to 43.0 min, and the crosslinking was accelerated by higher pH and higher monomer concentration (Table S1). Accordingly, the gelation time might be further extended by decreasing the crosslinking temperature 8 . We then estimated if the lower gel concentration considerably affects the hydrogel properties. The theoretical pore sizes for our crosslinked PEG were close for 7.5% and 10% gel (Table S1). Both would allow diffusion of functional molecules in our applications including water, ions, small DNA fragments, and proteins that size from below an angstrom to ~ 6 nm. 6 This estimation neglects non-ideality such as dangling ends or monomer self-interlinking, which might result in larger actual pore size. 1 Using the same PEG monomers, 10 w/v% hydrogel concentration has been utilized in cell encapsulation, multiple displacement amplification (MDA) and LAMP. [9][10] We reason that lowering the hydrogel concentration to 7.5 w/v% would benefit our applications by allowing more time for Gelbead generation and creating looser hydrogel network for reagent diffusion. Therefore, 7.5 w/v% PEG was used in further experiments.

Supplementary Note 2: Setup of the simulation model for centrifugal droplet generation
We performed finite element modeling using COMSOL Multiphysics 11 to investigate the droplet generation process at the needle tip. The fluid regime in the microcentrifuge tube outside the needle was modeled as a two-dimensional geometry, as illustrated in Figure S1a.  (Figure S1b) was used to measure the droplet size at each condition. Three measurement of droplet diameter was taken for the droplet that just pinched off, and the standard deviation was calculated. S-17

Supplementary Note 3: Droplet generation performance and sources of error
To achieve high throughput analysis of droplets with limited available instruments, we chose to analyze droplets through fluorescence imaging. As in the reported protocol, droplets were extracted into a viewing chamber, made by bonding a commercial plastic chamber onto a glass slide, for fluorescence imaging at 1.25× objective. We acknowledge that this protocol might have introduced systematic error in size characterizations by 1) pipetting droplets from the microcentrifuge tube into the viewing chamber, 2) noise difference from focus point to the edges within an image, and 3) image processing bias by MATLAB when identifying the circular-shaped edges. These sources of error would lead to overestimation of size distribution. Therefore, it is anticipated that the actual CV of the generated droplets or Gelbeads should be lower than reported.
Generally, the reported sizes and CVs of generated droplets, optimized at 175 µm diameter with CV of 5%, are comparable to those generated by centrifugal microfluidics reported in literature.

Supplementary Note 4: Effect of varying physical parameters on droplet generation
Through COMSOL simulation of the two-phase flow at the needle tip, the effect of the needle bent angle, aqueous phase viscosity, and interfacial tension on the size of the droplets generated is shown in Figure S1c-e, respectively. The needle bent angle, which is the angle between the stem and the bent tip, is shown to have limited impact on the droplet size ( Figure S1c). When this angle is varied between ±10°, the droplet sizes were all around 126 μm within the standard deviations of each other. This theoretically corroborates that, despite the seemingly not rigorous manual fabrication, the trial-to-trial error of NeaT droplet generation may still be fairly small. Aqueous phase viscosity itself has negligible impact on droplet size interestingly, although it was observed in other droplet generation studies that the higher viscosity led to larger generated droplets. 15 We speculate that the viscosity effect may manifest by changing the fluid velocity, and therefore optimization is needed for a new system with different viscosity of phases. It is also observed that the droplet size increases with increasing interfacial tension linearly (Figure S1e, R 2 = 0.989). This relationship indicates that, if smaller droplet size is desired, it can be achieved by adjusting the type and concentration of the surfactant for a lower interfacial tension 12 . This result also suggests that optimization is needed for a new combination of aqueous and oil phases when extending to other applications. We note that the absolute value of the simulated droplet sizes might not be directly comparable to those measured experimentally, as the simulation was set up with a simplified 2D approximation of the system. For the experimental condition that involves biological reagents in the aqueous phase, the actual interfacial properties may deviate from the ideal values used. In summary, the results from the simulations indicate that the developed NeaT droplet generation method is theoretically unsusceptible to fluctuations caused by manual fabrication, and this method can be easily extended to other applications that requires water-in-oil droplets. S-19

Supplementary Note 5: Numerical study of cell distribution under centrifugation
Monte Carlo experiments were carried out to investigate the patterns of particle compartmentalization before and after centrifugation. The fluid regime inside the Luer lock was represented by a 2D rectangular (5 mm × 4 mm) model, as shown in Figure S2a.

Supplementary Note 7: Overcoming challenges of PCR reagent infusion for Gelbeads
Initial attempts on PCR reagent infusion for phenotyped Gelbeads failed to yield any bright Gelbead, despite the visually successful Gelbead phase transfer from oil to aqueous phase and then back to the oil phase. Hydrogel network has been reported to possibly form smaller pores on the side of higher interfacial tension 17 . It was thus suspected that, during Gelbead crosslinking in fluorinated oil, the aqueous-oil interface might have smaller pore sizes. The surface barrier then might have hindered the inward diffusion of essential PCR macromolecules when the Gelbeads were transferred into aqueous phase. To overcome the Gelbead surface barrier, we performed a freeze-thaw treatment on the phenotyped Gelbeads prior to reagent infusion. The overnight freezing of the Gelbeads in water was intended for expanding the pores through ice crystal formation, which has been utilized to fabricate macroporous hydrogel for biomedical applications 18 . After adding the freeze-thaw treatment while other protocols remained unchanged, the Gelbead in situ PCR showed successful amplification.
It was observed that the contrast between positive and negative Gelbeads was less visually apparent than in direct gdPCR (Figure 5d-h), likely due to enlarged pore size. In our experiments, the fluorescence intensity difference was distinguishable for positive and negative Gelbeads, but the larger pore size might cause potential problems. For example, the viability phenotyping assay with other fluorescence dyes, such as PrestoBlue, that are more soluble in fluorinated oil thus easier to cross the interfacial barrier might have more crosstalking. It was also observed that, after phase transfers even without freeze-thaw treatment, the diameter of the Gelbead increased approximately from 160 µm to 220 µm. Gelbead swelling during the phase transfer process should not affect the interpretation of downstream PCR results, since the positive or negative reading on a Gelbead only reveals the molecular information of its own encapsulated cells since phenotyping.

S-23
However, the actual volume of the swollen Gelbeads need to be cautiously estimated while preparing the concentrated PCR reagent mixture, so that the final component concentrations can be achieved accurately.