Advertisement
Research Article|Articles in Press

Sorting single-cell microcarriers using commercial flow cytometers

  • Author Footnotes
    1 These authors contributed equally to this work.
    Joseph de Rutte
    Correspondence
    Corresponding author.
    Footnotes
    1 These authors contributed equally to this work.
    Affiliations
    Department of Bioengineering, University of California, Los Angeles, USA

    Partillion Bioscience Corporation, Los Angeles, CA, USA
    Search for articles by this author
  • Author Footnotes
    1 These authors contributed equally to this work.
    Robert Dimatteo
    Footnotes
    1 These authors contributed equally to this work.
    Affiliations
    Department of Chemical and Biomolecular Engineering, University of California, Los Angeles, USA
    Search for articles by this author
  • Sheldon Zhu
    Affiliations
    Partillion Bioscience Corporation, Los Angeles, CA, USA
    Search for articles by this author
  • Maani M Archang
    Affiliations
    Department of Bioengineering, University of California, Los Angeles, USA
    Search for articles by this author
  • Dino Di Carlo
    Affiliations
    Department of Bioengineering, University of California, Los Angeles, USA

    Department of Mechanical and Aerospace Engineering, University of California, Los Angeles, USA

    California NanoSystems Institute, University of California, Los Angeles, USA

    Jonsson Comprehensive Cancer Center, University of California, Los Angeles, USA
    Search for articles by this author
  • Author Footnotes
    1 These authors contributed equally to this work.
Open AccessPublished:October 13, 2021DOI:https://doi.org/10.1016/j.slast.2021.10.008

      Abstract

      The scale of biological discovery is driven by the vessels in which we can perform assays and analyze results, from multi-well plates to microfluidic compartments. We report on the compatibility of sub-nanoliter single-cell containers or “nanovials” with commercial fluorescence activated cell sorters (FACS). This recent lab on a particle approach utilizes 3D structured microparticles to isolate cells and perform single-cell assays at scale with existing lab equipment. Use of flow cytometry led to detection of fluorescently labeled protein with dynamic ranges spanning 2–3 log and detection limits down to ∼10,000 molecules per nanovial, which was the lowest amount tested. Detection limits were improved compared to fluorescence microscopy measurements using a 20X objective and a cooled CMOS camera. Nanovials with diameters between 35–85 µm could also be sorted with purity from 99–93% on different commercial instruments at throughputs up to 800 events/second. Cell-loaded nanovials were found to have unique forward and side (or back) scatter signatures that enabled gating of cell-containing nanovials using scatter metrics alone. The compatibility of nanovials with widely-available commercial FACS instruments promises to democratize single-cell assays used in discovery of antibodies and cell therapies, by enabling analysis of single cells based on secreted products and leveraging the unmatched analytical capabilities of flow cytometers to sort important clones.

      Keywords

      Introduction

      The ability to precisely manipulate and partition individual cells within miniaturized fluid volumes has expanded biological discovery to encompass the heterogeneity across cell populations [
      • Di Carlo D.
      • Lee L.P.
      Dynamic single-cell analysis for quantitative biology.
      ,
      • Di Carlo D.
      • Tse H.T.K.
      • Gossett D.R.
      Introduction: why analyze single cells?.
      ,
      • Chen Z.
      • Chen J.J.
      • Fan R.
      Single-cell protein secretion detection and profiling.
      ]. Traditional workflows focused on measuring bulk properties of interest from populations of cells have given way to novel microfluidic technologies enabling the parallelized assessment of these same features from each cell in a population simultaneously. Cursory insights gleaned from population averages are now being refined with a tremendous amount of single cell multi-omics data, fostering nuanced understandings of phenotypic heterogeneity and population dynamics [
      • Lu Y.
      • Xue Q.
      • Eisele M.R.
      • Sulistijo E.S.
      • Brower K.
      • Han L.
      • Amir E.A.D.
      • Pe'er D.
      • Miller-Jensen K.
      • Fan R.
      Highly multiplexed profiling of single-cell effector functions reveals deep functional heterogeneity in response to pathogenic ligands.
      ]. Early-stage adoption of these novel single cell functional screens have already proved critical across all stages of modern drug development, from antibody discovery [
      • Jin A.
      • Ozawa T.
      • Tajiri K.
      • Obata T.
      • Kondo S.
      • Kinoshita K.
      • Kadowaki S.
      • Takahashi K.
      • Sugiyama T.
      • Kishi H.
      • Muraguchi A.
      A rapid and efficient single-cell manipulation method for screening antigen-specific antibody-secreting cells from human peripheral blood.
      ,
      • Jones B.E.
      • Brown-Augsburger P.L.
      • Corbett K.S.
      • Westendorf K.
      • Davies J.
      • Cujec T.P.
      • Wiethoff C.M.
      • Blackbourne J.L.
      • Heinz B.A.
      • Foster D.
      • Higgs R.E.
      • Balasubramaniam D.
      • Wang L.
      • Zhang Y.
      • Yang E.S.
      • Bidshahri R.
      • Kraft L.
      • Hwang Y.
      • Žentelis S.
      • Jepson K.R.
      • Goya R.
      • Smith M.A.
      • Collins D.W.
      • Hinshaw S.J.
      • Tycho S.A.
      • Pellacani D.
      • Xiang P.
      • Muthuraman K.
      • Sobhanifar S.
      • Piper M.H.
      • Triana F.J.
      • Hendle J.
      • Pustilnik A.
      • Adams A.C.
      • Berens S.J.
      • Baric R.S.
      • Martinez D.R.
      • Cross R.W.
      • Geisbert T.W.
      • Borisevich V.
      • Abiona O.
      • Belli H.M.
      • de Vries M.
      • Mohamed A.
      • Dittmann M.
      • Samanovic M.I.
      • Mulligan M.J.
      • Goldsmith J.A.
      • Hsieh C.L.
      • Johnson N.V.
      • Wrapp D.
      • McLellan J.S.
      • Barnhart B.C.
      • Graham B.S.
      • Mascola J.R.
      • Hansen C.L.
      • Falconer E.
      The neutralizing antibody, LY-CoV555, protects against SARS-CoV-2 infection in non-human primates.
      ,
      • Winters A.
      • McFadden K.
      • Bergen J.
      • Landas J.
      • Berry K.A.
      • Gonzalez A.
      • Salimi-Moosavi H.
      • Murawsky C.M.
      • Tagari P.
      • King C.T.
      Rapid single B cell antibody discovery using nanopens and structured light.
      ] to cell line development [
      • Le K.
      • Tan C.
      • Gupta S.
      • Guhan T.
      • Barkhordarian H.
      • Lull J.
      • Stevens J.
      • Munro T.
      A novel mammalian cell line development platform utilizing nanofluidics and optoelectro positioning technology.
      ,
      • Josephides D.
      • Davoli S.
      • Whitley W.
      • Ruis R.
      • Salter R.
      • Gokkaya S.
      • Vallet M.
      • Matthews D.
      • Benazzi G.
      • Shvets E.
      • Gesellchen F.
      • Geere D.
      • Liu X.
      • Li X.
      • Mackworth B.
      • Young W.
      • Owen Z.
      • Smith C.
      • Starkie D.
      • White J.
      • Sweeney B.
      • Hinchliffe M.
      • Tickle S.
      • Lightwood D.J.
      • Rehak M.
      • Craig F.F.
      • Holmes D.
      Cyto-mine: an integrated, picodroplet system for high-throughput single-cell analysis, sorting, dispensing, and monoclonality assurance.
      ]. Unfortunately, even with the technical progress that has been achieved, single-cell screening capabilities are often limited in scale and restricted to researchers who have the capability to implement complex microfluidic tools or have access to a few high-priced commercial platforms.
      Researchers have developed techniques to address some of the challenges related to access and throughput by leveraging common flow cytometers for downstream analysis and sorting, instead of specialized instruments. For example, hybrid techniques using microfluidics to encapsulate cells within hydrogel particles [
      • Chokkalingam V.
      • Tel J.
      • Wimmers F.
      • Liu X.
      • Semenov S.
      • Thiele J.
      • Figdor C.G.
      • Huck W.T.S.
      Probing cellular heterogeneity in cytokine-secreting immune cells using droplet-based microfluidics.
      ,
      • Li M.
      • van Zee M.
      • Riche C.T.
      • Tofig B.
      • Gallaher S.D.
      • Merchant S.S.
      • Damoiseaux R.
      • Goda K.
      • Carlo Di
      • Gelatin D.A
      Microdroplet platform for high-throughput sorting of hyperproducing single-cell-derived microalgal clones.
      ,
      • Yanakieva D.
      • Elter A.
      • Bratsch J.
      • Friedrich K.
      • Becker S.
      • Kolmar H.
      FACS-based functional protein screening via microfluidic co-encapsulation of yeast secretor and mammalian reporter cells.
      ,
      • Fang Y.
      • Chu T.H.
      • Ackerman M.E.
      • Griswold K.E.
      Going native: direct high throughput screening of secreted full-length IgG antibodies against cell membrane proteins.
      ,
      • Eun Y.J.
      • Utada A.S.
      • Copeland M.F.
      • Takeuchi S.
      • Weibel D.B.
      Encapsulating bacteria in agarose microparticles using microfluidics for high-throughput cell analysis and isolation.
      ], double emulsions [
      • Brower K.K.
      • Khariton M.
      • Suzuki P.H.
      • Still C.
      • Kim G.
      • Calhoun S.G.K.
      • Qi L.S.
      • Wang B.
      • Fordyce P.M.
      Double emulsion picoreactors for high-throughput single-cell encapsulation and phenotyping via FACS.
      ,
      • Zinchenko A.
      • Devenish S.R.A.
      • Kintses B.
      • Colin P.Y.
      • Fischlechner M.
      • Hollfelder F.
      One in a million: flow cytometric sorting of single cell-lysate assays in monodisperse picolitre double emulsion droplets for directed evolution.
      ,
      • Terekhov S.S.
      • Smirnov I.V.
      • Stepanova A.V.
      • Bobik T.V.
      • Mokrushina Y.A.
      • Ponomarenko N.A.
      • Belogurov A.A.
      • Rubtsova M.P.
      • Kartseva O.V.
      • Gomzikova M.O.
      • Moskovtsev A.A.
      • Bukatin A.S.
      • Dubina M.V.
      • Kostryukova E.S.
      • Babenko V.V.
      • Vakhitova M.T.
      • Manolov A.I.
      • Malakhova M.V.
      • Kornienko M.A.
      • Tyakht A.V.
      • Vanyushkina A.A.
      • Ilina E.N.
      • Masson P.
      • Gabibov A.G.
      • Altman S.
      Microfluidic droplet platform for ultrahigh-throughput single-cell screening of biodiversity.
      ], or hollow particle shells [
      • Zee M.
      • Rutte J.
      • Rumyan R.
      • Williamson C.
      • Burnes T.
      • Radakovits R.
      • Eugenio A.S.
      • Badih S.
      • Lee D.H.
      • Archang M.
      • Carlo D.Di
      High-throughput selection of microalgae based on biomass accumulation rates in production environments using picoshell particles.
      ] have been developed to create small single-cell containers that can be analyzed and sorted with standard fluorescent activated cell sorters (FACS). Still, widespread adoption of these approaches is limited due to the significant expertise and specialized equipment required for the upstream formation of compartments using microfluidic devices, and limited capability to perform standard laboratory operations such as washing and reagent exchange once compartments have been formed. Further, the serial nature of forming compartments with these approaches limit potential throughput of the systems.
      Similarly, continued miniaturization of standard micro-titer plates, have significantly enhanced our ability to screen individual cells in parallel. However, the advantages gained from enabling direct scale down of common assay protocols utilized in more highly parallelized micro-titer plate formats are offset by the need to use liquid handlers and other automated robotics to accurately interface with the miniaturized well designs, significantly increasing assay costs and limiting throughputs. Additionally, there is a fundamental limit on the volume of these compartments as interfacial tension effects dominate at lower scales making it difficult to accurately perform liquid handling operations. To unlock the potential of single cell screening workflows, new technologies that extend compartmentalization to single-cell compatible volumes must be designed and standardized to work fully with existing infrastructure and common laboratory operations.
      Recently our group has reported on novel “lab on a particle” technologies that enable users to perform the same fundamental operations as in microtiter plates, but at orders of magnitude higher throughputs and at volumes on the scale of individual cells [
      • Wu C.Y.
      • Ouyang M.
      • Wang B.
      • de Rutte J.
      • Joo A.
      • Jacobs M.
      • Ha K.
      • Bertozzi A.L.
      • Di Carlo D.
      Monodisperse drops templated by 3D-structured microparticles.
      ,
      • Destgeer G.
      • Ouyang M.
      • Wu C.Y.
      • Di Carlo D.
      Fabrication of 3D concentric amphiphilic microparticles to form uniform nanoliter reaction volumes for amplified affinity assays.
      ,
      • de Rutte J.
      • Dimatteo R.
      • van Zee M.
      • Damoiseaux R.
      • Di Carlo D.
      Massively parallel encapsulation of single cells with structured microparticles and secretion-based flow sorting.
      ,
      • Lee S.
      • Rutte J.
      • Dimatteo R.
      • Koo D.
      • Carlo D.Di
      Scalable fabrication of 3D structured microparticles using induced phase separation.
      ]. These particle based platforms are unique in that they can be manufactured at scale in centralized locations and distributed to end users to perform assays utilizing existing lab infrastructure. We first applied the platform for screening and sorting individual cells based on their secreted products in high-throughput [
      • de Rutte J.
      • Dimatteo R.
      • van Zee M.
      • Damoiseaux R.
      • Di Carlo D.
      Massively parallel encapsulation of single cells with structured microparticles and secretion-based flow sorting.
      ,
      • Lee S.
      • Rutte J.
      • Dimatteo R.
      • Koo D.
      • Carlo D.Di
      Scalable fabrication of 3D structured microparticles using induced phase separation.
      ]. Our hydrogel nanovial technology acts as suspendable sub-nanoliter containers that can be functionalized with chemical moieties to promote cell binding and growth (Fig. 1). Fluids are easily exchanged around the nanovials by simple pipetting and centrifugation steps enabling exposure to different chemical stimuli or staining reagents. The cavity of the nanovials can be sealed in parallel through emulsification with biocompatible oils preventing molecular cross-talk between compartments [
      • de Rutte J.
      • Dimatteo R.
      • van Zee M.
      • Damoiseaux R.
      • Di Carlo D.
      Massively parallel encapsulation of single cells with structured microparticles and secretion-based flow sorting.
      ,
      • Ha K.
      • de Rutte J.
      • Di Carlo D.
      • Bertozzi A.
      Minimal surface configurations for axisymmetric microparticles.
      ]. Like a well-plate bottom, the hydrogel surfaces of nanovials can act as a substrate for performing biomolecular assays such as sandwich immuno-assays. Importantly, the emulsification process is reversible allowing for viable recovery of cells, staining, and analysis or sorting using FACS.
      Fig 1
      Fig. 1Nanovials: Sub-nanoliter containers for single cell analysis and sorting. (A) Overview of the key features of the nanovial platform. (i) Cavity-containing hydrogel microparticles (nanovials) can be modified with various binding moieties to directly facilitate the adhesion and growth of loaded cells. (ii) Reagents can be easily exchanged around cells contained in the nanovials using simple pipetting and centrifugation steps. (iii) Nanovials can be sealed to create uniform sub-nanoliter compartments by emulsifying with biocompatible oil and surfactants using a pipette or through vortexing. Nanovials can be un-sealed to perform downstream processes using a destabilizing agent to coalesce emulsions. (iv) The surface of nanovials can be used for capturing molecules of interest such as secreted proteins. (v) Nanovials can be sized to be compatible with commercial flow cytometers and FACS machines for high-throughput analysis and sorting. (B) Example flow cytometry scatter plot for a sample of nanovials loaded with B lymphocytes. Highlighted region corresponds to cell-loaded nanovials shown in the bright field and fluorescence images below.
      Because traditional FACS systems were designed to analyze and sort lymphocytes, and our nanovials are larger and possess a unique morphology when compared to typical mammalian cells, it is important to identify the operational parameters that yield the best performance with nanovials across a range of FACS instruments. In this work we perform a comprehensive characterization of the conditions suitable to analyze and sort nanovials with three different commercial FACS instruments. In particular, we characterized (1) the limit of detection (LOD) and dynamic range for measuring fluorescent protein binding to nanovials, (2) the sorting settings that led to optimal purity and efficiency, and (3) the ability to enrich cell-containing nanovials using scatter signals. To understand the limit of detection and dynamic range of affinity assays on nanovials we performed a dilution sweep using fluorescent streptavidin with biotinylated nanovials and compared with microscopy measurements. Using FACS we found that nanovials could bind ∼108 fluorescent molecules and had detectable concentrations down to ∼104 molecules, an order of magnitude improved over fluorescence microscopy measurements with a cooled, low dark noise, CMOS camera. For each instrument we also systematically adjusted the sorting parameters such as drop delay, sort masks, and sample buffers to identify the optimal conditions for sorting different size ranges of nanovials. We identified conditions that enabled sorting purities as high as 99% and sort recovery efficiencies >90%. Nanovials are unique in that they are transparent and scatter only weakly due to their hydrogel composition and small index of refraction difference compared with surrounding medium. By exploiting this property, we demonstrate that a scatter signal alone can be used to identify cell-containing nanovials from a background of empty nanovials and free cells. We further demonstrate the ability to enrich these populations with purities >90%.

      Materials and methods

      All materials were purchased from Sigma unless otherwise noted. Additional information on the flow cytometers used in this study is listed in Table S1.

      Nanovial fabrication and modification

      Biotinylated nanovials were fabricated using approaches described previously [
      • de Rutte J.
      • Dimatteo R.
      • van Zee M.
      • Damoiseaux R.
      • Di Carlo D.
      Massively parallel encapsulation of single cells with structured microparticles and secretion-based flow sorting.
      ]. Briefly an aqueous two-phase system combined with droplet microfluidics was employed to create hydrogel particles with exposed cavities. The hydrogel precursor component was composed of a 29% w/v solution of 10 kDa, 4 arm PEG norbornene, premixed with 10 mg/mL of 5 kDa biotin-PEG-thiol (Nanocs) to add biotin handles to the nanovial matrix. Different sizes of nanovials (35, 55, and 85 µm) were achieved by adjusting both the size of the droplet generators and the flow rates used to generate droplets. Intra-batch variation in the outer diameter of the nanovials was measured to be <5% for each size condition tested in this work as measured by brightfield microscopy.
      Nanovials were modified using fluorescent streptavidin for microscopy and flow cytometry experiments. A volume of concentrated nanovials was first diluted 5-10X in PBS containing 0.05% w/v Pluronic F-127. An equal volume of solution containing fluorescent streptavidin was added to the diluted solution of nanovials and incubated for a minimum of 15 min to allow binding of streptavidin to the biotin groups on the nanovials. Before analysis the nanovial solution was washed 3 times with PBS/Pluronic solution by centrifuging, aspirating supernatant, and resuspending. During all particle handling steps, tubes and pipette tips were precoated with PBS + 0.05% Pluronic solution to reduce unwanted sticking of nanovials onto surfaces.

      LOD and dynamic range measurements

      Biotinylated nanovials modified with AlexaFluorTM 488 Streptavidin (Invitrogen) were prepared over a range of concentrations using the methods described above. A fixed number of nanovials were added to various concentrations of fluorescent streptavidin solutions and incubated for 15 min to allow for binding. Nanovials were then washed in a large excess of PBS + 0.05% Pluronic solution 3 times to remove any unbound streptavidin molecules

      Microscopy

      The intensity of each particle was measured with an automated image analysis algorithm to detect each nanovial in the image and integrate the intensity across the particle. Localized background subtraction in our automated analysis algorithm was required to reach this performance level by reducing variation due to illumination non-uniformity across the images (Fig. S1). Brightfield and fluorescent images were taken using a Nikon Eclipse TI-S fluorescence microscope equipped with a Photometrics® PRIME™ cooled CMOS camera and SOLA Light Engine LED light source. All fluorescence images were taken using a 20X objective lens (Nikon CFI S Plan Fluor ELWD 20XC, 0.45 NA) with a 1 s exposure time. For quantification of fluorescence intensity a custom MATLAB script was developed. Briefly, nanovial locations were identified using a brightfield image and regions of interest were mapped onto a fluorescent image of the same location. Pixels within the border of the nanovial were integrated to obtain a calculated intensity and then subtracted from local background intensity to compensate for spatial background variation in illumination. >200 nanovials were quantified for each condition.

      Flow cytometry

      Nanovials were assessed for fluorescence signal using three different commercially available flow cytometers. For each trial nanovials were diluted into a 50-fold excess volume of PBS containing 0.05% w/v Pluronic F-127. PMT voltage/gain settings were optimized to ensure that the sample peak from the most concentrated particle sample was completely present within the bounds of the flow plot and unstained samples were between the first and second decades in the fluorescence histogram. Samples were gated based on FSC area and SSC area values and analyzed for Alexa FluorTM 488 signal area. At least 10,000 events were quantified for each condition.

      Sorting optimization

      General sample preparation

      To compare the nanovial sort efficiency across FACS machines we created mixtures of nanovials containing a defined subpopulation with a unique fluorescent label. Labeled nanovials were modified as described above. Labeled target nanovials were then mixed with an excess of background nanovials, labeled with a non-overlapping fluorophore, at desired concentrations.

      FACSAria

      Nanovial samples were sorted on a FACSAriaTM II instrument using a 130 μm nozzle. To optimize sorting efficiency the machine's drop delay values and sorting masks were adjusted until maximum target yields were obtained. Briefly, the instrument drop delay was initially set according to manufacturer instructions using Accudrop calibration beads. Drop delay was then offset +2.0 to −2.0 units from the set point in increments of 0.25. At each drop delay value a 1:5 mixture of unlabeled to fluorescently labeled particles was analyzed and 100 fluorescent target particles were sorted using a single-cell sort mask. Sorted samples were imaged to quantify the total number of recovered particles as well as the fraction of target vs off-target particles sorted. In general, optimal drop delay values shifted slightly between runs and a similar rapid test is recommended before each nanovial sort.
      The stringency of yield, purity, and phase mask settings were also calibrated to identify the optimal sort mask conditions for nanovial recovery. The FACS Aria II system discretizes droplets into 32 segments, and each mask can be varied between the values of 0 and 32 depending on how large a fraction of each interrogated droplet it is desired to affect. Each of the three masks was adjusted in units of 8 between the values of 0–32 (yield and purity) or 0–16 (phase). At each condition an identical mixture of AlexaFluorTM 488 streptavidin and AlexaFluorTM 568 streptavidin labeled particles (1:5 ratio) was analyzed and 100 AlexaFluorTM 488 Streptavidin labeled target particles were sorted into a single well of a well-plate. Total recovery and sample purity were determined using fluorescence microscopy.

      Sony SH800

      Samples were sorted using a 130 μm chip. Because the single cell sort mask on the Sony instrument is more lenient and sorts all droplets in the vicinity of detected target events as long as off target events are not nearby, we found calibration of drop delay values unnecessary. Instead the sort recovery and purity of nanovial sorts was evaluated using the four primary sort masks, yield, normal, purity, and single cell. In each condition 200 AlexaFluorTM 488 Streptavidin target nanovial events were sorted from a background of AlexaFluorTM 647 Streptavidin labeled particles. Recovered samples were analyzed using fluorescent microscopy. Note, the Sony SH800 sorter by default reports forward and back scatter data. Generated back scatter data was converted to side scatter upon export and is referred to as such throughout the manuscript.

      On-chip sort

      35 μm nanovial samples were sorted using an 80 μm chip following similar procedures outlined above. Larger nanovials (55 and 85 μm) were sorted using a 150 μm chip following similar procedures outlined above. Nanovials labeled with AlexaFluorTM 647 streptavidin were spiked into an unlabeled population at a ratio of 1:10. PBS containing 0.05% w/v PluronicTM F-127 was used as a sheath fluid for the sorts. For the sample solution study, we used PBS + 0.05% w/v Pluronic, Ficoll Paque Plus diluted to 40% v/v with PBS + 0.5% Pluronic, OptiprepTM diluted to 44% with PBS + 0.05% w/v Pluronic, and On-chip sample buffer (On-chip Bio). Particle samples were diluted approximately 1, 0.5, and 0.25 million nanovials per mL for 35, 55, and 85 μm nanovials respectively. Event rates were tracked by manually recording at fixed intervals during the sort. Samples were excited with a 637 nm laser and particles containing fluorescent streptavidin were gated based on peak fluorescence height collected through a 676/37 nm emission filter. Sorted samples were transferred from the collection reservoir into a well plate with a pipette and then imaged using a fluorescence microscope. Purity and recovery were quantified using a custom MATLAB image analysis code.

      Assessment of the amount of accessible biotin on particles

      To quantify the number of biotin available per nanovial an HRP ELISA was utilized. Evaluated nanovials were fabricated with 1 or 10 mg Biotin-PEG-thiol per mL of gel precursor material. For each condition, 2 µL of particles were first labeled with various amounts of unconjugated streptavidin ranging from 1 to 31000 ng SA per µL particle solution in 60 µL buffer to cap available biotins. After 45 min of incubation, 80 ng SA-3HRP was added to each vial and allowed to incubate an additional 30 min to allow binding to any free biotin. Particles were then washed 5 times with 1 mL PBS containing 0.05% Pluronic and 0.5% Bovine Serum Albumin (BSA), and aliquots of the supernatant of each wash were collected to confirm the level of remaining HRP is not biasing the experiment in any way. Following the last wash particles from each condition were resuspended in 50 µL washing buffer and moved to a 96 well plate and mixed with 100 µL of HRP substrate. To create a standard curve, 50 µL of washing buffer containing known amounts of SA-3HRP from 0.05 to 30 ng were mixed with 100 µL of substrate. Reactions were allowed to occur for 25 min on a shaker before proper development was observed. Reaction was stopped with 100 µL of 3N NaOH, and the absorbance was measured at 405 nm using a Gen5 plate reader. The high unconjugated SA concentrations was considered to saturate particles, practically not allowing any additional SA-3HRP being adsorbed to the particles, leading to no HRP signal in ELISA. The concentration of unconjugated SA below which the HRP signal started to appear was considered the lowest to saturate the particles and was used to estimate the capacity of particles for binding streptavidin.

      Scatter profiling and scatter-based enrichment of cell-containing nanovials

      All experiments involving animals and animal cells were performed in accordance with the Chancelor's Animal Research Committee ethical guidelines at the University of California Los Angeles under protocol no ARC-2015-125. Primary mouse B cells were isolated from spleen using an EasySepTM Mouse Pan-B cell isolation kit (STEMCELL Technologies) according to manufacturer instructions. Prior to conducting experiments, cells were labeled with a 1 μm solution of CellTrackerTM deep red dye at 37 °C over a period of 30 min. Nanovials were also labeled during this time first through a 15 min incubation within a 2 μg/mL solution of Alexa FluorTM 488 streptavidin and subsequently through a 45 min incubation in a 40 μg/mL solution of biotinylated anti-CD45 cell capture antibodies. After each conjugation step particles were washed three times in an excess of PBS + Pluronic F-127 solution.
      Once labeled with capture antibodies, 6 μL of nanovials were added into individual wells within a 24-well plate. Samples were incubated for 30 min to allow nanovials to settle uniformly on the bottom of each well with their cavities oriented upwards to promote cell capture. Cells were counted on a hemocytometer and roughly 125,000 cells were added into each well. Cells were co-incubated with particles for 1 h to allow adhesion onto the nanovial surfaces. After incubation samples from each well were pooled together and strained through a cell strainer (20 μm, CellTricksTM) to remove unbound cells from the suspension and nanovials and cell loaded nanovials were recovered by flipping the cell strainer and washing with buffer solution.
      Recovered nanovials were sorted via flow cytometry using a 130 μm nozzle on the Sony SH800 system in single cell mode. For fluorescence-based sorting, Alexa FluorTM 488 and CellTrackerTM deep red fluorescence was evaluated for all detected events and cell loaded nanovials were gated based on high fluorescence in both channels. For scatter-based sorting, samples were evaluated either using plots of FSC-A/SSC-A or FSC-H/SSC-A and cell-loaded nanovials were gated based on high scatter. For each gating strategy 500 events were sorted into a single well of a 96-well plate in triplicate. Post sort images were taken using fluorescence microscopy and the fraction of free cells, empty nanovials and filled nanovials was counted manually.

      Results

      In our studies we characterized the use of nanovials with three separate commercially available fluorescence activated cell sorters: (1) BD FACSAriaTM II (BD Biosciences), (2) Sony SH800 (Sony Biotechnology), and (3) On-chip Sort (On-Chip Biotechnologies). We compared each instrument's ability to detect fluorescence signal on nanovials with fluorescence microscopy, identified parameters to maximize the purity and efficiency of nanovial sorting, and characterized the ability to identify cell-loaded nanovials using scatter measurements alone. By identifying the optimal instrument parameters and trade-offs between each system, we aim to guide future experimental effort by easing cross-platform accessibility and aiding the selection of appropriate instrumentation to meet experimental needs.

      Dynamic range and limit of detection for microscopy and flow cytometry analysis of nanovials

      We found that nanovials could be analyzed over a large dynamic range using fluorescence microscopy with a cooled CMOS camera as well as flow cytometry. As previously shown, a unique feature of the nanovials is the ability to bind biomolecules to affinity reagents attached to the hydrogel matrix and detect their presence using fluorescent labels [
      • de Rutte J.
      • Dimatteo R.
      • van Zee M.
      • Damoiseaux R.
      • Di Carlo D.
      Massively parallel encapsulation of single cells with structured microparticles and secretion-based flow sorting.
      ]. A core advantage of flow cytometers is their ability to measure fluorescence signals associated with 100,000’s to 10’s of millions of unique events, such as microbeads and cells, easily and with limits of detections below 2000 molecules for common stains and down to 100’s of molecules for optimized systems [
      • Sklar L.A.
      • Carter M.B.
      • Edwards B.S.
      Flow Cytometry for drug discovery, receptor pharmacology and high-throughput screening.
      ]. We assessed the ability to detect fluorescent protein bound to nanovials using flow cytometry mimicking the signal from a biomolecular assay on a nanovial (e.g. single cell secretion assay). We bound fluorescent streptavidin to biotinylated nanovials across a 4-log range of concentrations to characterize the limit of detection and the max dynamic range of nanovial measurements with the instruments (Fig. 2, Table 1). Fluorescence microscopy measurements were performed in parallel as a baseline reference.
      Fig 2
      Fig. 2LOD and dynamic range for detecting nanovials using fluorescence microscopy and flow cytometry. (A) Biotinylated 35 µm nanovials were labeled with Alexa FluorTM 488 streptavidin (AF488 SA) across a 4-log range of molecules per nanovial. Example brightfield and corresponding fluorescence images are shown over the range of conditions. Fluorescence images of particles with the same LUT values are shown in the middle column and images with LUTs maximized for contrast are shown to the right. (B) The fluorescently-labeled nanovials were analyzed using fluorescence microscopy and three flow cytometers (BD FACSAriaTM II, SONY SH800, On-chip Sort). Intensity distributions for each condition are shown. Inset graph show intensity distributions at the lower concentration range using increased gain/voltage settings for the flow cytometers. (C) The mean intensity for each condition is plotted against the number of molecules per nanovial on a log-log plot. The vertical dashed line indicates the lowest concentration detected for each instrument (LOD) using gain/voltage settings tuned for high sensitivity. The shaded region indicates the max dynamic range when gain/voltage is adjusted for low electronic noise. For microscopy measurements n > 200 nanovials were measured for each condition. For the FACSAriaTM and Sony SH800 a total of 10,000 events were collected for each condition and for the On-Chip sort n > 2000 events were collected for each condition.
      Table 1Summary of Nanovial analysis and sorting capabilities across different FACS instruments.
      InstrumentNanovial Diameter (μm)Working Conc. (nanovials × 106/mL)Event rate (s−1)PurityRecovery EfficiencyLOD (# mol)Max Dynamic Range
      FACSAria II351–4200–80095-99%∼40–60%∼10
      • Lu Y.
      • Xue Q.
      • Eisele M.R.
      • Sulistijo E.S.
      • Brower K.
      • Han L.
      • Amir E.A.D.
      • Pe'er D.
      • Miller-Jensen K.
      • Fan R.
      Highly multiplexed profiling of single-cell effector functions reveals deep functional heterogeneity in response to pathogenic ligands.
      3 Log
      550.25–150–20095-99%∼40–60%
      85NANANANA
      Sony SH800351–4200-80095-99%70-80%∼10
      • Lu Y.
      • Xue Q.
      • Eisele M.R.
      • Sulistijo E.S.
      • Brower K.
      • Han L.
      • Amir E.A.D.
      • Pe'er D.
      • Miller-Jensen K.
      • Fan R.
      Highly multiplexed profiling of single-cell effector functions reveals deep functional heterogeneity in response to pathogenic ligands.
      3 Log
      550.25–150–200NA
      Able to perform analysis, but had low sorting recovery.
      NA
      Able to perform analysis, but had low sorting recovery.
      85NANANANA
      On-chip Sort350.5–2100–30080–90%70–95%∼10
      • Jin A.
      • Ozawa T.
      • Tajiri K.
      • Obata T.
      • Kondo S.
      • Kinoshita K.
      • Kadowaki S.
      • Takahashi K.
      • Sugiyama T.
      • Kishi H.
      • Muraguchi A.
      A rapid and efficient single-cell manipulation method for screening antigen-specific antibody-secreting cells from human peripheral blood.
      2 Log
      550.25–0.520–5080–90%70–95%
      850.1–120–10080–90%70–95%
      low asterisk Able to perform analysis, but had low sorting recovery.
      Using fluorescence microscopy, we found that signal could be detected on the nanovials, with 3 standard deviations above background noise, which we define as the limit of detection (LOD), starting at around 100,000 molecules per particle after background subtraction (Fig. S1). Fluorescence images reveal that the streptavidin first binds to the outer surface of the nanovials at lower concentrations and then can transport and react further into the hydrogel matrix of the particles as biotin groups become saturated. It was found that biotin groups were saturated at ∼108 molecules per nanovial for the fabrication conditions used as confirmed by both fluorescence measurements (Fig. 2) and colorimetric assay (Fig. S2). Of note, to calculate the number of bound molecules per nanovial we assume that nearly all fluorescent streptavidin in solution binds to the particles which is likely an over-estimate. Thus, the values reported herein may be a slight underestimate of the true sensitivity of the system.
      We found that measurements with the flow cytometers yielded comparable, and in some instances improved, limit of detection readings when compared to our fluorescence microscopy setup (Fig. S1). Both the FACS Aria II and Sony SH800 sorters were able to detect as few as ∼ 10,000 molecules / nanovial. This corresponds closely to the expected LOD for measurements of cell surface markers with similar fluorophores using flow cytometry [
      • Zola H.
      High-sensitivity immunofluorescence/flow cytometry: detection of cytokine receptors and other low-abundance membrane molecules.
      ]. In order to capture the high signal intensity of our high concentration positive labeled samples, particles labeled with <105 molecules were initially difficult to distinguish from non-labeled controls. However, through adjustment of the PMT voltage or gain parameters lower signal samples could be clearly resolved (Fig. 2B,C). This suggests that optimal parameters should be selected based on experimental need, with lower PMT voltage favoring increased resolution and higher values providing greater dynamic range. The FACSAriaTM instrument was able to resolve the lowest concentration tested with minor gain adjustments indicating a higher sensitivity. This is most likely due to the glass capillary used at the interrogation site as well as dedicated PMT's for each channel. Both the Sony SH800 and On-chip Sort have disposable polymer chips in the optical path which may slightly reduce the optical performance. Despite this the SH800 had a limit of detection comparable with the FACSAriaTM and a 3-log max dynamic range. The measured LODs are sufficient to detect captured secretions from single cells in a reasonable time period. For example, T cells were reported to secrete cytokines at rates of ∼1–10 molecules per second [
      • Han Q.
      • Bradshaw E.M.
      • Nilsson B.
      • Hafler D.A.
      • Love J.C
      Multidimensional analysis of the frequencies and rates of cytokine secretion from single cells by quantitative microengraving.
      ], while antibody-secreting plasma B cells can produce ∼2000 antibodies per second [

      B. Alberts; A. Johnson; J. Lewis; M. Raff; K. Roberts; P. Walter B Cells and Antibodies. 2002.

      ]. For low abundance targets, fluorophores such as R-phycoerythrin or polymer dyes which can be an order of magnitude brighter than other common dyes may be applied to further improve the sensitivity of the system [
      • Maciorowski Z.
      • Chattopadhyay P.K.
      • Jain P.
      Basic multicolor flow cytometry.
      ].

      Sorting nanovials using FACS

      We identified optimal settings and conditions for sorting nanovials of different sizes by systematically adjusting sorting settings on the different FACS instruments. Each instrument has unique settings and parameters that can be adjusted so we discuss the characterization of each system separately.

      FACSAriaTM II

      To test the sorting performance of the FACSAriaTM II a population of fluorescently-labeled nanovials were spiked into a population of nanovials with a second label at a ratio of 1:5 and sorted into a 96-well plate (Fig. 3). We first identified single nanovial events based on the side scatter area and forward scatter area, and then gated out target events based on high AF488 fluorescence and low AF568 fluorescence (Fig. 3B). Gated events were sorted and then imaged using fluorescence microscopy to assess both the purity of the sort and the sort recovery efficiency as defined by:
      SortRecoveryPurity=numberoftargetnanovialsinwellnumberoftotalnanovialsinwell×100%


      SortRecoveryEfficiency=numberoftargetnanovialsinwellnumberofreportedsorts×100%


      Fig 3
      Fig. 3Optimized Nanovial sorting parameters on the FACSAriaTM II. (A) Schematic showing an overview of the FACSAriaTM system and important parameters that can be adjusted to fine tune sorting. The drop delay defines the expected time between the interrogation of an event and formation of the aerosol droplets. Sorting masks adjust the stringency by which an event will be sorted based on its predicted positioning. (B) Scatter and fluorescence plots of Alexa FluorTM 488 streptavidin labeled nanovials (35 µm) spiked into Alexa FluorTM 568 streptavidin labeled nanovials (1:5 ratio). Inset image shows a post sort example. Scale: 200 µm. (C) It was found that for both 35 and 55 µm nanovials on a 130 µm nozzle that manually adjusting the drop delay from the system calibrated baseline resulted in improved sort recovery efficiency. For each condition 100 events were sorted into different wells of a 96-well plate and performance was quantified by counting recovered events. It was found that over a 5-fold improvement in recovery was achieved by adjusting the delay settings. (D) Various sort masks were tested and evaluated based on both purity of the sort and the sort recovery efficiency. Fluorescently labeled nanovials were spiked into an unlabeled population at a ratio of 1:5, sorted and then characterized using fluorescence microscopy. At each condition 200 target events were sorted in triplicate. The numbers above each bar report the average number of target events recovered (green) over the average total number of recovered events over the three sorts. The optimal settings for purity and sort recovery efficiency are highlighted in purple.
      The Aria system utilizes an integrated Accudrop system to calibrate the drop delay between the detected event and the sorting point. This value can shift from day to day and it is recommended to calibrate it before each experiment utilizing the associated Accudrop calibration beads. We found that the sort recovery efficiency for both 35 and 55 micron diameter nanovials was low when using the baseline Accudrop calibrated value (Fig. 3C). In previous work manual adjustment of the drop delay parameter was shown to improve results for double emulsion sorting [
      • Brower K.K.
      • Brower K.K.
      • Carswell-Crumpton C.
      • Klemm S.
      • Cruz B.
      • Kim G.
      • Calhoun S.G.K.
      • Nichols L.
      • Fordyce P.M.
      • Fordyce P.M.
      • Fordyce P.M.
      • Fordyce P.M.
      Double emulsion flow cytometry with high-throughput single droplet isolation and nucleic acid recovery.
      ]. Sweeping drop delay measurements manually in 0.25 increments we found that a delay value of −1.75 offset from the Accudrop value led to the best recovery efficiency for both nanovial sizes tested with the 130 µm nozzle, improving recovery 4–8 fold over baseline for 55 and 35 µm nanovials. Similarly, we found manual adjustment improved recovery with 35 µm nanovials sorted with a 100 µm nozzle (Fig. S3). We hypothesize that this deviation is due to a change in velocity of the nanovials in the nozzle. Nanovials cover a larger fraction of the cross section of the flow compared to a cell which could lead to variation in velocity for a non-uniform flow cross-section. For best practice it is recommended to perform manual drop delay adjustment before each sort or at least perform a test for different instruments or nozzles.
      Sorting masks can be adjusted to maximize purity or sort recovery efficiency. The Aria system defines its masks in reference to the interrogated droplet (the droplet where the target is predicted to be), the leading droplet (the droplet in front of the target), and the trailing droplet (that follows the target droplet) (Fig. 3A). Each droplet is broken up into 32 fractions that define the relative coverage of each mask. The purity mask defines the fraction of the leading and trailing droplet in which the presence of another event will abort the sort. The phase mask defines which region of the interrogated droplet the target should be in to proceed with a sort. To prevent sample loss from target events partitioning into the incorrect droplet, phase masks will only trigger sorts when events are “centered” within a droplet, rather than localized to the droplets leading or trailing edges. Higher phase values define the center region of each droplet more stringently than lower values.
      Using a similar strategy as with the drop delay, we swept these conditions and quantified the sort purity, sort recovery efficiency, and the abort rate for each (Fig. 3D). As expected, the best purity performance was found for the most restrictive purity masks. We further found that more restrictive phase masks improved the sort recovery from 29 to 55%. We hypothesize that since the particles are larger than normally sorted objects, target events near the edge of each drop may disrupt the stability of the aerosolizing stream, leading to over or under-deflection during sorting and a missed collection in the target container.

      Sony SH800

      To test the sorting performance of the Sony SH800 a population of fluorescently-labeled nanovials were spiked into a population of nanovials with a second label at a ratio of 1:5 and sorted into a 96-well plate using its index sorting feature (Fig. 4). We found that both 35 and 55 µm nanovials were compatible with the larger 130 µm microfluidic chip available for the system and focused optimization for this setup. For the SH800 sort masking takes into account the target event droplet, adjacent droplets and the relative position of the target event to off target events (Fig. 4A). We chose the four following sorting modes to evaluate in this study: yield, normal, purity, and single cell. Descriptions and sorting criteria are shown in Fig. 4B. To characterize each sorting mode, AlexaFluorTM 488 labeled nanovials were spiked into a population of AlexaFluorTM 647 labeled nanovials at a 1:5 ratio. Single nanovial events were gated using forward scatter area and side scatter area, and target events were sorted based on high AF488 fluorescence and low AF647 fluorescence (Fig. 4C). For each condition 200 nanovials were sorted into separate wells of a 96-well plate and sort purity and sort recovery efficiency was quantified using fluorescence microscopy. As expected, the lowest purity resulted from yield mode, as this mode does not reject a sort in the presence of adjacent off-target events. Despite this, purity was quite high (96%) and increased incrementally with the more restricted sorting modes, with the highest being for Single Cell recovery mode which had 100% purity across 3 separate sorts. Interestingly, the Single Cell mode had significantly higher sort recovery efficiency compared to any of the other modes. This was attributed to the fact that single cell mode will sort the target event droplet and both of the adjacent droplets. As described above, the larger size of the nanovials can potentially cause more inconsistency in sort timing as well as affect droplet deflection accuracy. Thereby including 3 droplets in the sort is expected to have better recovery compared to the other modes which only sort 1–2 of the droplets depending on the event conditions. One downside to this is that the Single Cell mode will only sort if there are no overlapping events in the adjacent droplets leading to a higher frequency of aborted target events during analysis of more concentrated nanovial samples (Fig. S4). However, we note that even at the highest nanovial concentration tested in this work, the single-cell mode abort rate of ∼26% was sufficiently counteracted by the increased recovery efficiency and sort purity to obtain higher total target yields than could be achieved using any alternative sort mask at any sample concentration.
      Fig 4
      Fig. 4Optimized Nanovial sorting parameters on the Sony SH800. (A) An overview schematic of the Sony SH800 system and sorting definitions. Sorting decisions are typically made in reference to the relative position of the target event and the distance between target and off target events. (B) Overview of four common sorting modes on the SH800 and the sort decision logic for the sort mode. (C) Scatter and fluorescence plots of Alexa FluorTM 488 streptavidin labeled nanovials (35 µm) spiked into Alexa FluorTM 647 streptavidin labeled nanovials (1:5 ratio). Inset image shows post sort example. Scale: 200 µm. (D) The different sort modes were tested and evaluated based on both purity of the sort and the sort recovery efficiency. Fluorescently-labeled nanovials were spiked into a non-labeled population at a ratio of 1:5, sorted and then characterized using fluorescence microscopy. At each condition 200 target events were sorted in triplicate. The numbers above each bar report the average number of target events recovered (green) over the average total number of recovered events over the three sorts. The optimal settings for purity and sort recovery efficiency are highlighted.

      On-chip sort

      The On-chip Sort system is unique compared to the other two systems in that it is aerosol free and has a larger range of particle size compatibilities. Like with the studies using the SH800 and FACSAriaTM a fluorescently labeled subpopulation of nanovials was spiked into a larger population of unlabeled nanovials at a 1:10 ratio and sorted to characterize purity and recovery efficiency. Forward scatter height, side scatter height and fluorescence were used to gate single target nanovial events for sorting (Fig. 5A). We found that we could easily sort nanovials up to 85 microns in diameter with the larger 150 micron chip available for the instrument achieving ∼10 fold enrichment (Fig. 5B). The instrument does not have a sample mixer and it was noted that nanovials settled and tended to aggregate at the sample inlet (Fig. 5C). This often resulted in an initial moderate event rate followed by little to no events and eventually to a large spike in events as the last of the sample fluid was pushed through the channel. To address this, we tested Ficoll and Optiprep, two well-known density-modulating agents used for cell separation processes to tune the density of the sample solution such that the nanovials are nearly neutrally buoyant. We also tested the On-chip sample buffer which increases settling time through increased viscosity. Each of these experiments was conducted with 55-micron nanovials as these samples gave us the most inconsistent event rate in PBS leading to lower purities. In addition to the different solutions the On-chip instrument has two standard pressure modes, low-pressure mode, optimized for lower viscosity solutions such as PBS and high-pressure mode used for sorting in more viscous solutions. We found that in general using the higher-pressure mode as well as using the buoyancy matched and higher viscosity solutions led to much more consistent event rates and improved the purity and recovery rates. Optiprep resulted in the highest purity of the solutions, however, it tended to cause large variation in scatter signal which may be non-optimal depending on application requirements (Fig. S5). For further studies we utilized the Ficoll sample solution as it had the best balance of sorting results and optical properties. Although purity was lower for the On-chip Sort, the sort recovery efficiency was high. Since no aerosols are generated, the device is more forgiving with compatible nanovial sizes and there is no risk of sorted droplets missing the recovery container. During default operation, the On-chip Sort targets maximizing yield. However, because sample loading, sorting, and collection all occurs within the same self-contained chip interface, higher purities can be achieved by collecting and re-sorting previously sorted samples.
      Fig 5
      Fig. 5Optimized Nanovial sorting parameters on the On-chip Sort. (A) Example scatter plot of 55 µm nanovials and fluorescence gating of Alexa FluorTM 647 labeled nanovial sub-populations. (B) Pre and post sort images of 35, 55, and 85 µm nanovials. Fluorescence and brightfield channels are overlaid to aid in visualization. (C) Density matching to increase uniformity of sort event rates and improve performance. Event rate over time shown for 55 µm nanovials suspended in different buffers. (D) Purity and recovery efficiency of 55 µm nanovials for different buffers and pressure settings.

      Scatter based enrichment of cell-containing subpopulations

      Unique scatter signatures can be used to identify populations of cell-containing nanovials among a background of free cells and nanovial-containing cells. Scatter based gating is powerful for both cleaning up data by removing debris or unwanted populations of cells and identifying sub-populations of interest. This can be advantageous for getting more accurate results, purer samples, and can free up fluorescence channels for additional stains. In previous nanovial workflows, cell-containing nanovial populations were either not discriminated from other populations or were identified using fluorescent staining.
      To improve the purity of the sample and free up fluorescence channels we systematically identify unique scatter signatures associated with nanovials containing cells using fluorescent stains as a ground truth. B lymphocytes stained with Deep Red CellTrackerTM were bound to Alexa FluorTM 488 labeled nanovials. A mixture of the cell-containing nanovials, empty nanovials, and free cells were analyzed using the Sony SH800 (Fig. 6). Using the unique fluorescent signatures of each population we were able to back gate each target population to their associated scatter signals. In general cells were observed to have lower side scatter and higher forward scatter, while nanovials had higher side scatter and lower forward scatter. It was noted that both scatter height or area could be used to identify the subpopulation of interest (Fig. S6). Cell-containing nanovials were enriched by gating off of high forward and side scatter (Fig. 6C). As a control, cell-containing nanovials were also sorted based off of a fluorescence gate (Fig. 6B). Samples were then imaged using fluorescence microscopy and analyzed to compare enrichment based on a fluorescence signature and scatter signature (Fig. 6D). Nanovials containing cells were successfully enriched over 10-fold using both the scatter and fluorescence gating with no significant difference in purity (90%, and 88% respectively).
      Fig 6
      Fig. 6Nanovial scatter characterization and scatter-based enrichment of cell-containing nanovials. (A) A mixture of free cells, nanovials, and cell-containing nanovials were analyzed to identify unique scatter signatures for each population and then sorted to enrich cell-containing nanovials. (B) The nanovials and cells were stained with unique fluorescent labels in order to accurately discriminate each population. (C) The fluorescent signatures were then correlated back to the scatter plots to identify each population based on scatter signature. (D) Samples were sorted based on fluorescent gating alone and scatter gating alone and imaged to compare enrichment of the cell-containing nanovials.

      Discussion

      In this work we systematically characterized the analysis and sorting of hydrogel nanovials with commercial FACS instruments (Table 1). By employing the high signal to noise of FACS readouts we are able to detect protein bound to nanovial surfaces with higher fluorescence intensity and dynamic range than high end fluorescence microscopy readouts, improving performance in terms of LOD and dynamic range by at least an order of magnitude. We characterized sorting performance on the different FACS instruments and identified optimal parameters that improved purity to up to 93% and sample recovery up to 8-fold over baseline settings. Using more widely-available FACS instruments such as the FACSAriaTM and Sony SH800 we were able to easily sort 35 μm nanovials, and larger 55 μm nanovials were successfully sorted with the FACSAriaTM. Using the microfluidic chip based On-chip Sort system we were able to sort nanovials up to 85 microns in diameter which could further open up possibilities for performing assays with larger cells, multiple cells, or cell colonies. Further, this instrument is capable of sorting water-in-oil emulsions, enabling analysis and sorting of emulsified nanovials which may open up other homogenous assays such as substrate-based FRET or nucleic acid assays.
      Beyond the nanovials we use here, our results may provide a guide for other large particle sorting using FACS, such as for sorting of hydrogel beads [
      • Chokkalingam V.
      • Tel J.
      • Wimmers F.
      • Liu X.
      • Semenov S.
      • Thiele J.
      • Figdor C.G.
      • Huck W.T.S.
      Probing cellular heterogeneity in cytokine-secreting immune cells using droplet-based microfluidics.
      ,
      • Li M.
      • van Zee M.
      • Riche C.T.
      • Tofig B.
      • Gallaher S.D.
      • Merchant S.S.
      • Damoiseaux R.
      • Goda K.
      • Carlo Di
      • Gelatin D.A
      Microdroplet platform for high-throughput sorting of hyperproducing single-cell-derived microalgal clones.
      ,
      • de Rutte J.M.
      • Koh J.
      • Di Carlo D.
      Scalable high-throughput production of modular microgels for in situ assembly of microporous tissue scaffolds.
      ], hollow shell (e.g. PicoShell) particles [
      • Zee M.
      • Rutte J.
      • Rumyan R.
      • Williamson C.
      • Burnes T.
      • Radakovits R.
      • Eugenio A.S.
      • Badih S.
      • Lee D.H.
      • Archang M.
      • Carlo D.Di
      High-throughput selection of microalgae based on biomass accumulation rates in production environments using picoshell particles.
      ], plant cells, cell clusters, and oocytes. In addition our results can inform sorting of double emulsions (e.g. water-in-oil-in-water), which have been applied to microfluidic droplet assays [
      • Brower K.K.
      • Khariton M.
      • Suzuki P.H.
      • Still C.
      • Kim G.
      • Calhoun S.G.K.
      • Qi L.S.
      • Wang B.
      • Fordyce P.M.
      Double emulsion picoreactors for high-throughput single-cell encapsulation and phenotyping via FACS.
      ,
      • Zinchenko A.
      • Devenish S.R.A.
      • Kintses B.
      • Colin P.Y.
      • Fischlechner M.
      • Hollfelder F.
      One in a million: flow cytometric sorting of single cell-lysate assays in monodisperse picolitre double emulsion droplets for directed evolution.
      ,
      • Terekhov S.S.
      • Smirnov I.V.
      • Stepanova A.V.
      • Bobik T.V.
      • Mokrushina Y.A.
      • Ponomarenko N.A.
      • Belogurov A.A.
      • Rubtsova M.P.
      • Kartseva O.V.
      • Gomzikova M.O.
      • Moskovtsev A.A.
      • Bukatin A.S.
      • Dubina M.V.
      • Kostryukova E.S.
      • Babenko V.V.
      • Vakhitova M.T.
      • Manolov A.I.
      • Malakhova M.V.
      • Kornienko M.A.
      • Tyakht A.V.
      • Vanyushkina A.A.
      • Ilina E.N.
      • Masson P.
      • Gabibov A.G.
      • Altman S.
      Microfluidic droplet platform for ultrahigh-throughput single-cell screening of biodiversity.
      ].
      Given the wide compatibility and analysis and sorting performance, we anticipate that nanovial-based approaches can transform almost any bulk assay currently performed in a well plate and analyzed with a well-plate reader into a single-cell assay performed on a nanovial and analyzed with a flow cytometer. The presence of a cell on the solid phase of the nanovial unlocks the capability to perform sequential assays on the same single cell where information from the assay is stored on the nanovial for downstream sensitive analysis, e.g. by flow cytometry. Unlike traditional well plate assays, the ability to easily sort the cell associated with a reaction in a nanovial opens up new opportunities for screening at throughputs unmatched by liquid handling and microfluidic techniques. Imaging of nanovials using next-generation imaging flow cytometry and image-activated cell sorters [
      • Nitta N.
      • Sugimura T.
      • Isozaki A.
      • Mikami H.
      • Hiraki K.
      • Sakuma S.
      • Iino T.
      • Arai F.
      • Endo T.
      • Fujiwaki Y.
      • Fukuzawa H.
      • Hase M.
      • Hayakawa T.
      • Hiramatsu K.
      • Hoshino Y.
      • Inaba M.
      • Ito T.
      • Karakawa H.
      • Kasai Y.
      • Koizumi K.
      • Lee S.W.
      • Lei C.
      • Li M.
      • Maeno T.
      • Matsusaka S.
      • Murakami D.
      • Nakagawa A.
      • Oguchi Y.
      • Oikawa M.
      • Ota T.
      • Shiba K.
      • Shintaku H.
      • Shirasaki Y.
      • Suga K.
      • Suzuki Y.
      • Suzuki N.
      • Tanaka Y.
      • Tezuka H.
      • Toyokawa C.
      • Yalikun Y.
      • Yamada M.
      • Yamagishi M.
      • Yamano T.
      • Yasumoto A.
      • Yatomi Y.
      • Yazawa M.
      • Di Carlo D.
      • Hosokawa Y.
      • Uemura S.
      • Ozeki Y.
      • Goda K.
      Intelligent image-activated cell sorting.
      ,
      • Isozaki A.
      • Mikami H.
      • Tezuka H.
      • Matsumura H.
      • Huang K.
      • Akamine M.
      • Hiramatsu K.
      • Iino T.
      • Ito T.
      • Karakawa H.
      • Kasai Y.
      • Li Y.
      • Nakagawa Y.
      • Ohnuki S.
      • Ota T.
      • Qian Y.
      • Sakuma S.
      • Sekiya T.
      • Shirasaki Y.
      • Suzuki N.
      • Tayyabi E.
      • Wakamiya T.
      • Xu M.
      • Yamagishi M.
      • Yan H.
      • Yu Q.
      • Yan S.
      • Yuan D.
      • Zhang W.
      • Zhao Y.
      • Arai F.
      • Campbell R.E.
      • Danelon C.
      • Di Carlo D.
      • Hiraki K.
      • Hoshino Y.
      • Hosokawa Y.
      • Inaba M.
      • Nakagawa A.
      • Ohya Y.
      • Oikawa M.
      • Uemura S.
      • Ozeki Y.
      • Sugimura T.
      • Nitta N.
      • Goda K.
      Intelligent image-activated cell sorting 2.0.
      ,
      • Nitta N.
      • Iino T.
      • Isozaki A.
      • Yamagishi M.
      • Kitahama Y.
      • Sakuma S.
      • Suzuki Y.
      • Tezuka H.
      • Oikawa M.
      • Arai F.
      • Asai T.
      • Deng D.
      • Fukuzawa H.
      • Hase M.
      • Hasunuma T.
      • Hayakawa T.
      • Hiraki K.
      • Hiramatsu K.
      • Hoshino Y.
      • Inaba M.
      • Inoue Y.
      • Ito T.
      • Kajikawa M.
      • Karakawa H.
      • Kasai Y.
      • Kato Y.
      • Kobayashi H.
      • Lei C.
      • Matsusaka S.
      • Mikami H.
      • Nakagawa A.
      • Numata K.
      • Ota T.
      • Sekiya T.
      • Shiba K.
      • Shirasaki Y.
      • Suzuki N.
      • Tanaka S.
      • Ueno S.
      • Watarai H.
      • Yamano T.
      • Yazawa M.
      • Yonamine Y.
      • Di Carlo D.
      • Hosokawa Y.
      • Uemura S.
      • Sugimura T.
      • Ozeki Y.
      • Goda K.
      Raman image-activated cell sorting.
      ], may further expand the capabilities, leading to high-content single-cell assays at extreme throughputs that can drive the next decades of discovery.

      CRediT authorship contribution statement

      Joseph de Rutte: Conceptualization, Visualization, Formal analysis, Writing – review & editing, Writing – original draft. Robert Dimatteo: Conceptualization, Methodology, Visualization, Formal analysis, Writing – review & editing. Sheldon Zhu: Visualization, Formal analysis. Maani M Archang: Resources. Dino Di Carlo: Conceptualization, Visualization, Writing – review & editing.

      Declaration of Competing Interest

      The Regents of the University of California have filed a provisional patent related to the work described in the manuscript that D.D., J.D., and R.Di. are inventors on. J.D. and D.D. are co-founders and have a financial interest in Partillion Bioscience, which is commercializing the Nanovial platform. S. Z. if an employee of and has financial interests in Partillion Bioscience.

      Acknowledgments

      We acknowledge support from the National Institutes of Health Grants R21GM126414, T32GM008042, T32AR071307, R43GM142252 and the Simons Foundation Math+X Investigator Award #510776. Sorting experiments were performed in the UCLA Jonsson Comprehensive Cancer Center (JCCC) and Center for AIDS Research Flow Cytometry Core Facility that is supported by National Institutes of Health awards P30 CA016042 and 5P30 AI028697, and by the JCCC, the UCLA AIDS Institute, the David Geffen School of Medicine at UCLA, the UCLA Chancellor's Office, and the UCLA Vice Chancellor's Office of Research. The authors would like to thank Iris Williams for her technical expertise and guidance related to nanovial sorting using the FACSAriaTM II instrument.

      Appendix. Supplementary materials

      References

        • Di Carlo D.
        • Lee L.P.
        Dynamic single-cell analysis for quantitative biology.
        Anal Chem. 2006; (American Chemical Society December 1): 7918-7925https://doi.org/10.1021/ac069490p
        • Di Carlo D.
        • Tse H.T.K.
        • Gossett D.R.
        Introduction: why analyze single cells?.
        Methods Mol Biol. 2012; 853: 1-10https://doi.org/10.1007/978-1-61779-567-1_1
        • Chen Z.
        • Chen J.J.
        • Fan R.
        Single-cell protein secretion detection and profiling.
        Annu Rev Anal Chem. 2019; https://doi.org/10.1146/annurev-anchem-061318-115055
        • Lu Y.
        • Xue Q.
        • Eisele M.R.
        • Sulistijo E.S.
        • Brower K.
        • Han L.
        • Amir E.A.D.
        • Pe'er D.
        • Miller-Jensen K.
        • Fan R.
        Highly multiplexed profiling of single-cell effector functions reveals deep functional heterogeneity in response to pathogenic ligands.
        Proc Natl Acad Sci U S A. 2015; 112: E607-E615https://doi.org/10.1073/pnas.1416756112
        • Jin A.
        • Ozawa T.
        • Tajiri K.
        • Obata T.
        • Kondo S.
        • Kinoshita K.
        • Kadowaki S.
        • Takahashi K.
        • Sugiyama T.
        • Kishi H.
        • Muraguchi A.
        A rapid and efficient single-cell manipulation method for screening antigen-specific antibody-secreting cells from human peripheral blood.
        Nat Med. 2009; 15: 1088-1092https://doi.org/10.1038/nm.1966
        • Jones B.E.
        • Brown-Augsburger P.L.
        • Corbett K.S.
        • Westendorf K.
        • Davies J.
        • Cujec T.P.
        • Wiethoff C.M.
        • Blackbourne J.L.
        • Heinz B.A.
        • Foster D.
        • Higgs R.E.
        • Balasubramaniam D.
        • Wang L.
        • Zhang Y.
        • Yang E.S.
        • Bidshahri R.
        • Kraft L.
        • Hwang Y.
        • Žentelis S.
        • Jepson K.R.
        • Goya R.
        • Smith M.A.
        • Collins D.W.
        • Hinshaw S.J.
        • Tycho S.A.
        • Pellacani D.
        • Xiang P.
        • Muthuraman K.
        • Sobhanifar S.
        • Piper M.H.
        • Triana F.J.
        • Hendle J.
        • Pustilnik A.
        • Adams A.C.
        • Berens S.J.
        • Baric R.S.
        • Martinez D.R.
        • Cross R.W.
        • Geisbert T.W.
        • Borisevich V.
        • Abiona O.
        • Belli H.M.
        • de Vries M.
        • Mohamed A.
        • Dittmann M.
        • Samanovic M.I.
        • Mulligan M.J.
        • Goldsmith J.A.
        • Hsieh C.L.
        • Johnson N.V.
        • Wrapp D.
        • McLellan J.S.
        • Barnhart B.C.
        • Graham B.S.
        • Mascola J.R.
        • Hansen C.L.
        • Falconer E.
        The neutralizing antibody, LY-CoV555, protects against SARS-CoV-2 infection in non-human primates.
        Sci Transl Med. 2021; 13: eabf1906https://doi.org/10.1126/scitranslmed.abf1906
        • Winters A.
        • McFadden K.
        • Bergen J.
        • Landas J.
        • Berry K.A.
        • Gonzalez A.
        • Salimi-Moosavi H.
        • Murawsky C.M.
        • Tagari P.
        • King C.T.
        Rapid single B cell antibody discovery using nanopens and structured light.
        MAbs. 2019; 11: 1025-1035https://doi.org/10.1080/19420862.2019.1624126
        • Le K.
        • Tan C.
        • Gupta S.
        • Guhan T.
        • Barkhordarian H.
        • Lull J.
        • Stevens J.
        • Munro T.
        A novel mammalian cell line development platform utilizing nanofluidics and optoelectro positioning technology.
        Biotechnol Prog. 2018; 34: 1438-1446https://doi.org/10.1002/btpr.2690
        • Josephides D.
        • Davoli S.
        • Whitley W.
        • Ruis R.
        • Salter R.
        • Gokkaya S.
        • Vallet M.
        • Matthews D.
        • Benazzi G.
        • Shvets E.
        • Gesellchen F.
        • Geere D.
        • Liu X.
        • Li X.
        • Mackworth B.
        • Young W.
        • Owen Z.
        • Smith C.
        • Starkie D.
        • White J.
        • Sweeney B.
        • Hinchliffe M.
        • Tickle S.
        • Lightwood D.J.
        • Rehak M.
        • Craig F.F.
        • Holmes D.
        Cyto-mine: an integrated, picodroplet system for high-throughput single-cell analysis, sorting, dispensing, and monoclonality assurance.
        SLAS Technol Transl Life Sci Innov. 2020; 25: 177-189https://doi.org/10.1177/2472630319892571
        • Chokkalingam V.
        • Tel J.
        • Wimmers F.
        • Liu X.
        • Semenov S.
        • Thiele J.
        • Figdor C.G.
        • Huck W.T.S.
        Probing cellular heterogeneity in cytokine-secreting immune cells using droplet-based microfluidics.
        Lab Chip. 2013; 13: 4740https://doi.org/10.1039/c3lc50945a
        • Li M.
        • van Zee M.
        • Riche C.T.
        • Tofig B.
        • Gallaher S.D.
        • Merchant S.S.
        • Damoiseaux R.
        • Goda K.
        • Carlo Di
        • Gelatin D.A
        Microdroplet platform for high-throughput sorting of hyperproducing single-cell-derived microalgal clones.
        Small. 2018; 141803315https://doi.org/10.1002/smll.201803315
        • Yanakieva D.
        • Elter A.
        • Bratsch J.
        • Friedrich K.
        • Becker S.
        • Kolmar H.
        FACS-based functional protein screening via microfluidic co-encapsulation of yeast secretor and mammalian reporter cells.
        Sci Rep. 2020; 10: 1-13https://doi.org/10.1038/s41598-020-66927-5
        • Fang Y.
        • Chu T.H.
        • Ackerman M.E.
        • Griswold K.E.
        Going native: direct high throughput screening of secreted full-length IgG antibodies against cell membrane proteins.
        MAbs. 2017; 9: 1253-1261https://doi.org/10.1080/19420862.2017.1381812
        • Eun Y.J.
        • Utada A.S.
        • Copeland M.F.
        • Takeuchi S.
        • Weibel D.B.
        Encapsulating bacteria in agarose microparticles using microfluidics for high-throughput cell analysis and isolation.
        ACS Chem Biol. 2011; 6: 260-266https://doi.org/10.1021/cb100336p
        • Brower K.K.
        • Khariton M.
        • Suzuki P.H.
        • Still C.
        • Kim G.
        • Calhoun S.G.K.
        • Qi L.S.
        • Wang B.
        • Fordyce P.M.
        Double emulsion picoreactors for high-throughput single-cell encapsulation and phenotyping via FACS.
        Anal Chem. 2020; 92: 13262-13270https://doi.org/10.1021/acs.analchem.0c02499
        • Zinchenko A.
        • Devenish S.R.A.
        • Kintses B.
        • Colin P.Y.
        • Fischlechner M.
        • Hollfelder F.
        One in a million: flow cytometric sorting of single cell-lysate assays in monodisperse picolitre double emulsion droplets for directed evolution.
        Anal Chem. 2014; 86: 2526-2533https://doi.org/10.1021/ac403585p
        • Terekhov S.S.
        • Smirnov I.V.
        • Stepanova A.V.
        • Bobik T.V.
        • Mokrushina Y.A.
        • Ponomarenko N.A.
        • Belogurov A.A.
        • Rubtsova M.P.
        • Kartseva O.V.
        • Gomzikova M.O.
        • Moskovtsev A.A.
        • Bukatin A.S.
        • Dubina M.V.
        • Kostryukova E.S.
        • Babenko V.V.
        • Vakhitova M.T.
        • Manolov A.I.
        • Malakhova M.V.
        • Kornienko M.A.
        • Tyakht A.V.
        • Vanyushkina A.A.
        • Ilina E.N.
        • Masson P.
        • Gabibov A.G.
        • Altman S.
        Microfluidic droplet platform for ultrahigh-throughput single-cell screening of biodiversity.
        Proc Natl Acad Sci U S A. 2017; 114: 2550-2555https://doi.org/10.1073/pnas.1621226114
        • Zee M.
        • Rutte J.
        • Rumyan R.
        • Williamson C.
        • Burnes T.
        • Radakovits R.
        • Eugenio A.S.
        • Badih S.
        • Lee D.H.
        • Archang M.
        • Carlo D.Di
        High-throughput selection of microalgae based on biomass accumulation rates in production environments using picoshell particles.
        bioRxiv. 2021; (2021.02.03.429271)https://doi.org/10.1101/2021.02.03.429271
        • Wu C.Y.
        • Ouyang M.
        • Wang B.
        • de Rutte J.
        • Joo A.
        • Jacobs M.
        • Ha K.
        • Bertozzi A.L.
        • Di Carlo D.
        Monodisperse drops templated by 3D-structured microparticles.
        Sci Adv. 2020; 6: eabb9023https://doi.org/10.1126/sciadv.abb9023
        • Destgeer G.
        • Ouyang M.
        • Wu C.Y.
        • Di Carlo D.
        Fabrication of 3D concentric amphiphilic microparticles to form uniform nanoliter reaction volumes for amplified affinity assays.
        Lab Chip. 2020; 20: 3503-3514https://doi.org/10.1039/d0lc00698j
        • de Rutte J.
        • Dimatteo R.
        • van Zee M.
        • Damoiseaux R.
        • Di Carlo D.
        Massively parallel encapsulation of single cells with structured microparticles and secretion-based flow sorting.
        bioRxiv. 2020; (2020.03.09.984245)https://doi.org/10.1101/2020.03.09.984245
        • Lee S.
        • Rutte J.
        • Dimatteo R.
        • Koo D.
        • Carlo D.Di
        Scalable fabrication of 3D structured microparticles using induced phase separation.
        bioRxiv. 2021; (2021.07.14.451688)https://doi.org/10.1101/2021.07.14.451688
        • Ha K.
        • de Rutte J.
        • Di Carlo D.
        • Bertozzi A.
        Minimal surface configurations for axisymmetric microparticles.
        UCLA Comput Appl Math Rep. 2020; (No. 20–38)
        • Sklar L.A.
        • Carter M.B.
        • Edwards B.S.
        Flow Cytometry for drug discovery, receptor pharmacology and high-throughput screening.
        Curr Opin Pharmacol. 2007; (Elsevier October 1): 527-534https://doi.org/10.1016/j.coph.2007.06.006
        • Zola H.
        High-sensitivity immunofluorescence/flow cytometry: detection of cytokine receptors and other low-abundance membrane molecules.
        Curr Protoc Cytom. 2004; (Chapter 6)https://doi.org/10.1002/0471142956.cy0603s30
        • Han Q.
        • Bradshaw E.M.
        • Nilsson B.
        • Hafler D.A.
        • Love J.C
        Multidimensional analysis of the frequencies and rates of cytokine secretion from single cells by quantitative microengraving.
        Lab Chip. 2010; 10: 1391-1400https://doi.org/10.1039/b926849a
      1. B. Alberts; A. Johnson; J. Lewis; M. Raff; K. Roberts; P. Walter B Cells and Antibodies. 2002.

        • Maciorowski Z.
        • Chattopadhyay P.K.
        • Jain P.
        Basic multicolor flow cytometry.
        Curr Protoc Immunol. 2017; 2017: 5.4.1-5.4.38https://doi.org/10.1002/cpim.26
        • Brower K.K.
        • Brower K.K.
        • Carswell-Crumpton C.
        • Klemm S.
        • Cruz B.
        • Kim G.
        • Calhoun S.G.K.
        • Nichols L.
        • Fordyce P.M.
        • Fordyce P.M.
        • Fordyce P.M.
        • Fordyce P.M.
        Double emulsion flow cytometry with high-throughput single droplet isolation and nucleic acid recovery.
        Lab Chip. 2020; 20: 2062-2074https://doi.org/10.1039/d0lc00261e
        • de Rutte J.M.
        • Koh J.
        • Di Carlo D.
        Scalable high-throughput production of modular microgels for in situ assembly of microporous tissue scaffolds.
        Adv Funct Mater. 2019; https://doi.org/10.1002/adfm.201900071
        • Nitta N.
        • Sugimura T.
        • Isozaki A.
        • Mikami H.
        • Hiraki K.
        • Sakuma S.
        • Iino T.
        • Arai F.
        • Endo T.
        • Fujiwaki Y.
        • Fukuzawa H.
        • Hase M.
        • Hayakawa T.
        • Hiramatsu K.
        • Hoshino Y.
        • Inaba M.
        • Ito T.
        • Karakawa H.
        • Kasai Y.
        • Koizumi K.
        • Lee S.W.
        • Lei C.
        • Li M.
        • Maeno T.
        • Matsusaka S.
        • Murakami D.
        • Nakagawa A.
        • Oguchi Y.
        • Oikawa M.
        • Ota T.
        • Shiba K.
        • Shintaku H.
        • Shirasaki Y.
        • Suga K.
        • Suzuki Y.
        • Suzuki N.
        • Tanaka Y.
        • Tezuka H.
        • Toyokawa C.
        • Yalikun Y.
        • Yamada M.
        • Yamagishi M.
        • Yamano T.
        • Yasumoto A.
        • Yatomi Y.
        • Yazawa M.
        • Di Carlo D.
        • Hosokawa Y.
        • Uemura S.
        • Ozeki Y.
        • Goda K.
        Intelligent image-activated cell sorting.
        Cell. 2018; 175 (e13): 266-276https://doi.org/10.1016/j.cell.2018.08.028
        • Isozaki A.
        • Mikami H.
        • Tezuka H.
        • Matsumura H.
        • Huang K.
        • Akamine M.
        • Hiramatsu K.
        • Iino T.
        • Ito T.
        • Karakawa H.
        • Kasai Y.
        • Li Y.
        • Nakagawa Y.
        • Ohnuki S.
        • Ota T.
        • Qian Y.
        • Sakuma S.
        • Sekiya T.
        • Shirasaki Y.
        • Suzuki N.
        • Tayyabi E.
        • Wakamiya T.
        • Xu M.
        • Yamagishi M.
        • Yan H.
        • Yu Q.
        • Yan S.
        • Yuan D.
        • Zhang W.
        • Zhao Y.
        • Arai F.
        • Campbell R.E.
        • Danelon C.
        • Di Carlo D.
        • Hiraki K.
        • Hoshino Y.
        • Hosokawa Y.
        • Inaba M.
        • Nakagawa A.
        • Ohya Y.
        • Oikawa M.
        • Uemura S.
        • Ozeki Y.
        • Sugimura T.
        • Nitta N.
        • Goda K.
        Intelligent image-activated cell sorting 2.0.
        Lab Chip. 2020; 20: 2263-2273https://doi.org/10.1039/d0lc00080a
        • Nitta N.
        • Iino T.
        • Isozaki A.
        • Yamagishi M.
        • Kitahama Y.
        • Sakuma S.
        • Suzuki Y.
        • Tezuka H.
        • Oikawa M.
        • Arai F.
        • Asai T.
        • Deng D.
        • Fukuzawa H.
        • Hase M.
        • Hasunuma T.
        • Hayakawa T.
        • Hiraki K.
        • Hiramatsu K.
        • Hoshino Y.
        • Inaba M.
        • Inoue Y.
        • Ito T.
        • Kajikawa M.
        • Karakawa H.
        • Kasai Y.
        • Kato Y.
        • Kobayashi H.
        • Lei C.
        • Matsusaka S.
        • Mikami H.
        • Nakagawa A.
        • Numata K.
        • Ota T.
        • Sekiya T.
        • Shiba K.
        • Shirasaki Y.
        • Suzuki N.
        • Tanaka S.
        • Ueno S.
        • Watarai H.
        • Yamano T.
        • Yazawa M.
        • Yonamine Y.
        • Di Carlo D.
        • Hosokawa Y.
        • Uemura S.
        • Sugimura T.
        • Ozeki Y.
        • Goda K.
        Raman image-activated cell sorting.
        Nat Commun. 2020; 11https://doi.org/10.1038/s41467-020-17285-3