Massively Parallel Profiling of Single‐Cell RNA Dynamics Using Well‐TEMP‐seq

Abstract

Single‐cell RNA sequencing (scRNA‐seq) reveals the transcriptional heterogeneity of cells, revolutionizing our understanding of cellular processes. However, the static snapshots obtained from scRNA‐seq fail to reveal the time‐resolved dynamics of transcription, which impedes critical insights into various biological processes, such as cellular differentiation, embryonic development, disease progression, and responses to external stimuli. Here, we describe Well‐TEMP‐seq, a protocol for massively parallel profiling of the temporal dynamics of single‐cell gene expression. Well‐TEMP‐seq combines metabolic RNA labeling with a microwell‐based scRNA‐seq method, Well‐paired‐seq, to distinguish newly transcribed RNAs marked by T‐to‐C substitutions from pre‐existing RNAs in each of thousands of single cells. Well‐TEMP‐seq is high‐throughput, cost‐effective, accurate, and provides a low cell loss rate and high single cell/bead pairing efficiency. More importantly, Well‐TEMP‐seq can be easily set up in other labs, and the loading of cells and beads can be easily accomplished by an optical microscope and a pipette. We believe that Well‐TEMP‐seq will be widely adopted and help researchers perform transformative research to unveil the dynamics of single‐cell gene expression in diverse biological processes. © 2026 Wiley Periodicals LLC.

Basic Protocol 1: Well‐paired‐seq chip fabrication

Basic Protocol 2: Well‐TEMP‐seq sample processing

Basic Protocol 3: Bioinformatics analysis

INTRODUCTION

Gene expression of cells is a heterogeneous and temporally dynamic program in various biological processes, such as cell differentiation, embryo development, disease progression, and response to stimuli (de Nadal et al., ²⁰¹¹; Emilsson et al., ²⁰⁰⁸; Hamatani et al., ²⁰⁰⁴; Saitou et al., ²⁰⁰²; Toyooka et al., ²⁰⁰⁸). Single‐cell RNA sequencing (scRNA‐seq) reveals the transcriptional heterogeneity of cells and is transforming our understanding of biology (Chen et al., 2021). However, scRNA‐seq captures static snapshots of gene expression and fails to temporally resolve the RNA dynamics (Lin et al., 2021). Metabolic RNA labeling‐based scRNA‐seq methods, such as scSLAM‐seq, NASC‐seq, scEU‐seq, sci‐fate, and scNT‐seq, have emerged as promising solutions to unveil the temporal dynamics of single‐cell gene expression by combining nucleoside analog (e.g., 4‐thiouridine, 4sU, and 5‐ethynyl uridine, 5‐EU) labeling and scRNA‐seq workflow (Battich et al., 2020; Cao et al., ²⁰²⁰; Erhard et al., ²⁰¹⁹; Hendriks et al., ²⁰¹⁹; Qiu et al., ²⁰²⁰). Plate‐based scSLAM‐seq, NASC‐seq, and scEU‐seq show high gene detection sensitivity but they are low‐throughput, costly, time‐consuming, and labor‐intensive. These methods are more suitable for the study of dozens to hundreds of cells, such as the early stage of embryo development. High‐throughput methods, such as sci‐fate and scNT‐seq, can handle thousands of cells in one experiment, which is beneficial for the study of heterogenous cell population and cell state transition. However, these high‐throughput methods suffer from a high cell‐loss rate, or low barcoding efficiency, which hamper the accurate capture and deeper understanding of temporal dynamics in single cells and the spread of metabolic labeling‐based scRNA‐seq technology to advance new biological applications (Erhard et al., 2022).

To meet these challenges, we developed Well‐TEMP‐seq, which combines metabolic RNA labeling and Well‐paired‐seq chip‐based scRNA‐seq for massively parallel characterization of the temporal dynamics of gene expression in thousands of single cells (Fig. 1) (Lin et al., 2023). Well‐paired‐seq is a high‐throughput scRNA‐seq platform (Yin et al., 2022). The Well‐paired‐seq chip consists of thousands of dual wells for single cells/barcoded beads trapping and pairing. Under the principles of size‐exclusion and locally quasi‐static hydrodynamics, single cells are trapped in the bottom wells, and single beads are trapped in the top wells. It breaks the limitations of the Poisson distribution, achieving high throughput (up to 100,000 cells per assay), high efficiency of single cell/bead pairing (∼80%), and low collision rate (0% to 6.2%). It also allows cell‐free RNA removal for highly accurate single‐cell gene expression profiling. After metabolic RNA labeling of 4sU, cells are subjected to single cell/barcoded bead pairing, cell lysing, RNA capture, and chemical conversion. Notably, the iodoacetamide (IAA)‐based chemical conversion recodes the base's hydrogen‐bonding pattern and transforms 4sU to a cytosine analog by nucleophilic substitution, resulting in T‐to‐C substitutions at 4sU‐labeled sites. Newly transcribed RNAs marked by T‐to‐C substitutions can be accurately distinguished from pre‐existing RNAs after next‐generation sequencing. Therefore, Well‐TEMP‐seq unbiasedly captures the RNA dynamics.

Figure 1.

Overview of the working principle of Well‐TEMP‐seq. Reproduced from Lin et al. (2023). Copyright 2023 Springer Nature.

Well‐TEMP‐seq is superior to other metabolic RNA labeling‐based scRNA‐seq methods in terms of throughput, cost, accuracy, and efficiency. Well‐TEMP‐seq and other unique molecular identifier (UMI)‐based approaches require a lower sequencing depth than full‐length transcript‐based approaches. Well‐TEMP‐seq costs <$0.10 per cell for library preparation, which is even lower than other UMI‐based approaches. Due to the adoption of Well‐paired‐seq chip and the improved alkylation chemistry on barcoded beads, a high single cell/barcoded bead pairing rate (∼80% vs ∼1% in scNT‐seq) and a low chemical conversion‐induced cell loss (∼67.5% recovery vs ∼5% recovery in sci‐fate) have been achieved. The Well‐TEMP‐seq protocol is compatible with fixed cells to perform temporal scRNA‐seq. It allows researchers to collect samples at different time points and parallel process the fixed cells of different groups on the same day. It is noteworthy to mention that Well‐TEMP‐seq can handle up to 8 parallel samples in one chip, which is beneficial for reducing batch effects. Moreover, the whole process of Well‐TEMP‐seq can be accomplished with a pipette, an optical microscope, and a thermocycler, which makes this technique readily affordable and reproducible in a standard lab.

In this article, we detail the working protocol of the previously published Well‐TEMP‐seq (Lin et al., 2023), step‐by‐step without any major changes, aiming to help researchers in the field to apply this technology to diverse biological processes. The original Well‐TEMP‐seq protocol is divided into three Basic Protocols, including Well‐paired‐seq chip fabrication for single cell/single bead pairing, Well‐TEMP‐seq sample processing for sequencing library preparation, and bioinformatics analysis for biological studies. The dual‐well chip is fabricated by soft lithography. Mask fabrication with twice overlay exposure produces the two‐layer pillars. The polydimethylsiloxane (PDMS)‐based Well‐paired‐seq chip is then obtained via a replication molding process. The Well‐TEMP‐seq sample processing includes metabolic RNA labeling, cell fixation, single cell/bead pairing, cell lysing, RNA capture, IAA treatment, scRNA‐seq library preparation, and next‐generation sequencing. The Well‐TEMP‐seq bioinformatic analysis includes read alignment and quantification and estimation of the portion of newly transcribed transcripts. This article is expected to make Well‐TEMP‐seq accessible to scientists in the related fields of biological, chemical, and clinical sciences and help address fundamental questions in biology and biomedicine.

Basic Protocol 1. WELL‐PAIRED‐SEQ CHIP FABRICATION

This protocol describes the procedure to fabricate the microwell‐based Well‐paired‐seq chip for single cells and single barcoded beads trapping and pairing (Fig. 2). The Well‐paired‐seq chip consists of thousands of dual wells and is obtained via soft lithography. The obtained Well‐paired‐seq chip can handle up to 8 parallel samples in one chip, which greatly increases the throughput and minimizes potential batch effects.

Figure 2.

Overview of the experimental procedure of Well‐paired‐seq chip fabrication.

Materials

• SU‐8 3050 photoresist (MicroChem, cat. no. SU‐8 3050)

• SU‐8 2015 photoresist (MicroChem, cat. no. SU‐8 2015)

• 1‐Methoxy‐2‐propyl acetate (Aladdin Co., cat. no. P299435‐500ml)

• Isopropanol (Sinopharm Chemical Reagent, cat. no. 40064360)

• DEPC H₂O (Sangon Biotech, cat. no. B501005‐0500)

• ¹H,¹H,²H,²H‐perfluorooctyldimethyl‐chlorosilane (Alfa Aesar, cat. no. 044543)

• HFE‐7500 (3M, cat. no. FS001)

• PDMS precursor and curing agent (DowCorning, cat. no. DC184)

• Bovine serum albumin (BSA) (New England Biolabs, cat. no. B9000S)

• Oven (Shanghai Yiheng, cat. no. PH‐010)

• Silicon wafer (CChip Scientific Instrument Co., cat. no. ZX32941)

• Source of N₂ gas (Linde Industrial Gases, cat. no. 5112‐1‐KS)

• Plasma cleaner (Harrick Plasma, cat. no. PDC‐002‐HP)

• Spin coater (Wenhao, cat. no. KW‐4A)

• UV light source (SUSS MicroTec Lithography, cat. no. MA8BA8)

• Mask (Qingyi Photomask Limited, Chromium plate)

• Optical microscope (Caikang Optics, cat. no. XDS‐200C)

• 100‐mm‐diameter Petri dish

• 200‐µl pipette (Eppendorf, cat. no. 3123000250)

• 1In an oven, heat silicon wafer at 135°C to remove moisture on the surface and blow the surface with pressurized N₂ gas.

  Any moisture on the surface of silicon wafer affects the coating of photoresist. Pressurized N₂ gas can be obtained from N₂ gas cylinder or piped N₂ gas supply. To ensure a clean silicon wafer surface, use a rubber tube (inner diameter ∼5 mm) connecting to the N₂ gas cylinder or piped N₂ gas supply and carefully blow the surface to remove any dust.

• 2Treat the silicon wafer with O₂ plasma for 5 min and blow the surface with N₂ gas to obtain a clean surface.

  The chip fabrication should be performed in the cleanroom. The silicon wafer can be placed in the chamber of any plasma cleaner that equipped with pure O₂ gas. The chamber should be vacuumed first and filled with O₂ at a pressure of 600 bar.

• 3Spin coat SU‐8 3050 photoresist onto the surface of silicon wafer (2900 rpm for 35 s) to a height of 50 µm and bake the photoresist (65°C for 5 min, and then 105°C for 15 min).

  Set up the spin coating parameters before dispensing ∼3 ml of SU‐8 3050 onto the center of the wafer.

• 4Perform UV exposure under a bead‐layer mask to produce the bead‐layer pattern and bake the photoresist (65°C for 5 min, and then 105°C for 15 min).

  This step is intended to generate a bead‐layer pattern by selectively exposing the photoresist to UV light.

• 5Coat SU‐8 2015 photoresist onto the surface of silicon wafer (500 rpm for 10 s, and then 1400 rpm for 30 s) to a height of 25 µm and bake the photoresist (65°C for 5 min, and then 105°C for 15 min).

  Prepare two hot plates with temperature set to 65°C and 105°C, respectively. Therefore, the silicon wafer can be transferred from 65°C plate and 105°C plate without delay.

• 6Perform UV exposure under a cell‐layer mask to produce the cell‐layer pattern and bake the photoresist (65°C for 5 min, and then 105°C for 15 min).

  Keep the bead‐layer pattern and the cell‐layer pattern center‐aligned under an optical microscope for producing perfect two‐layer pillars.

• 7Develop the two‐layer pillars with 1‐methoxy‐2‐propyl acetate, clean with isopropanol and H₂O successively, and heat at 135°C for 2 hr for hard baking.

  Use fresh 1‐methoxy‐2‐propyl acetate to enhance the reaction efficiency and ensure the silicon wafer with photoresist is immersed in the solution.

• 8Prepare a 1% ¹H,¹H,²H,²H‐perfluorooctyldimethyl‐chlorosilane/HFE‐7500 (v/v) solution and dispense ∼0.1 ml of the solution onto the center of the wafer before spin coating (1500 rpm for 60 s) to make the surface hydrophobic.

  The hydrophobic coating of the silicon mold surface is important for successful reverse replication by PDMS.

• 9Place the completed master wafer in a plastic 100‐mm‐diameter Petri dish. Pour polydimethylsiloxane (PDMS) precursor solution (10:1 of PDMS precursor and curing agent) onto the mold and cure at 100°C for 15 min to solidify the elastomer.

  Mix PDMS precursor and curing agent thoroughly and avoid visible bubbles.

• 10Bond the patterned PDMS to a grooved PDMS containing an inlet and an outlet after plasma activation to obtain the Well‐paired‐seq chip. The height, width, and length of the flow channel are 500 µm, 0.5 cm, and 1.2 cm, respectively.

  During plasma activation, the PDMS slab should be placed in the plasma chamber with the channel side facing upward. The structure of the patterned PDMS can be examined by microscopy.

• 11Inject 200 µl of 0.5% BSA aqueous solution into the chip and dry at 60°C overnight to block the surface.

  The channels should be filled with BSA solution and bubbles in the channels should be avoided. BSA coating significantly reduces the non‐specific binding of cell‐free nucleic acids, proteins, and cells on the surface of the chip.

Basic Protocol 2. WELL‐TEMP‐SEQ SAMPLE PROCESSING

This protocol provides steps to a standard Well‐TEMP‐seq library for sequencing (Fig. 3). The Well‐TEMP‐seq sample processing relies on the following key steps:

• (1)Incubation of cells with 4‐thiouridine (4sU), the biocompatible thymidine analog, to label newly transcribed RNAs.
• (2)Cells and barcoded microbeads are successively loaded into the microwells to achieve single‐cell/bead pairing. Cell‐free RNAs can be removed by washing before loading microbeads.
• (3)Cells are lysed and RNAs with poly(A) tails are captured by the microbeads with oligo(dT) primers.
• (4)Microbeads are pooled and subjected to a one‐pot chemical reaction with iodoacetamide (IAA) to recode base's hydrogen‐bonding pattern and transform 4sU to a cytosine analog by nucleophilic substitution, resulting in U‐to‐C substitutions at 4sU labeled sites.
• (5)RNAs on the microbeads are reverse transcribed, and the resulting cDNAs are amplified by PCR and tagmented by Tn5 transposase for library preparation and sequencing.

Figure 3.

Overview of the experimental procedure of Well‐TEMP‐seq sample processing.

Materials

• Cells, e.g., Human K562 cells (ATCC, CCL‐243) and human colorectal cancer HCT116 cells (ATCC, CCL‐247)

• 5‐AZA‐CdR (Sigma‐Aldrich, cat. no. 189825‐25MGCN)

• 4‐Thiouridine (4sU) (Aladdin Limited, cat. no. T122953)

• 0.25% trypsin‐EDTA (Gibco, cat. no. 25200056)

• Dulbecco's phosphate‐buffered saline (DPBS) (Sangon Biotech, cat. no. E607009‐0500)

• RNase inhibitor (Thermo Fisher, cat. no. EO0381)

• Methanol (Sinopharm Chemical Reagent, cat. no. 10014108)

• Enzyme blocking buffer (see recipe)

• BSA (New England Biolabs, cat. no. B9000S)

• DNA barcoded beads [Macosko‐2011‐10(V+), cat. no. Chemgenes]

• Pre‐lysis buffer (see recipe)

• Sodium lauroyl sarcosine (Sangon Biotech, cat. no. A600486‐0250)

• Mineral oil (Sigma‐Aldrich, cat. no. 69794)

• Saline sodium citrate (Invitrogen, cat. no. AM9763)

• DEPC H₂O (Sangon Biotech, cat. no. B501005‐0500)

• Alkylation master mix (see recipe)

• Dithiothreitol (DTT) (Sangon Biotech, cat. no. B645939‐0001)

• Maxima H Minus reverse transcriptase (Thermo Fisher, cat. no. EP0751)

• Maxima RT master mix (see recipe)

• Tris·HCl pH 8.0 (Leagene Biotechnology, cat. no. NR0073)

• EDTA (Leagene Biotechnology, cat. no. NR0009)

• Sodium dodecyl sulfate (SDS) (Sinopharm Chemical Reagent, cat. no. 30166428)

• Tween 20 (Sigma‐Aldrich, cat. no. P9416)

• Exonuclease I (New England Biolabs, cat. no. M0293L)

• KAPA HiFi HotStart ReadyMix (Kapa Biosystems, cat. no. KK2601)

• ISPCR oligo primer (5’‐AAGCAGTGGTATCAACGCAGAGT‐3’)

• VAHTS DNA Clean Beads (Vazyme Biotech, cat. no. N411‐01)

• Qubit dsDNA HS Assay Kit (Thermo Fisher, cat. no. Q32851)

• TruePrep DNA Library Prep Kit V2 (Vazyme Biotech, cat. no. TD502/TD503)

• P5‐TSO‐hybrid primer (5’‐AATGATACGGCGACCACCGAGATCTACACGCCTGTCCGCGGAAGCAGTGGTATCAACGCAGAGT*A*C‐3’; * indicates phosphorothioate modification)

• Read 1 primer (5’‐GCCTGTCCGCGGAAGCAGTGGTATCAACGCAGAGTAC‐3’)

• 0.5‐ and 1.5‐ml tubes

• Refrigerated centrifuge (Eppendorf, cat. no. 5425R), 4°C

• Vortex mixer

• Well‐paired‐seq chip (from Basic Protocol 1)

• Ultrasonic cleaner (Supmile, cat. no. KQ2200DV)

• 10‐, 20‐, 200‐, and 1000‐µl pipettes and tips (Eppendorf, cat. nos. 3123000225, 3123000233, 3123000250, and 3123000268)

• Decolorization shaker (Qilinbeier, cat. no. TS‐100)

• Optical microscope (Caikon, cat. no. XDS‐200C)

• Rotator (AS ONE, cat. no. ACR‐100)

• Thermocycler (Bio‐gener, cat. no. GE4T)

• Bioanalyzer (Bioptic, Qsep‐100)

• 1Treat cells with drugs (e.g., 5‐AZA‐CdR) or other stimuli for different durations (optional). Incubate cells with 200 µM 4sU for 2 hr before harvest.

  Disperse 4sU in a new cell culture medium and replace the old cell culture medium to start labeling.

• 2Digest cells with 0.25% trypsin‐EDTA and harvest cells in 1.5‐ml tube.

  Digest cells into single cells. Insufficient digestion may lead to cell clusters while over digestion may lead to low cell viability.

• 3Centrifuge cells 5 min at 300 × g, 4°C, and wash once with 1× DPBS.
• 4Re‐disperse cells in 1 volume (100 µl) of ice‐cold DPBS (with 0.4 U/µl RNase inhibitor) and dropwise add 9 volumes (900 µl) of ice‐cold methanol (pre‐chilled to −20 °C) with gentle vortexing.
• 5Fix cells for 10 min on ice in the dark and centrifuge 3 min at 900 × g, 4°C.

  4sU is sensitive to light. Keep the sample in the dark and avoid potential RNA degradation.

• 6Wash (not re‐suspended) cell pellet with ice‐cold DPBS (with 0.4 U/µl RNase inhibitor) and re‐suspended in 100 µl enzyme blocking buffer (saturated ammonium sulfate solution with 50 mM EDTA, 0.8 U/µl RNase inhibitor, pH 5.2). Store cells at −20 °C in the dark.

  Do not re‐suspend the cells during ice‐cold DPBS washing. Fixed cells in DPBS are difficult to be pelleted by centrifuge.

• 7Pipet 70 µl of 1× DPBS into the Well‐paired‐seq chip and sonicate it for 2 min to remove air bubbles in wells.

  Immerse the chip partially in the water bath of the ultrasonic cleaner.

• 8Centrifuge fixed cells 5 min at 900 × g, 4°C, and re‐suspend cells in 0.05% BSA solution (in 1× DPBS) to make single‐cell suspension.

  0.05% BSA solution help disperse single cells. Higher concentrations of BSA may generate bubbles and affect the single‐cell loading process.

• 9Aspirate the 1× DPBS from the Well‐paired‐seq chip and then pipet 70 µl of single‐cell suspension into the chip.
• 10Oscillate the chip horizontally for 5 min to capture single cells in the bottom wells of the chip by gravitational settling.
• 11Aspirate the single‐cell suspension in the chip and wash with 1× DPBS three times.
• 12Check the cell occupations in the chip by the optical microscope. If the cell capture efficiency is too low, repeat steps 10 to 11 as necessary.

  We recommend a minimum cell occupancy of 60% for a good practice of the Well‐TEMP‐seq protocol.

• 13Re‐suspend DNA barcoded beads in pre‐lysis buffer (200 mM Tris·HCl, 20 mM EDTA, 50 mM DTT, pH 7.5). Aliquot ∼50,000 beads into a 0.5‐ml tube, centrifuge 30 s at 300 × g, room temperature, wash twice with pre‐lysis buffer, and re‐suspend beads in pre‐lysis buffer.

  During washing, carefully remove the supernatant and avoid touching the beads by the tip of pipette.

• 14Aspirate the buffer in the chip and load 70 µl bead suspension.

  Beads easily precipitate at the bottom of the centrifuge tube by gravity. Pipet up and down for several times before taking 70 µl bead suspension.

• 15Oscillate the chip horizontally for 2 min to capture single beads in the top wells of the chip by gravitational settling.
• 16Aspirate the bead suspension in the chip and wash the chip twice with pre‐lysis buffer.
• 17Check the bead occupations in the chip by the optical microscope. If the bead occupancy rate is <99%, repeat steps 14 to 16 as necessary.
• 18Disperse 50 mg sodium lauroyl sarcosine in 1 ml mineral oil by ultrasonication.

  Sodium lauroyl sarcosine is a strong surfactant for cell lysis. Sodium lauroyl sarcosine can be easily dissolved in water rather than mineral oil. Therefore, the surfactant forms aggregates in the oil phase, which is highlighted in Figure 1.

• 19Aspirate the buffer in the chip and pipet 70 µl of the mineral oil containing sodium lauroyl sarcosine into the chip. Wait 15 min for cell lysis and mRNA capture.

  Mineral oil isolates each of dual‐layer microwells, resulting in thousands of aqueous droplets in the chip. Then, surfactant aggregates in the mineral oil settle down due to gravity and dissolve in the water phase of the aqueous droplets. This unique design is critical for single‐cell lysis in the protocol and avoid RNA diffusion‐induced cross‐well contamination.

• 20Aspirate the mineral oil in the chip and cut out the PDMS above the microwells.

  After cutting, a well is generated at the center of the PDMS for bead retrieval.

• 21Wash out beads from the chip by 6× saline sodium citrate (SSC) buffer and transfer beads to a 1.5‐ml tube.

  The 6× SSC buffer is diluted with DEPC H₂O from a commercial 20× SSC stock.

• 22Wash beads twice with 200 µl of 6× SSC, and then once with 200 µl alkylation buffer (50 mM phosphate buffer saline, pH 8.0).
• 23Prepare alkylation master mix by adding 10 µl of 2‐iodoacetamide (IAA, in DMSO) and 90 µl of alkylation buffer (see Reagents and Solutions).

  The final concentration of IAA and DMSO is 10 mM and 10% (v/v), respectively. The optimal pH for alkylation is 8.0.

• 24Incubate beads with 100 µl of alkylation master mix at 37°C for 1 hr.
• 25Add 1 µl of 1 M DTT to stop the alkylation reaction and wash beads with 1× DPBS and 1× RT buffer successively.

  DTT reacts with excessive IAA and quenches the alkylation reaction.

• 26Prepare Maxima RT master mix by adding 4 µl of Maxima 5× RT buffer, 2 µl of 10 mM dNTPs, 0.5 µl of RNase inhibitor, 1 µl of 50 µM template switching oligo (TSO), 1 µl of 200 U/µl Maxima H Minus reverse transcriptase, and 11.5 µl DEPC H₂O.

  The sequence of TSO is 5’‐AAGCAGTGGTATCAACGCAGAGTGAATrGrGrG‐3’.

• 27Re‐suspend beads in 20 µl Maxima RT master mix, incubate at room temperature for 30 min with rotation, and heat at 42°C for 90 min in thermocycler.
• 28Collect the beads by centrifuging for 30 s at 300 × g, room temperature, and wash once with 200 µl TE‐SDS (10 mM Tris·HCl, 1 mM EDTA, 0.5% sodium dodecyl sulfate, pH 8.0), once with 200 µl TE‐TW (10 mM Tris·HCl, 1 mM EDTA, 0.01% Tween 20, pH 8.0) and once with 200 µl Tris buffer (10 mM Tris·HCl, pH 8.0).

  Stop point: The microbeads after RT can be stored at 4°C for 1 day.

• 29Prepare the Exonuclease I master mix by adding 2 µl of 10× Exo I buffer, 17 µl of DEPC H₂O, and 1 µl of 20 U/µl Exonuclease I.

  Exonuclease I removes excessive poly(dT) primers on the surface of barcoded beads.

• 30Re‐suspend beads in 20 µl Exonuclease I master mix and incubate at 37°C for 45 min in the thermocycler.
• 31Collect the beads by centrifuging for 30 s at 300 × g, room temperature, and wash once with 200 µl TE‐SDS, once with 200 µl TE‐TW and once with 200 µl DEPC H₂O.
• 32Prepare PCR mix by adding 25 µl of 2× KAPA HiFi HotStart ReadyMix, 1 µl of 0.8 µM ISPCR oligo primer, 24 µl DEPC H₂O.

  The sequence of ISPCR oligo primer is 5’‐AAGCAGTGGTATCAACGCAGAGT‐3’.

• 33
  Re‐suspend beads in 50 µl PCR mix and incubate in the thermocycler with the thermal cycling parameters as follows:

  • 95°C for 3 min
  • 4 cycles of 98°C for 20 s, 65°C for 45 s, and 72°C for 3 min
  • 10 to 12 cycles of 98°C for 20 s, 67°C for 20 s, and 72°C for 3 min
  • 72°C for 5 min
  • Hold at 4°C.
• 34Centrifuge the tube for 30 s at 300 × g, room temperature, to separate beads and PCR product. Transfer the supernatant to a new a 0.5‐ml tube and purify the cDNAs twice with 0.6× VAHTS DNA Clean Beads according to the manufacturer's instructions.

  0.6× DNA clean beads remove undesired small fragments.

• 35Measure the concentration of the purified PCR product by Thermo Fisher Qubit 4.0 fluorometer.

  Stop point: The PCR product can be stored at −20° or −80°C for ≥6 months.

• 36
  Prepare the 3′‐end enriched sequencing library by TruePrep DNA Library Prep Kit V2 for Illumina according to the manufacturer's instructions, except that P5 primer is replaced by a customized P5‐TSO‐hybrid primer. The thermal cycling parameters are as follows:

  • 72°C for 3 min
  • 98°C for 30 s
  • 10 to 12 cycles of 98°C for 15 s, 60°C for 30 s, and 72°C for 3 min
  • 72°C for 5 min
  • Hold at 4°C.
    Replace standard P5 primer with customized P5‐TSO‐hybrid primer. The sequence of P5‐TSO‐hybrid primer is: 5’‐AATGATACGGCGACCACCGAGATCTACACGCCTGTCCGCGGAAGCAGTGGTATCAACGCAGAGT*A*C‐3’ (* indicates phosphorothioate modification).
• 37Centrifuge the tube 5 s at 300 × g, room temperature, transfer the PCR product to a new 0.5‐ml tube, and purify the library once with 0.6× VAHTS DNA Clean Beads according to the manufacturer's instructions.
• 38Measure the concentration of the purified library by Thermo Fisher Qubit 4.0 fluorometer.

  Stop point: The library product can be stored at −20° or −80°C for ≥6 months.

• 39Analyze the fragment size distribution by Bioptic Qsep‐100.
• 40Sequence the library with Illumina Hiseq X Ten (paired‐end, 150 bp). Read 1 primer is replaced by the customized Read 1 primer.

  The sequence of Read 1 primer is 5’‐GCCTGTCCGCGGAAGCAGTGGTATCAACGCAGAGTAC‐3’.

Basic Protocol 3. BIOINFORMATICS ANALYSIS

This protocol describes how to process sequencing data to generate new RNA and old RNA matrices for single‐cell temporal RNA dynamics analysis (Fig. 4). Read alignment is first performed, and each of the mapped reads is tagged with a cell barcode and a unique molecular identifier (UMI). Then, reads with T‐to‐C substitutions are identified as new RNA reads and are distinguished from old RNA reads. To address the insufficiency of metabolic RNA labeling, a binomial mixture model is applied to approximate the real distribution of T‐to‐C substitution in single cells and estimate the real mutation rate. After the statistical correction, the obtained new RNA and old RNA matrices are ready for downstream analysis. These contents are automatically carried out by the code provided in the GitHub pipeline.

Figure 4.

Overview of the bioinformatics analysis procedure of Well‐TEMP‐seq sequencing data.

Necessary Resources

Hardware

A computer with internet access and ≥16 GB memory

Software

Rstudio (version 4.0.0 or newer)

Files

• Well‐TEMP‐seq pipeline (https://github.com/songjiajia2018/Well‐TEMP‐Seq)

• Human reference genome (GRCh38) (https://asia.ensembl.org/index.htmlI)

• 1Extract the cell barcode (base 1‐12) and UMI (base 13‐20) in Read 1 and tag to the corresponding mRNA read (Read 2).
• 2Trim sequencing adaptors and poly(A) sequences from the reads and align the reads to the human reference genome assembly (GRCh38) using STAR v2.7.3a.
• 3Retain both exonic and intronic reads that are uniquely mapped to predicted strands of annotated genes with a mapping score >10.

  The original Drop‐seq pipeline only retains exonic reads. However, a large fraction of newly transcribed RNAs contains unspliced introns. Therefore, both exonic and intronic reads should be kept for downstream analysis.

• 4Identify the labeled reads with at least one T‐to‐C substitution (base Phred quality score >27).
• 5Count the total number of labeled and unlabeled RNA for each gene in each cell. Assemble the numbers into matrices using the gene name as rows and the cell barcode as columns. Each cell is associated with two digital gene expression matrices (labeled and unlabeled RNAs).
• 6Generate a consensus sequence for each transcript by gathering reads with the same UMI index and picking the most frequent variant at each site.

  Generating consensus sequence is intended to remove background mutations induced by PCR amplification or sequencing error.

• 7Approximate the real distribution of T‐to‐C substitution in single cells using the binomial mixture model.
• 8Calculate θ_gene, the level of new transcript for each gene at each condition according to the statistical model.
• 9Calculate α_cell, the mean detection rate in each cell, by dividing the number of all observed labeled transcripts by the number of all estimated new transcripts for each cell.
• 10Calculate N_gene, the estimated new RNA level for each gene in each cell.
• 11Generate the corrected digital gene expression matrices (new and old RNAs).

  Total RNA gene expression matrix can be generated by merging the new and old RNA matrices.

REAGENTS AND SOLUTIONS

Alkylation master mix

Alkylation master mix is used to perform alkylation on labeled 4sU sites of new RNAs and introduce T‐to‐C substitutions. Weigh 18.5 mg of 2‐iodoacetamide (IAA) powder (TCI, cat. no. I0044) and dissolve it in 1 ml dimethyl sulfoxide (DMSO) (Sangon Biotech, cat. no. A600163‐0250) to make 100 mM IAA. Mix 94.7 ml of 0.2 M Na₂HPO₄ (Sinopharm Chemical Reagent, cat. no. 10020318) and 5.3 ml of 0.2 M NaH₂PO₄ (Sinopharm Chemical Reagent, cat. no. 20040717) to make phosphate buffer. The pH of the obtained phosphate buffer is ∼8.0. Dilute 100 mM IAA to 10 mM IAA using pH 8.0 phosphate buffer. The obtained solution is denoted as alkylation master mix. Alkylation master mix should be freshly prepared before use.

Enzyme blocking buffer

Enzyme blocking buffer is used to precipitate protein after methanol fixation of cells and prevent RNA degradation by RNase in the cells. Weigh 8.0 g ammonium sulfate (Sinopharm Chemical Reagent, cat. no. 10002917) and dissolve it in 10 ml DEPC H₂O (Sangon Biotech, cat. no. B501005‐0500) while stirring at room temperature. Allow the solution to equilibrate for 20 min and filter the supernatant using a 0.22‐µm membrane to remove undissolved solid. In 9.55 ml saturated ammonium sulfate, add 250 µl of 2 M EDTA (Leagene Biotechnology, cat. no. NR0009) and 200 µl of 40 U/µl RNase inhibitor (Thermo Fisher, cat. no. EO0381) and mix thoroughly. The pH of the obtained enzyme blocking buffer is ∼5.2. Store up to 1 month at 4 °C.

Maxima RT master mix

• Maxima RT master mix is used for reverse transcription reaction. Mix the following well:

• 11.5 µl DEPC H₂O (Sangon Biotech, cat. no. B501005‐0500)

• 4 µl of Maxima 5× RT buffer (supplied with transcriptase; Thermo Fisher, cat. no. EP0751)

• 2 µl of 10 mM dNTPs (GenScript, cat. no. C01582‐250)

• 0.5 µl of 40 U/µl RNase inhibitor (Thermo Fisher, cat. no. EO0381)

• 1 µl of 50 µM template switching oligo (TSO) (5’‐AAGCAGTGGTATCAACGCAGAGTGAATrGrGrG‐3’)

• 1 µl of 200 U/µl Maxima H Minus reverse transcriptase (Thermo Fisher, cat. no. EP0751)

• Prepare fresh

Pre‐lysis buffer

Pre‐lysis buffer is used to remove TE‐TW in the stocking solution of DNA barcoded beads and to re‐suspend DNA barcoded beads before bead loading. Dilute 1 M Tris·HCl pH 7.5 stock solution (Leagene Biotechnology, cat. no. NR0072) to 200 mM Tris·HCl pH 7.5 solution. Add 100 µl of 2 M EDTA (Leagene Biotechnology, cat. no. NR0009) and 250 µl of 2 M dithiothreitol (DTT) (Sangon Biotech, cat. no. B645939‐0001) into 9.65 ml of 200 mM Tris·HCl pH 7.5 solution. Mix well to make 10 ml pre‐lysis buffer. Store up to 1 month at 4 °C.

COMMENTARY

Background Information

In the past decade, scRNA‐seq has been widely applied in revealing gene expression heterogeneity and uncovering new cell subtypes. However, scRNA‐seq only captures static snapshots of gene expression at specific time points, leaving the dynamics of single‐cell gene expression unexplored. To solve this problem, metabolic RNA labeling‐based time‐resolved scRNA‐seq technologies have been developed very recently by combining nucleoside analog labeling and scRNA‐seq. Among the developed time‐resolved scRNA‐seq methods, Well‐TEMP‐seq represents a well‐rounded method with high throughput, high sensitivity, and low cost to profile the single‐cell temporal RNA dynamics. Due to the time‐resolved scRNA‐seq ability, Well‐TEMP‐seq can detect the short‐term changes in gene expression, which is critical for gene regulations with rapid transcriptional responses. The applications of Well‐TEMP‐seq could be extended but not restricted to unveiling the heterogeneous state transitions of cells in embryogenesis, cellular differentiation, and response to external stimuli (e.g., drug treatment, virus infection, CRISPR perturbation). We anticipate that Well‐TEMP‐seq will be broadly applicable to addressing fundamental questions in biology and biomedicine.

Critical Parameters

Chip fabrication and pre‐processing

The Well‐paired‐seq chip fabrication process is sensitive to environmental dust. All the chip fabrication processes should be conducted in a cleanroom. The two‐layer pillar patterns should be examined under a microscope to ensure the size of microwells is as designed. The size of microwells is essential to achieve size‐exclusion and quasi‐static hydrodynamics principle for single cell capture. After chip fabrication, surface coating of the PDMS chip with a 0.5% BSA aqueous solution is also critical to block the surface and avoid non‐specific adsorption of cell‐free RNAs and cells in the chip.

Metabolic RNA labeling and cell fixation

In Well‐TEMP‐seq, 4sU is used for metabolic RNA labeling. 4sU contains a thiol group, which is critical for the alkylation reaction to introduce a T‐to‐C substitution. Therefore, 4sU should be protected from light throughout the experiment to avoid light‐induced hydrolysis of 4sU. As for the cell fixation, the methanol solution should be kept ice‐cold to avoid RNA degradation. Moreover, the fixed cells should be stored in an enzyme blocking buffer at –20°C and used as soon as possible. These experimental details contribute to an enhanced library complexity and accuracy of Well‐TEMP‐seq.

Cell capture efficiency

In the section of single cell capture, cells are loaded in the Well‐paired‐seq chip and captured in the bottom wells of the chip by gravitational settling. Due to the principle of size‐exclusion and quasi‐static hydrodynamics, only one cell is captured in each of the bottom wells. Excessive cells or cell clusters in the upper wells can be washed out. However, cells may not occupy every well in one round of cell loading. Therefore, the cell loading and washing should be repeated, ensuring that the capture cells occupy >60% of the wells. Fewer captured single cells may lead to less input RNA for the alkylation reaction, which may further decrease the RNA integrity and affect the library complexity of Well‐TEMP‐seq.

Troubleshooting

See Table 1 for a list of problems, possible causes, and solutions while executing the Basic Protocols.

Table 1.

Troubleshooting Guide for Well‐TEMP‐seq Protocols

Problem	Possible cause	Solution
Defects in the PDMS chip	Insufficient mixing of PDMS precursor and curing agent	Mix PDMS precursor and curing agent thoroughly and avoid air bubbles in the mixture before solidification
Cells aggregate together	Cell dissociation reagent is not working	Use a fresh trypsin‐EDTA solution or warm it to 37°C for 30 min before use
Low single‐cell capture rate	The concentration of single‐cell suspension is too low	Increase the concentration of single‐cell suspension or make sure the cell counting is accurate
Low cell recovery rate	DNA barcoded beads are lost during centrifugation and washing	Do not touch or disturb the beads when aspirating supernatant, and leave a small amount of supernatant in each round of washing
Low cDNA yield	The number of captured single cells is less than expected	Increase the occupancy of single cells in the microwells or increase the number of rounds of PCR
Low T‐to‐C substitution rate	RNAs are partially degraded, or IAA is not working well	After methanol fixation, quickly transfer cells into the enzyme blocking buffer and store at –20°C; prepare a fresh IAA working solution and make sure the pH is 8.0

Understanding Results

After Well‐paired‐seq chip fabrication, the obtained chips can be examined under a microscope. Figure 5 shows an example of Well‐paired‐seq chip characterization. The scanning electron microscopy (SEM) image (Fig. 5A) reveals bottom wells with a width of 25 µm and upper wells with an inscribed circle diameter of 50 µm. The side view of the microwells (Fig. 5B) further shows bottom wells with a height of 25 µm and upper wells with a height of 50 µm.

Figure 5.

Characterization of the Well‐paired‐seq chip. (A) The top‐view structure of the dual‐layer microwells imaged by scanning electron microscopy. (B) The side‐view structure of the dual wells by optical microscopy. Reproduced from Yin et al. (2022). Copyright 2022 Wiley‐VCH.

In the Well‐TEMP‐seq sample processing, the library is constructed via a 3’‐end method. The obtained cDNAs are tagmented by Tn5 transposase. We selectively amplify the sequences with one end of Customer Read 1 and the other end of Nextera Read 2. Therefore, we can extract cell barcode and UMI information from Read 1 reads and gene expression information from Read 2 reads. A typical size distribution of the sequencing library is shown in Figure 6. The expected size of the library is 350 to 1000 bp.

Figure 6.

A typical size distribution of the Well‐TEMP‐seq library for NGS sequencing.

For the bioinformatic analysis, the Well‐TEMP‐seq pipeline generates new RNA and old RNA matrices. The substitution rate, fraction of labeled transcripts per cell, and the fraction of labeled transcripts for each gene can be used to evaluate the performance of the Well‐TEMP‐seq protocol. For example, we label K562 cells with 200 µM 4sU for 2 hr and collect for Well‐TEMP‐seq sample processing. Only the group with 4sU labeling and IAA treatment exhibits high T‐to‐C substitution rates (Fig. 7A). Other groups without 4sU labeling or IAA treatment showed negligible substitution rates, indicating a high signal‐to‐noise ratio for Well‐TEMP‐seq. As shown in Figure 7B, 4sU labeling does not induce significant T‐to‐C substitution. Further IAA treatment recodes 4sU's hydrogen‐bonding pattern and induces T‐to‐C substitution. To examine the accuracy of Well‐TEMP‐seq in identifying newly transcribed RNAs, we can analyze the fraction of labeled transcripts per cell for genes encoding mRNAs with different turnover. In principle, mRNAs with high turnover should have a larger fraction of new RNAs, as indicated by a larger fraction of labeled transcripts with T‐to‐C substitutions. As shown in Figure 7C, transcription factor genes such as MYC encode mRNAs with high turnover, supported by the large fraction of labeled transcripts per cell. In contrast, housekeeping genes such as GAPDH encode mRNAs with low turnover, supported by the small fraction of labeled transcripts per cell.

Figure 7.

Evaluation of the Well‐TEMP‐seq performance. (A) Bar plot of nucleoside substitution rates in K562 cells (K562), 4sU‐labeled K562 cells (K562_4sU), IAA‐treated K562 cells (K562_IAA), and 4sU‐labeled and IAA‐treated K562 cells (K562_4sU_IAA). (B) Box plot of the fraction of labeled transcripts per cell in 4sU‐labeled K562 cells and 4sU‐labeled and IAA‐treated K562 cells. (C) Violin plots showing the fraction of labeled transcripts per cell of different genes in 4sU‐labeled K562 cells and 4sU‐labeled and IAA‐treated K562 cells. Left, high turnover gene (MYC); right, low turnover gene (GAPDH). Reproduced from Lin et al. (2023). Copyright 2023 Springer Nature.

Collectively, these results confirm the successful practice of the Well‐TEMP‐seq protocol.

Time Considerations

The construction of two‐layer pillars on a silicon wafer takes ∼6 hr. Further replication molding process takes <1 hr. The blocking process can be performed overnight. Therefore, the Well‐paired‐seq chip fabrication is done within 24 hr. We can prepare multiple Well‐paired‐seq chips in parallel and save time. Once the silicon wafer with two‐layer pillars is constructed, the mold can be used for a long time unless the patterns are broken.

The Well‐TEMP‐seq sample processing is critical for the successful practice of Well‐TEMP‐seq. The metabolic RNA labeling and fixation processes take ∼3 hr. The time spent on single cells and single beads loading depends on the proficiency of operation. We usually finish the cell loading, bead loading, cell lysis, bead recovery, and IAA treatment processes within 2 hr. Further reverse transcription and library construction are routine operations and take ∼6 hr. We recommend minimizing the time of operation before reverse transcription, as RNAs in the fixed cells tend to degrade after removing the enzyme blocking buffer.

Finally, the time needed for bioinformatic analysis depends on the accessible hardware, such as CPU and GPU. It is almost impossible to process the large sequencing raw data with a personal laptop. We recommend using a high‐performance computing server to run the Well‐TEMP‐seq pipeline. The digital gene expression matrix generation and correction can be finished within 2 days.

Author Contributions

Di Wang: Conceptualization; writing—original draft. Qiqi Lv: Writing—original draft. Shichao Lin: Conceptualization; funding acquisition; project administration; supervision; writing—review and editing.

Conflict of Interest

The authors declare no competing interests.

Wang, D. , Lv, Q. , & Lin, S. (2026). Massively parallel profiling of single‐Cell RNA dynamics using well‐TEMP‐seq. Current Protocols, 6, e70458. doi: 10.1002/cpz1.70458

Data Availability Statement

All sequencing data can be downloaded from Gene Expression Omnibus (GEO) with the accession code GSE194357. The human reference genome (GRCh38) used in this study can be downloaded from: https://asia.ensembl.org/index.html.