Overview

In iGEM 2025, we incorporated two measurement protocols for our project: Retro-Cascorder and keratinase activity assay. For Retro-Cascorder, we adopted an existing method to record transcriptional events as barcodes in the CRISPR array. Beyond the original protocol, we introduced several key improvements. First, we optimized critical parameters to minimize measurement bias by combining computational simulation and experimental validation. Second, inspired by the spirit of local people solving local problems, we made reagents available to the Chinese iGEM community, lowering the barrier for others to apply this method. As a result, we developed two versions of the protocol to expand its accessibility: one tailored for Chinese labs with limited access to certain reagents and another one standardized for international teams. Finally, we conducted a cost analysis that highlights the affordability of our approach, enabling wider adoption across iGEM teams.

For the keratinase activity assay, we designed a measurement strategy to address the complication caused by the insolubility of keratin by using a heterogeneous catalysis scheme based on the Folin-Ciocalteus assay. In addition, by reporting enzymatic activity in absolute units rather than relative values, we ensured that our results can be compared across labs and experiments with greater consistency. This emphasis on adaptability and reproducibility strengthens the robustness of our assay for future applications.

Retro-Casorder: A method to record transcriptional events

Introduction

Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) are adaptive immune systems found in bacteria and archaea that integrate short fragments of foreign DNA into their CRISPR arrays for future defense (Horvath & Barrangou, 2010). CRISPR-associated (Cas) proteins carry out the acquisition of these fragments and are widely applied in genome engineering and molecular recording (Askary et al., 2024); (Cong et al., 2013). Previous studies have combined Cas1–Cas2 with retrons, enabling retron-derived ncRNA transcribed from a target promoter to be reverse-transcribed into DNA and integrated as new spacers in the CRISPR array (Bhattarai-Kline et al., 2022). In this way, the order of transcriptional events can be permanently recorded into the bacterial genome.

Specifically, Retro-Cascorder harnesses the Cas1–Cas2 adaptation machinery together with a retron module to capture transcriptional events in real time (Fig. 1a). When a promoter of interest is activated, it drives transcription of a non-coding RNA (retron-ncRNA) containing specific barcodes, which is then reverse-transcribed by a constitutively expressed retron reverse transcriptase (retron-RT) into complementary DNA. These DNA fragments are subsequently integrated into the CRISPR array as new spacers. Besides retron-derived spacers, Cas1–Cas2 also incorporates spacers derived from other double-stranded DNA sources such as the host genome or plasmids. These non-retron-derived spacers, collectively referred to "N," accumulate alongside retron-derived spacers during the course of recording. Thus, each array can contain three types of spacers: retron-derived spacers corresponding to signal A or B, and background spacers denoted as N. Over time, the sequential acquisition of A, B, and N spacers forms a chronological record within the CRISPR array, which can be decoded by sequencing to reconstruct the order of transcriptional events.

To quantitatively assess recording outcomes, the ordering scores were defined based on the relative arrangement of spacers within CRISPR arrays (Eqn. 1-3). The A/N and B/N scores compare the positions of retron-derived spacers (A and B) relative to background spacers (N), which are assumed to be constitutively acquired throughout the recording period. If signal A was present earlier in the recording timeframe, then arrays of the form A → N → leader would be more abundant than N → A → leader, which gives positive A/N score. Hence, a positive A/N score therefore indicates that signal A was expressed earlier in the recording timeframe. Conversely, a negative A/N score suggests that A spacers appear later relative to N. For the B/N score, the logic is reversed: a positive value indicates that signal B spacers were acquired later in the recording period. The A/B score directly compares the relative positions of signal A and signal B spacers, providing a measure of the temporal order between the two retron signals. A positive A/B score indicates that A tends to be acquired earlier than B, while a negative score indicates the reverse. Together, these metrics offer a quantitative framework to interpret the relative timing of promoter activities and evaluate the temporal resolution and accuracy of Retro-Cascorder recordings.

Figure 1 | Principles of Retro-Cascorder transcriptional recording. a, Upon activation of a promoter of interest by signal A or B, the downstream retron non-coding RNA (RT-ncRNA) would be transcribed, which is reverse-transcribed by retron reverse transcriptase (RT, shown in blue) into complementary DNA. The Cas1–Cas2 integrase complex (yellow and brown) then integrates DNA sequentially into the CRISPR array as spacer A (green) or B (purple). Meanwhile, double-stranded DNA from other sources like genome or plasmids could also be incorporated by Cas1-Cas2 as spacer N (yellow). Over time, different signals are captured as new spacers, creating a chronological transcriptional history that can be read out by sequencing. b, The plasmid expressing Cas1, Cas2, and Eco1 retron-specific RT. The LacI-regulated T7 promoter drives the expression of Cas1 and Cas2, while Eco1 RT is constitutively expressed by the J23115 promoter. c, Retron ncRNA encoding barcode A and B were expressed by aTc- or Cho-inducible promoters regulated by TetR and BetI. Created by biorender.com.

Experimental workflow

The experimental workflow of Retro-Cascorder is outlined in Figure 2. The protocol begins with the transformation of the ncRNA plasmid (pSBK.134) into BL21(AI) chemically competent cells. In the next step, transformants were used to prepare electrocompetent cells, which were then electroporated with the plasmid carrying the retron reverse transcriptase (RT), Cas1, and Cas2 (pSBK.079). Transformants were recovered on LB agar plates containing dual antibiotics, and single colonies harboring both plasmids were selected for recording experiments. Each colony represented a biological replicate and was inoculated into LB medium for overnight culture.

The following day, overnight cultures were diluted into fresh LB supplemented with IPTG and L-arabinose to induce Cas1–Cas2 expression. Meanwhile, the first set of inducers, such as anhydrotetracycline (aTc) or choline chloride (Cho), was added to initiate recording. Genetic recording was carried out by serial dilution of cultures to maintain exponential growth. After 16 h, cultures were diluted into fresh LB with the same inducer composition to sustain log-phase growth while continuing recording. At 24 h, a second dilution was performed into LB containing the second set of inducers to activate an additional promoter. Cultures were further diluted at regular intervals to preserve exponential growth until a total of 48 h was reached.

At the end of the recording period, cells from each culture were harvested and boiled to release genomic DNA. Next-generation sequencing (NGS) libraries were then prepared using a two-step PCR: the first round amplified CRISPR arrays from the BL21(AI) genome, and the second round introduced Illumina adapters. The resulting library products were confirmed to be 286 bp by electrophoresis. NGS was performed on an Illumina NovaSeq X Plus platform generating paired-end 150 bp (PE150) reads. Finally, custom Python scripts were employed to process FASTQ files, extract newly acquired spacers, identify retron-derived barcodes, and compute ordering scores (A/B, A/N, B/N), enabling reconstruction of the temporal order of promoter activities.

Figure 2 | Experiment workflow of Retro-Cascorder. The protocol begins with chemical transformation of BL21(AI) cells with the ncRNA plasmid (pSBK.134), followed by electroporation of the expression plasmid (pSBK.079) carrying retron RT and Cas1–Cas2 (Steps 1–2). Colonies containing both plasmids were picked as biological replicates and grown in LB medium (Step 3). Cultures were induced with IPTG and L-arabinose to express Cas1–Cas2, and sequential exposure to inducers such as aTc or Cho initiated temporal recording (Step 4). After 48 hours of recording, cells were harvested and boiled. CRISPR arrays were amplified using a two-step PCR, purified, and validated by electrophoresis prior to sequencing (Step 5). Sequencing reads were processed to extract spacers, identify retron-derived barcodes, and reconstruct the temporal order of promoter activities (Step 6). Created by biorender.com.

Parameter optimization

PCR bias simulation

In the experimental workflow shown above, NGS library preparation requires two rounds of PCR. In the first round, primers anneal to the CRISPR array in the BL21(AI) genome to amplify the spacer region while appending unique molecular identifier and Illumina read primers; in the second round, additional primers append Illumina sequencing adapters and indices to create the final NGS-ready library (Fig. 3a). During this process, genomes containing different CRISPR array sequences were pooled and amplified together, which raises the concern of amplification bias that could distort the true representation of variants in the final library. As a thought experiment, if the PCR amplification is perfect, then each template DNA molecule would be duplicated, so the relative fraction of each variant would be preserved across cycles (Fig. 3b, left). However, in practice, amplification is likely imperfect, which causes only a fraction of template being duplicated. As a result, variants might not be amplified proportionally, so the relative abundances of variants would shift away from their original proportions (Fig. 3b, right).

Figure 3 | PCR amplification bias in NGS library preparation. a, Schematic of two-step PCR used in library preparation. Illumina sequencing adapters were appended to the 5'- and 3'-ends in the final library by primers. The number of cycles and annealing temperatures (Tm) are parameters to be optimized. b, Illustration of amplification bias. Under perfect amplification (100%), all templates are duplicated, and the relative abundance of variants is preserved. In contrast, incomplete amplification (for example, 50% duplication efficiency) can lead to disproportionate representation of variants, shifting the observed fractions away from the initial distribution. Created by biorender.com.

To verify this thought experiment, we simulated PCR amplification in Python starting with an experimental dataset from Bhattarai-Kline et al. (Bhattarai-Kline et al., 2022). The dataset reported the number of sequencing reads corresponding to different CRISPR array categories, including wild type, single insertions (spacers A, B, and N), and double insertions (spacers AA, AB, AN, BA, BB, BN, NA, NB, and NN). From this dataset, we calculated the initial fraction of all 13 categories (Fig. 4a). Since the spacer AA category contained no reads, it was excluded in subsequent analysis.

The simulation process was shown in Fig. 4b. For the first round of PCR, we assumed an initial population of 800,000 individual genomes as reported previously (Lear et al., 2023). The number of genomes assigned to each category was obtained by multiplying this total by its fractional abundance and rounding up to the nearest integer. To model amplification, we tested a gradient of amplification fractions ranging from 0.1 to 1.0 in increments of 0.1, where a value of 1.0 represents perfect duplication of all templates in each cycle. At every cycle, the number of templates to be replicated was calculated as the amplification fraction multiplied by the total population size. These templates were then allocated across categories using multivariate hypergeometric sampling, such that the number of new copies for each category was drawn in proportion to its relative abundance in the population. The amplification was simulated for 25 cycles. The second round of PCR was simulated in a similar manner but started with about 2.33 × 10¹⁰ genomes, corresponding to the addition of 5 ng of first-round PCR product (210 bp) as calculated by NEBCalculator. This round was modeled for 8 cycles. In both simulations, the amplification fraction varied across a gradient from 0.1 to 1.0, and a total of 100 replications were performed for each amplification fraction. After each replication was simulated, the final fractions of each category were compared to the original fractions to assess the extent of bias introduced by PCR.

To evaluate the deviation between the initial and final relative fractions in our PCR simulations, we employed the Kullback–Leibler divergence (KLD). Let \(P\) and \(Q\) denote the initial and final fraction distribution across all \(n\) categories of CRISPR array variants before and after PCR amplification. For category \(i\), \(p_i\) and\(q_i\) represent the relative abundances of that category in the initial and final distributions, respectively.

The difference at category \(i\) can be quantified by the log-likelihood ratio shown below, which reflects how much more (or less) likely category \(i\) is under the initial distribution compared to the final distribution.

Taking the expectation of this log-likelihood ratio under the true distribution \(P\) yields the KLD.

Simulation results for round 1 and round 2 are shown in Fig. 4c and Fig. 4d, respectively. In both cases, the Kullback–Leibler divergence (KLD) between the initial and final distributions decreased as the amplification fraction increased, indicating that higher amplification efficiency reduces PCR bias. Because round 1 involved more cycles than round 2 (25 versus 8), it produced higher overall KLD values on the order of 1E-5, whereas round 2 showed lower KLD values on the order of 1E-10. Despite this difference in magnitude, both rounds demonstrated the same negative correlation between KLD and amplification fraction. These results highlight that amplification efficiency is a critical parameter in preserving the true distribution of variants during library preparation.

Figure 4 | PCR bias simulation. a, Distribution of CRISPR array categories reported by Bhattarai-Kline et al. (2022), including wild type (WT), single insertions (N, B, A), and double insertions (NN, BN, NA, NB, AN, BA, BB, AB, AA). Read counts and fractions were calculated from sequencing data; category AA contained no reads and was excluded from further analysis. b, Schematic of the PCR simulation workflow. Round 1 starts with an initial pool of 800,000 genomes that were amplified over 25 cycles. For round 2, 2.33 × 10¹⁰ templates (equivalent to 5 ng of round 1 product) were amplified for 8 cycles. At each cycle, a fraction of templates was selected by multivariate hypergeometric sampling. Selected templates were duplicated and mixed back with the rest unselected templates. The final distributions after amplification were compared to the initial fractions using Kullback–Leibler divergence (KLD). A total of 100 replicates were performed for each amplification fraction ranging from 0.1 to 1.0. c, PCR simulation for round 1 (25 cycles, 100 replicates). Higher amplification fractions reduce the Kullback–Leibler divergence (KLD) between the initial and final distributions. d, PCR simulation for round 2 (8 cycles, 100 replicates). As in round 1, greater amplification efficiency lowers KLD. For all data, the mean KLD values of 100 replicates were shown, and error bars indicate the standard deviations. Created by biorender.com.

PCR parameters and reagents optimization

To maximize amplification efficiency, we need to carefully optimize PCR parameters. First, the number of PCR cycles is a critical factor (Fig. 5a). If too few cycles were performed, the product yield would be insufficient for the next round of PCR or for sequencing. If too many cycles were run, the reaction would enter a plateau phase where amplification efficiency declines, leading to poor effective amplification fractions and significant PCR bias as simulated before. To estimate an appropriate cycle number, we performed a simple calculation. In the first round of PCR, we began with 800,000 E. coli BL21(AI) genomes as the initial template, which corresponds to 3.706 ng of dsDNA according to the NEBioCalculator. Because the target region to amplify is only 142 bp within the 4,530,564 bp E. coli BL21(AI) genome (Bhawsinghka et al., 2020), the effective starting DNA mass for the amplicon is given by

We then assumed an amplification efficiency of 85% per cycle (i.e., 1.85-fold amplification). To ensure sufficient product for downstream steps while avoiding the plateau phase, we targeted a final concentration of 10 ng/µL in a 50 µL reaction. Hence, the expected yield after (x) cycles can be expressed as shown below. Solving this equation indicated that 25 PCR cycles were required.

For the second round of PCR, we used 5 ng of the round 1 PCR product (210 bp) as the starting material. This amount corresponds to approximately 2.3 × 10¹⁰ double-stranded DNA molecules according to NEBioCalculator, which is vastly greater than the number of genomes present in the initial template (8.0 × 10⁵). This enabled the full coverage of all variants in the initial library. Because the target region in the second round of PCR spans the entire length of the round 1 amplicon, the effective template mass remains 5 ng. Using the same rationale as before, we estimated that 8 cycles would be sufficient for the second amplification. Based on these estimations, we set the number of PCR cycles as 25 for the first round and 8 for the second round. These values were applied in both the simulations described earlier and all subsequent experiments.

The other two factors influencing PCR yield are annealing temperature and types of DNA polymerase (Fig. 5b-c). To optimize annealing temperature, we performed gradient PCR ranging from 50 °C to 65 °C. The yield was determined by the band brightness on agarose gel electrophoresis. For polymerase selection, the original paper employed Q5 High-Fidelity DNA Polymerase (NEB, cat. no. M0491L) (Lear et al., 2023). However, KAPA HiFi DNA Polymerase (Roche, cat. no. KK2102) is more widely used in NGS library preparation due to its robust and consistent amplification performance (Han et al., 2023), and we therefore chose it for gradient PCR test instead of Q5. Nevertheless, because KAPA HiFi is relatively expensive and not easily accessible in China, we also evaluated T8 High-Fidelity Master Mix (Tsingke, cat. no. TSE111), a domestically produced alternative. Using the same annealing temperature gradient for both KAPA and T8, we found that T8 consistently achieved high yields comparable to KAPA (Fig. 5d-e). Based on these results, we recommend T8 for other Chinese iGEM teams, while maintaining KAPA as the polymerase of choice for international teams. The optimal annealing temperature for the first round of PCR was determined to be about 52 °C for both KAPA and T8 (Fig. 5d). In contrast, the second round proved less sensitive to temperature (Fig. 5e), so we chose 55 °C, the midpoint of the tested gradient, for all subsequent experiments.

Figure 5 |PCR parameter optimization. a, Schematic showing the effect of cycle numbers on PCR yield. Too few cycles result in insufficient product, whereas too many cycles cause saturation and amplification bias. b, Diagram illustrating the effect of annealing temperature (Tm) on primer binding. Excessively high Tm prevents binding; optimal Tm promotes specific binding; low Tm leads to nonspecific binding. c, Comparison of high- versus low-yield DNA polymerases (DNAPs). They differ in amplification efficiency and hence bias. d, Gradient PCR for the first round of amplification using KAPA HiFi DNA Polymerase (left) and T8 High-Fidelity Master Mix (right) across annealing temperatures from 50 °C to 65 °C. e, Gradient PCR for the second round of amplification using KAPA (left) and T8 (right). Product yield was less sensitive to annealing temperature in this round. Created by biorender.com.

In summary, our simulations demonstrated that increased yield reduces PCR bias, highlighting the importance of maximizing amplification efficiency. To achieve this, we optimized three key parameters: the number of cycles, annealing temperature, and choice of DNA polymerase. These optimized conditions were subsequently adopted in our Retro-Cascorder protocols to ensure accurate preservation of variant distributions during library preparation.

Protocols for Retro-Cascorder

Here, we present the detailed experimental procedures of the Retro-Cascorder system, adapted from Bhattarai-Kline et al., 2022 with several modifications. Briefly, magnetic bead–based purification steps described in the original protocol were replaced with commercial purification kits to reduce overall cost and simplify handling. Moreover, this protocol assumes the use of the original expression plasmid pSBK.079 and signal plasmid pSBK.134 (Bhattarai-Kline et al., 2022). However, users may adapt the workflow for custom plasmid constructs as needed. The sources of most reagents and equipment are specified in the materials section to make our protocol more replicable. When the vendor is not critical (e.g., common laboratory chemicals such as NaCl), we indicate the source as “general” in the reagents and equipment list. The procedures documented here include two representative experimental modules: event ordering and acquisition efficiency testing, which serve as example measurement workflows that can be customized for specific applications. In addition, because some reagents and consumables used in our work were obtained from Chinese domestic suppliers, we provide both domestic and international versions of the protocol. In the international version, domestically sourced materials such as T8 High-Fidelity Master Mix and TIANquick Mini Purification Kit have been substituted with functionally equivalent reagents available from globally accessible vendors such as Roche and Zymo. To minimize redundancy on the wiki, the domestic version is presented in the main text, while both versions are available as downloadable PDF files at the end of this section.

Materials

Plasmids

• Expression plasmid pSBK.079 (Addgene, cat. no. 187218)

• Signal pSBK.134 (Addgene, cat. no. 187219)

Strains

• BL21(AI) super competent cells (Beyotime, cat. no. D1011S)

Recording experiments

• LB media powder (general)

• Agar (general)

• L-arabinose (general)

• IPTG (general)

• Chloramphenicol (general)

• Ampicillin (general)

• Anhydrotetracycline hydrochloride (general)

• Choline chloride (general)

• Glycerol (general)

• 14 mL bacterial culture tubes (JET BIOFIL, cat. no. TUB011140)

• 50 mL Bio-reaction tube (JET BIOFIL, cat. no. BRT010050)

• L-shaped cell spreader (JET BIOFIL, cat. no. CSP012014)

• Gene Pulser/MicroPulser Electroporation Cuvettes, 0.1 cm gap (Bio-Rad, cat. no. 1652083)

Next-generation sequencing library preparation

• Nuclease-free water (general)

• KAPA HiFi DNA Polymerase (Roche, cat. no. KK2102)

• T8 High-Fidelity Master Mix (Tsingke, cat. no. TSE111)

• 250 bp DNA Ladder (Tsingke, cat. no. TSJ105-100)

• DNA Clean & Concentrator-5 (Zymo Research, cat. no. D4014)

• TIANquick Mini Purification Kit (TIANGEN, cat. no. DP203)

• Anhydrous ethanol (general)

• Agarose (general)

• 50X TAE buffer (242 g Tris base, 57.1 mL/L glacial acetic acid, 100 mL/L of 0.5 M EDTA solution, pH = 8.0)

• Primers for deep sequencing (general)

• 1.5 mL microcentrifuge tubes (general)

• PCR tubes (general)

Equipment

• Water bath (general)

• Bacterial shaker (general)

• Thermocycler (general)

• Electrophoresis device (general)

• MicroPulser Electroporator (Bio-Rad, cat. no. 1652100)

• Benchtop microcentrifuge (general, compatible with 200 µL PCR tube)

• Centrifuge (general, compatible with 1.5 and 2.0 mL microcentrifuge tubes)

• NovaSeq X Plus Sequencing System (Illumina, cat. no. 20084804)

Procedure

Media and stock solution preparation

1. Prepare 10% glycerol as follows.
   Mix 50 mL glycerol with ddH₂O until the total volume reaches 500 mL. Autoclave at 121 °C 20 min.

2. Prepare LB media as follows.
   Dissolve 25 g LB media powder into 1 L ddH₂O. Autoclave at 121 °C 20 min.
   For agar plate, mix 25 g LB media powder and 18 g agar with 1 L ddH₂O. Autoclave at 121 °C 20 min.
   Note: users could also prepare LB media using 5 g/L yeast extract, 10 g/L tryptone, and 10 g/L NaCl if the pre-mixed LB media power were not available.

3. Prepare antibiotics and inducer stock solutions. Sterilize all solutions by filtering through 0.22 µm membrane.
   100 mg/mL ampicilin: dissolve 1000 mg ampicilin sodium into 10 mL ddH₂O
   34 mg/mL chloramphenicol: dissolve 340 mg chloramphenicol into 10 mL ddH₂O
   1 M IPTG: dissolve 2.383 g IPTG into 10 mL ddH₂O
   1 M L-arabinose: dissolve 3 g L-arabinose into 20 mL ddH₂O
   2 mM anhydrotetracycline hydrochloride: dissolve 18.5 mg anhydrotetracycline hydrochloride into 20 mL anhydrous ethanol
   1 M choline chloride: dissolve 2.7926 g choline chloride into 20 mL ddH₂O

4. When supplementing antibiotics or inducers into LB media, follow the dilution fold shown below.
   100 mg/mL Ampicillin: 1000X (e.g., add 40 µL stock solution into 40 mL LB media)
   34 mg/mL Chloramphenicol: 1000X (e.g., add 40 µL stock solution into 40 mL LB media)
   1 M IPTG: 1000X (e.g., add 40 µL stock solution into 40 mL LB media)
   1 M L-arabinose: 100X (e.g., add 400 µL stock solution into 40 mL LB media)
   2 mM Anhydrotetracycline hydrochloride: 10000X (e.g., add 4 µL stock solution into 40 mL LB media)
   1 M Choline chloride: 100X (e.g., add 400 µL stock solution into 40 mL LB media)

Transformation of signal and expression plasmids into BL21(AI)

1. Place one aliquot of E. coli BL21-AI super competent cells on ice to thaw. When the aliquot is fully thawed (5–10 min), add 10 ng signal (pSBK.134) into in 1–3 µL total volume to the competent cell and gently swirl solution with pipette tip. Keep on ice for 30 min.

Note: As suggested in the original report (Lear et al., 2023), the transformation order of plasmids does matter. The signal should be transformed before the expression plasmid to avoid spacer acquisition before recording starts.

2. Heat shock cells by placing the tube in a water bath heated to 42 °C for 1 min. Place the tube back on ice for 1 min.

3. Add 900 µL LB medium to the tube and place in the bacterial shaker at 37 °C at 220 r.p.m. for 1 h for recovery.

4. Centrifuge the recovered culture at 4,000 g for 5 min. Remove 800 µL supernatant and use the rest 200 µL media to resuspend cell pellet by pipetting up and down. Plate entire transformation on a LB agar plate supplemented with 34 µg/mL chloramphenicol. Incubate the plate overnight at 37 °C for 16 h.

5. The following morning, check the LB agar plate for the presence of bacterial colonies. Take LB agar plates from 37 °C and leave them at 4 °C until the evening.

6. In the evening, add 3 mL of LB medium containing chloramphenicol (34 µg/mL) into a 14 mL bacterial culture tube. Inoculate the tube with one bacterial colony from the LB agar plate. Transfer tube to bacterial shaker set at 37 °C at 220 r.p.m. overnight for 16 h.

7. The following morning, prepare two 50 mL bio-reaction tubes, each containing 15 mL of LB medium supplemented with 34 µg/mL chloramphenicol. Inoculate each tube with 30 µL of the overnight culture. Incubate the tubes at 37 °C with shaking at 220 r.p.m. until DO600 reaches 0.6

Note: according to our experiences, this step typically requires approximately 4 hours of incubation.

8. During the incubation, place one 1 mm gap electroporation cuvette, four 1.5 mL microcentrifuge tubes, and about 50 mL of 10% glycerol on ice to prechill.

9. After approximately 4 hours, or once the OD600 reaches 0.6, centrifuge each culture at 4,000 × g for 10 minutes at 4 °C. Discard the supernatant and keep the tubes on ice.

10. Add 10 mL of ice-cold 10% (v/v) glycerol to each bio-reaction tube and resuspend the cell pellets by gentle pipetting. Centrifuge at 4,000 × g for 10 minutes at 4 °C. Discard the supernatant and keep the tubes on ice.

11. Repeat the previous step to ensure thorough washing with glycerol.

12. Resuspend the pellet from each tube in 1 mL of ice-cold 10% glycerol, then combine the two suspensions into a single tube. Aliquot 600 µL of the combined cell suspension into four 1.5 mL microcentrifuge tubes.

13. Centrifuge all four 1.5 mL microcentrifuge tubes at 4,000 × g for 10 minutes at 4 °C. Carefully remove the supernatant using a pipette, and resuspend each pellet in 60 µL of ice-cold 10% (v/v) glycerol.

14. Take one tube of resuspended cells and add 1 ng of the Cas1–Cas2 expression plasmid (pSBK.079). Mix thoroughly by gentle pipetting. Transfer the cell–plasmid mixture into a pre-chilled electroporation cuvette. Dry the outside of the cuvette by wiping with a paper towel and place the cuvette into the electroporation system. Electroporate using the Ec1 setting on BioRad MicroPulser. If using an electroporation system other than Bio-Rad, then the parameter setting should be 1.8 kV, 25 µF and 200 Ω.

15. Immediately after electroporation, add 900 µL of LB medium to the cuvette and gently mix to resuspend the cells. Transfer the mixture to a clean, sterile 1.5 mL microcentrifuge tube and incubate at 37 °C with shaking at 220 r.p.m. for 1 hour for recovery.

16. Spread 200 µL of the recovered culture onto an LB agar plate supplemented with 34 µg/mL chloramphenicol and 100 µg/mL ampicillin. Incubate the plate at 37 °C overnight.

17. The next morning, check the formation of bacterial colonies on the LB agar plate. Then, proceed to the next section.

Recording transcriptional activity

In this section, we document the procedures for transcriptional event ordering and spacer acquisition efficiency testing, presented as two separate parts. These serve as representative experimental workflows demonstrating how the Retro-Cascorder system records and quantifies transcriptional signals over time.

Option 1: Transcriptional event ordering

1. In the morning, prepare six 14 mL bacterial culture tubes, each containing 3 mL of LB medium supplemented with 100 µg/mL ampicillin and 34 µg/mL chloramphenicol. Inoculate each tube with a single bacterial colony harboring both the signal plasmid (pSBK.134) and expression plasmid (pSBK.079). Incubate the cultures at 37 °C with shaking at 220 r.p.m. for 9 hours.

2. After 9 hours of incubation, split each culture into three subcultures by transferring 60 µL of the original culture into three new 14 mL bacterial culture tubes, each containing 3 mL of fresh LB medium supplemented with 100 µg/mL ampicillin, 34 µg/mL chloramphenicol, 1 mM IPTG, and 10 mM L-arabinose.

   1. For the first subculture, add 200 nM anhydrotetracycline to induce the pTet promoter on the signal plasmid, initiating the first transcriptional event to express barcode A.

   2. For the second subculture, add 10 mM choline chloride to initiate the second transcriptional event, which corresponds to barcode B.

   3. For the third subculture, do not add any additional inducers to serve as the negative control.

In total, this results in 18 tubes (6 biological replicates × 3 induction conditions). Incubate all cultures at 37 °C with shaking at 220 r.p.m. for 16 hours.

3. The following morning, dilute 150 µL of culture into a new bacterial culture tube with 3 mL LB containing the same antibiotics and inducers as previous step. Transfer tubes to bacterial shaker set at 37 °C at 220 r.p.m. Incubate for 8 h.

4. After 8 h, perform a second induction round by transferring 60 µL of each culture into a new 14 mL bacterial culture tube containing 3 mL of fresh LB medium with the same antibiotics and basal inducers (1 mM IPTG and 10 mM L-arabinose).

   1. For subcultures originally induced with aTc (barcode A), add 10 mM choline chloride to initiate the second transcriptional event (barcode B).

   2. For subcultures originally induced with choline chloride (barcode B), add 200 nM anhydrotetracycline to initiate the second transcriptional event (barcode A).

   3. For negative control subcultures, do not add any additional inducers.

This induction swap generates all possible transcriptional order combinations (A→B, B→A, and negative control). Incubate all cultures at 37 °C with shaking at 220 r.p.m. for 16 hours.

5. The following morning, dilute 150 µL of each culture into a new 14 mL bacterial culture tube containing 3 mL of fresh LB medium with the same antibiotics and inducers as in the previous step. Incubate the cultures at 37 °C with shaking at 220 r.p.m. for 8 h.

6. After 8 hours, the total culture duration will have reached 48 hours. Stop the incubation and collect 0.5 mL of culture from each tube into clean 1.5 mL microcentrifuge tubes.

7. Centrifuge the samples at 12,000 × g for 1 minute. Discard the supernatant and resuspend each pellet in 1 mL of nuclease-free water.

8. Repeat the previous step once more to wash the cells thoroughly.

9. Transfer 50 µL of each resuspended sample into individual PCR tubes. Boil the samples at 95 °C for 10 minutes in a thermocycler to lyse the cells, then allow them to cool to room temperature for 5 minutes before storing at –20 °C until further use.

Option 2: Spacer acquisition efficiency testing

1. In the morning, prepare three 14 mL bacterial culture tubes, each containing 3 mL of LB medium supplemented with 100 µg/mL ampicillin and 34 µg/mL chloramphenicol. Inoculate each tube with a single bacterial colony harboring both the signal plasmid (pSBK.134) and expression plasmid (pSBK.079). Incubate the cultures at 37 °C with shaking at 220 r.p.m. for 9 hours.

2. After 9 hours of incubation, split each culture into three subcultures by transferring 60 µL of the original culture into three new 14 mL bacterial culture tubes, each containing 3 mL of fresh LB medium supplemented with 100 µg/mL ampicillin and 34 µg/mL chloramphenicol.

   1. For the first subculture, add 1 mM IPTG and 10 mM L-arabinose to induce Cas1–Cas2 expression, and supplement with the appropriate signal inducer (e.g., 200 nM anhydrotetracycline or 10 mM choline chloride) to activate expression of retron-derived barcodes on the signal plasmid.

   2. For the second subculture, add 1 mM IPTG and 10 mM L-arabinose to induce Cas1–Cas2 expression only, without any signal inducers. This serves as the baseline control for retron-independent spacer acquisition.

   3. For the third subculture, do not add any inducers to serve as the negative control.

In total, this results in 9 tubes (3 biological replicates × 3 induction conditions). Incubate all cultures at 37 °C with shaking at 220 r.p.m. for 16 hours.

3. The following morning, dilute 150 µL of culture into a new bacterial culture tube with 3 mL LB containing the same antibiotics and inducers as previous step. Transfer tubes to bacterial shaker set at 37 °C at 220 r.p.m. Incubate for 8 h.

4. After 8 hours, the total culture duration will have reached 24 hours. Stop the incubation and collect 0.5 mL of culture from each tube into clean 1.5 mL microcentrifuge tubes.

5. Centrifuge the samples at 12,000 × g for 1 minute. Discard the supernatant and resuspend each pellet in 1 mL of nuclease-free water.

6. Repeat the previous step once more to wash the cells thoroughly.

7. Transfer 50 µL of each resuspended sample into individual PCR tubes. Boil the samples at 95 °C for 10 minutes in a thermocycler to lyse the cells, then allow them to cool to room temperature for 5 minutes before storing at –20 °C until further use.

Next-generation sequencing library preparation

1. Thaw frozen bacterial samples from the previous step.

2. Prepare the first-round PCR as follows. The UMI primer should be unique for each sample to enable the identification of its source after next-generation sequencing (NGS). These unique molecular identifiers allow demultiplexing of pooled sequencing data by linking each read to its original experimental condition. We recommend preparing a master mix containing all reagents except the boiled bacterial sample and the UMI primer. Dispense 46.4 µL master mix into each PCR tube.

Note: For the transcriptional event ordering experiment, two 50 µL PCR reactions should be prepared for each sample using two different UMI primers. This ensures that a sufficiently large number of unique genomes are sampled, allowing the calculated ordering scores to more accurately reflect the true distribution of CRISPR array patterns within the population.

3. Run the PCR temperature program by following the steps shown below.

4. Purify all PCR products by following the manual provided by TIANGEN PCR Purification Kit. Remember to add anhydrous ethanol to the concentrated wash buffer for the first time using it. Elute the PCR product using 50 µL nuclease-free water. Use nanodrop to determine concentrations.

5. Dilute all PCR products to about 5 ng/µL.

6. Prepare the second-round PCR as follows. We recommend preparing a master mix containing all reagents except the template. Dispense 49 µL master mix into each PCR tube.

Note: Again, for the transcriptional event ordering experiment, two 50 µL PCR reactions should be prepared for each first-round PCR product from previous steps.

7. Run the PCR temperature program by following the steps shown below.

8. Purify all PCR products by following the manual provided by TIANGEN PCR Purification Kit. Remember to add anhydrous ethanol to the concentrated wash buffer for the first time using it. Elute the PCR product using 50 µL nuclease-free water. Use nanodrop to determine concentrations.

9. Take 5 µL from each sample and mix them together into a 1.5 mL centrifuge tube. Then, take 5 µL mixture and add gel loading dye for electrophoresis to verify the band size. The main band should be 286 bp. If the band showed expected size, then the sample could be sent for sequencing using Illumina NovaSeq X Plus PE150.

10. Submit the library to NGS vendors like Azenta. Order 1 million reads for each sample.

Note: for example, since we have 18 samples in the event ordering experiment, we should order 18 million reads for the whole library. This is suggested in the original report .

10. Analyze NGS data using Python pipelines.

The international and Chinese domestic versions of the protocol, as well as the primers used in this study, are available for download at the following links.

For the international version, download the full PDF [here].

For the Chinese domestic version, download the full PDF [here].

For all primers used in NGS library preparation, download the spreadsheet [here].

Software

Upon receiving the raw NGS data, we processed them using Python pipelines adapted from previously reported methods (Lear et al., 2023). While the overall framework of the original pipeline was retained, several key modifications were introduced to tailor the workflow to the specific requirements of our experimental design (Fig. 6).

As the first step, we employed the paired-end 150 bp sequencing (PE150) on the Illumina NovaSeq X Plus platform, rather than the more costly 250-cycle strategy used in the original report. To reconstruct full-length CRISPR arrays, we merged the forward and reverse reads using FLASH (Fast Length Adjustment of SHort reads), which leverages the overlapping region between paired reads (Magoč & Salzberg, 2011). This merging strategy reliably recovered contiguous sequences spanning up to four spacers: two newly incorporated and two old spacers on the E. coli BL21(AI) genome. In this way, the event ordering scores could still be calculated as it only requires coverage of two new spacers. Meanwhile, PE150 is the standard configuration for Illumina sequencing, so our sample could be pooled with others to split the lane. This would substantially reduce the sequencing cost and hence make our protocol more affordable for other iGEM teams.

Following read assembly, we first identified the unique molecular identifier (UMI) incorporated during library preparation to assign each read to its sample that corresponds to the specific replicate and induction condition (Fig. 6, Step 2). Next, CRISPR repeat sequences were detected, and the intervening sequences were extracted as spacers (Fig. 6, Step 3). Each spacer was then compared against reference sequences to determine its identity as spacer A, spacer B, an ancestral spacer (“old”), or background (N). Based on the identities of all spacers within a read, the entire CRISPR array was classified into one of several possible states: wild-type (no new spacers), single-expansion (containing one new spacer: A, B, or N), or double-expansion (containing two new spacers in all possible combinations including AA, AB, AN, BA, BB, BN, NA, NB, or NN). Finally, arrays were quantified to determine the frequency of each state, and A/N, B/N, and A/B scores were calculated to infer the relative order of transcriptional events (Fig. 6, Step 4). These curated data formed the basis for downstream visualization and interpretation of CRISPR-based signal recording.

We have uploaded all raw sequencing data, index–sample mapping files, and Python analysis scripts to our Software Repository on iGEM GitLab. These resources include the complete data processing workflow, from read merging and CRISPR array classification to score calculation and visualization. Detailed instructions for running the analysis, including setup, dependencies, and usage examples, are provided in the repository’s documentation, which can be accessed [Here]

Figure 6 | Data processing pipeline for Retro-Cascorder. Raw sequencing reads generated on the Illumina NovaSeq X Plus platform using paired-end 150 bp sequencing were merged with FLASH to recover full-length CRISPR arrays (Step 1). Unique molecular identifiers (UMIs) incorporated during library preparation were identified to assign reads to their corresponding samples (Step 2). CRISPR repeat boundaries were then detected, and the intervening sequences were extracted as spacers (Step 3). Each spacer was compared against reference sequences and classified as retron-derived spacer A, retron-derived spacer B, ancestral (old) spacer, or background (N). Arrays were subsequently categorized into wild-type, single-expansion, or double-expansion states based on spacer content and order. Quantification of these array states enabled calculation of A/N, B/N, and A/B scores to infer the relative timing of transcriptional events (Step 4). The curated results were visualized as violin plots to illustrate event-ordering outcomes. Created by biorender.com.

Cost analysis

To assist other iGEM teams in planning and budgeting for similar experiments, we conducted a cost analysis to estimate the total expenses required for one complete batch of the event ordering experiment. Although we provided both domestic and international versions of the protocol, the cost analysis was performed for the Chinese domestic version only. The reason is that the relative prices of experimental materials may vary significantly across countries. Since we have little knowledge of reagent pricing and sourcing in other countries, any international cost estimation would likely be unreliable. Foreign teams are encouraged to adapt the Table 1 shown below to generate their own region-specific cost estimates. All reported prices are based on the actual values we obtained from our suppliers at the time of purchase. The exact costs for other teams may vary depending on institutional vendor contracts, promotions, sponsorships, or other external factors. All prices were reported in Chinese Yuan (CNY).

To ensure our estimation remained conservative, we deliberately overestimated the cost by calculating reagent usage based on quantities higher than the minimum required for each material. As a result, the total estimated cost is likely higher than the actual experimental expense. Overall, performing one complete batch of the event ordering experiment required approximately 1,500 CNY, corresponding to an average of 83.7 CNY per sample (1,500 CNY / 18 samples). This analysis demonstrates that the Retro-Cascorder workflow is cost-effective for most iGEM teams. For comparison, the early registration fee for iGEM 2025 is 5,500 USD (approximately 38,500 CNY), meaning that the cost of conducting one batch of the event ordering experiment accounts for less than 5% of a team's registration fee. Therefore, we expect that other iGEM teams should be able to adopt and perform our measurement protocols without significant financial burden.

Results

Recording the orders of transcriptional events

In this experiment, we tested the ordering of transcriptional events using the Retro-Cascorder system in E. coli BL21(AI) carrying the signal and expression plasmids (Fig. 7a). Six single colonies were picked to establish independent biological replicates, which were cultured in LB medium supplemented with chloramphenicol (Cm) and ampicillin (Amp) to maintain plasmids, along with IPTG and L-arabinose (Ara) to induce the expression of the recording machinery throughout the experiment. Anhydrotetracycline (aTc) was used as an inducer to trigger the expression of barcode A, while choline chloride (Cho) was used to trigger barcode B. These six cultures were then split into three groups with different inducer conditions: (i) aTc followed by Cho (aTc–Cho), (ii) Cho followed by aTc (Cho–aTc), and (iii) no inducers as a negative control. Each group was maintained in the initial condition for 24 hours before switching to the second condition for another 24 hours. For the control group, all cultures remained uninduced for 48 hours. Following growth, samples were processed for next-generation sequencing (NGS) library preparation using both KAPA High-Fidelity DNA Polymerase and T8 High-Fidelity Master Mix. This allowed us to compare whether PCR reagent choice influences the outcome of Retro-Cascorder event recording.

The resulting spacer acquisition profiles were analyzed to calculate A/N, B/N, and A/B scores for all groups (Fig. 7b–c). Ideally, the aTc–Cho (AB) condition should yield positive values across all three scores, whereas the Cho–aTc (BA) condition should give uniformly negative values. However, the violin plots showed that the distributions were highly noisy, with scores either spreading broadly between –1.0 and 1.0 or even displaying opposite trends in both of the two libraries prepared by KAPA (Fig. 7b) and T8 (Fig. 7c). Additionally, the A/B scores were left blank because no informative arrays with the expected A → B → leader or B → A → leader patterns were observed. These results suggest that the failure to produce cleanly separated scores arises from low recording efficiency, leaving too few informative arrays to confidently resolve event order.

To further investigate this issue, we examined the percentage of arrays expanded with spacers A, B, and N (Fig. 7d–e). The frequency of arrays carrying A or B spacers was on the order of 0.001%, which is two orders of magnitude lower than the ~0.1% reported in the original Retro-Cascorder study (Bhattarai-Kline et al., 2022). The incorporation frequency of N spacers was also slightly lower than the reported ~10%. Together, these results indicate that inefficient spacer acquisition underlies the noisy and inconsistent temporal determination observed in this experiment.

Figure 7 | Recording the orders of transcriptional events by Retro-Cascorder. a, Schematic of the experimental workflow. Six independent E. coli BL21(AI) colonies carrying pSBK.134 (ncRNA) and pSBK.079 (retron-RT and Cas1-Cas2) were picked and tested in aTc→Cho (AB), Cho→aTc (BA), and uninduced conditions. After 48 hours of recording, NGS libraries were prepared using both KAPA and T8 enzymes. b, Violin plots of A/N, B/N, and A/B scores from the KAPA-prepared library. c, Violin plots of A/N, B/N, and A/B scores from the T8-prepared library. d, Bar plots of the fraction of arrays incorporating spacers A, B, or N in the KAPA-prepared library. e, Bar plots of the fraction of arrays incorporating spacers A, B, or N in the T8-prepared library. Mean values from six replicates were shown for all data points in d-e, with error bars representing standard deviation. Individual replicate values are shown as black dots. Created by biorender.com.

Characterizing Cas1 variants with higher activity

To improve the spacer acquisition efficiency of the Cas1–Cas2 complex, we tested three previously reported variants with enhanced activity (Yosef et al., 2023), E269G, V76L, and the double mutant E269G–V76L, and compared them to the wild-type Cas1 (Fig. 8a). For each variant, the Cas1–Cas2 plasmid was co-transformed with an ncRNA plasmid carrying the aTc-inducible expression of barcode A. Cultures were grown in the presence of IPTG and L-arabinose (Ara) to activate the recording machinery, and each culture was split into two conditions: one induced with aTc to trigger barcode A expression and one left uninduced. Three biological replicates were performed for each Cas1 variant. Samples were collected at 12-hour intervals (0, 12, 24, 36, and 48 hours after induction) for NGS library preparation (Fig. 8b).

Sequencing results suggested that the percentage of arrays with spacer A incorporation remained extremely low across all variants, in both uninduced (Fig. 8c) and aTc-induced (Fig. 8d) conditions. Among them, WT and V76L exhibited slightly higher A array incorporation compared to E269G and the E269G–V76L double mutant. Because barcode A incorporation was limited, we next examined barcode N, which is assumed to be generated at a constant rate (Bhattarai-Kline et al., 2022) and can therefore serve as a proxy for Cas1-mediated integration efficiency. Similar incorporation frequencies of N arrays were observed under uninduced (Fig. 8e) and induced (Fig. 8f) conditions. Consistent with the A array data, V76L showed higher N array incorporation than WT, whereas E269G and the double mutant unexpectedly showed lower incorporation, despite previous reports suggesting that these mutations enhance spacer acquisition efficiency. These discrepancies might be due to differences in measurement methods between our study and the original report, or the possibility that N array generation rates were not identical across all conditions.

Figure 8 | Characterization of Cas1 variants in Retro-Cascorder recording. a, Schematic showing the Cas1 variants tested: WT, E269G, V76L, and the double mutant E269G–V76L. Each Cas1–Cas2 plasmid was co-transformed with an ncRNA plasmid carrying the aTc-inducible barcode A. b, Experimental workflow. Three colonies of E. coli BL21(AI) harboring the two plasmids in a were cultured under IPTG and L-arabinose (Ara) induction, with or without aTc to trigger barcode A expression. Samples were collected every 12 hours (0, 12, 24, 36, and 48 h) for NGS library preparation. c, Percentage of arrays incorporating barcode A without aTc induction. d, Percentage of arrays incorporating barcode A with aTc induction. e, Percentage of arrays incorporating barcode N without aTc induction. f, Percentage of arrays incorporating barcode N with aTc induction. Mean values from three replicates were shown for all data points in c-f, with error bars representing standard deviation. Created by biorender.com.

Troubleshooting retron-derived spacer acquisition

To troubleshoot the low efficiency of retron-derived spacer acquisition, we tested whether both barcode A and B could be incorporated under induction conditions (Fig. 9a). Three colonies of E. coli BL21(AI) harboring the original Cas1–Cas2 plasmid and the ncRNA plasmid carrying barcodes A and B were used. This time, the expression plasmid was electroporated into BL21(AI) harboring the signal plasmid to ensure efficient transformation. Cultures were then split into three conditions: (i) IPTG + Ara + aTc + Cho, where spacers A, B, and N were expected to be incorporated simultaneously; (ii) IPTG + Ara only, where only spacer N was expected; and (iii) no inducers, serving as a negative control. After 24 hours of culture, samples were collected for NGS.

Sequencing results revealed clear differences in spacer acquisition across conditions (Fig. 9b). For spacer N, we observed an approximately 7-fold increase in incorporation when IPTG and Ara were added, confirming that the Cas1 and Cas2 were induced and functional. For spacer A, its incorporation also increased upon IPTG and Ara induction, indicating not only active Cas1–Cas2 but also a functional retron-RT capable of reverse-transcribing ncRNA into DNA spacers for integration. However, additional induction with aTc did not further enhance incorporation, suggesting that spacer A was largely derived from transcriptional leakage of the TetR-regulated promoter and that aTc failed to effectively drive barcode A expression. For spacer B, the BetI-regulated promoter is known to exhibit high basal leakage, and Cho induction resulted in about a 2-fold increase in incorporation, which is consistent with previously reported values (Bhattarai-Kline et al., 2022). Taken together, these results indicate that the poor performance of retron-derived spacer acquisition stems primarily from weak or ineffective induction of barcode A by aTc, rather than from dysfunction of the Cas1–Cas2 complex or the retron-RT. This finding accounts for the insufficient barcode A incorporation observed in both the event-ordering and Cas1 variant characterization experiments, which in turn led to noisy and inconclusive results.

Figure 9 | Troubleshooting retron-derived spacer acquisition. a, Troubleshooting experiment workflow. Three colonies of E. coli BL21(AI) carrying the original Cas1–Cas2 plasmid and an ncRNA plasmid encoding barcodes A and B were picked. Cultures were divided into three conditions: IPTG + Ara + aTc + Cho , IPTG + Ara only , and no inducers. After 24 hours, samples were collected for NGS library preparation. b, Spacer acquisition results. The percentage of arrays incorporating spacers A, B, or N was shown for all three conditions. Bar heights represent the mean of three replicates, with error bars showing standard deviation. Individual replicate values are shown as black dots; replicates with zero values are not visible due to the log scale. Created by biorender.com

Since aTc is known to be light-sensitive, a possible explanation for the poor incorporation of spacer A is that the reagent had degraded during storage. To test whether the aTc stock was still active, we used a positive control strain from the Marionette collection (Meyer et al., 2019), which carries a plasmid expressing EYFP under the control of an aTc-inducible promoter (Fig. 10a). Cultures were grown with and without aTc, and fluorescence was compared. No yellow fluorescence was detected upon aTc induction (Fig. 10b), indicating that the reagent had lost activity. We then repeated the assay using a freshly prepared aTc solution, and this time robust yellow fluorescence was observed (Fig. 10c). These results suggest that the experimental failures were caused by degradation of the aTc stock. For future experiments, freshly prepared aTc should be used to ensure reliable recording performance.

Figure 10 | Verifying aTc activity using an aTc-inducible EYFP reporter strain. a, Schematic of the Marionette plasmid in which EYFP is expressed under control of an aTc-inducible promoter. b, The fluorescence image of cultures grown with or without the original aTc stock. c, The fluorescence image of cultures grown with freshly prepared aTc. Created by biorender.com.

Discussion

In this study, we evaluated the performance of the Retro-Cascorder system for recording the order of transcriptional event using retron-derived spacers. Our event-ordering experiments produced noisy temporal scores, largely due to the low acquisition efficiency of retron-derived spacers that resulted in too few informative double-expansion CRISPR arrays for reliable analysis. Characterizing Cas1 variants showed that the V76L mutation modestly improved spacer acquisition, as reflected by a higher percentage of N arrays compared to the wild-type Cas1. However, spacer A incorporation remained limited across all variants, indicating that the bottleneck lies in retron-derived spacer acquisition rather than Cas1 catalytic activity alone. Troubleshooting confirmed that the Cas1–Cas2 complex was functional and that spacer B incorporation was inducible, but incorporation of spacer A was consistently poor. Further investigation revealed that the root cause was degradation of the aTc inducer during long-term storage, which explained the lack of effective barcode A induction observed across all experiments.

One procedural difference may also have contributed to the variation in performance. In the original Retro-Cascorder protocol, the ncRNA plasmid is first transformed into BL21(AI) chemically competent cells. Transformants were then made electrocompetent to transform the Cas1–Cas2 plasmid. By contrast, in our event-ordering and Cas1 mutant experiments, we introduced both plasmids by chemical transformation to simplify the protocol and avoid the need for an electroporator. This adjustment was motivated by the desire to make this measurement approach more accessible to the iGEM community, particularly high school teams with limited access to specialized equipment like the electroporator. However, in our troubleshooting experiments, where we reverted to the original electroporation-based protocol, spacer B incorporation appeared normal. Whether this difference in spacer B acquisition was caused by transformation method or an experimental artifact remains unclear, but for now we recommend following the original electroporation protocol to ensure consistent spacer acquisition.

Looking ahead, addressing the low efficiency of retron-derived spacer acquisition will be the key to improve the recording of transcriptional event orders. One potential solution is to adopt engineered retrons with modified RNA structures that confer greater intracellular stability, as has been demonstrated in mammalian cells (Cattle et al., 2025). Fusing xrRNA elements to the retron ncRNAs encoding barcodes A and B might help protect transcripts from degradation and improve their availability for reverse transcription and hence integration into the CRISPR array. In parallel, further testing of Cas1 variants, such as V76L, in combination with these retron enhancements could help determine whether integration efficiency can be synergistically improved. Finally, adhering closely to the original electroporation-based transformation workflow and using freshly prepared reagents would reduce technical variability and help ensure reproducibility. Together, these improvements could overcome the current limitations of retron-derived spacer acquisition and make Retro-Cascorder a more robust and accessible tool for recording the orders of transcriptional events.

Folin-Ciocalteus assay: A method to characterize keratinase activity

Introduction

Keratinases are a group of enzymes that cleave peptide bonds within keratin substrates, such as cat fur, thereby releasing free tyrosine residues. The concentration of free tyrosine increases proportionally with the extent of keratin degradation and thus serves as a quantitative indicator of enzyme activity. As previously reported, the Folin–Ciocalteu reagent can be used to detect tyrosine (Fig. 11). The phenolic group of tyrosine reduces the reagent's tungstate and molybdate components, which generates a blue chromophore that can be quantified by measuring the absorbance at 660 nm wavelength using a spectrophotometer (Pérez et al., 2023). Conventionally, one unit of enzyme activity is defined as an increase of 0.01 Abs₆₆₀. Inspired by the iGEM 2018 Interlab study (Beal et al., 2020), we also reported keratinase activity as the amount of tyrosine released, providing an absolute unit of measurement to enable direct comparison of our results with those from other iGEM teams.

Figure 11 | Principle of the Folin–Ciocalteu assay. Keratinase-catalyzed hydrolysis of keratin (e.g., cat hair) releases free tyrosine residues, which reduce the Folin–Ciocalteu reagent containing tungsten and molybdenum oxides to their lower oxidation states, forming a blue chromophore. The intensity of the blue color correlates with the concentration of released tyrosine and thus reflects keratinase activity. Created by biorender.com.

Experimental workflow

The experimental workflow for the Folin–Ciocalteu assay is illustrated in Fig. 12. The protocol begins with the establishment of a standard curve. The L-tyrosine stock solution was serially diluted with reaction buffer to generate a gradient of tyrosine concentrations for calibration. To each tube, Folin–Ciocalteu (FC) solution A was added and incubated at room temperature for 10 minutes, followed by the addition of FC solution B and incubation for another 30 minutes. During this step, blue coloration gradually develops in the test tube. After incubation, the solutions should be immediately transferred to cuvettes, as the blue chromophore fades rapidly. Abs₆₆₀ was measured using a spectrophotometer and plotted against tyrosine concentration, and a standard curve was generated by linear regression.

For the enzyme activity assay, keratinase samples are diluted to 2 mg/mL. A 0.5 mL aliquot of the diluted enzyme is added to the reaction buffer containing cat fur so that each reaction contains 1 mg of keratinase, which minimizes variability arising from differences in enzyme concentration and enables straightforward activity calculation. All samples are incubated at 37 °C for 16 hours with three biological replicates per keratinase. After incubation, the reactions are terminated by filtration through a 0.22 µm membrane filter, which removes residual cat fur. The filtrate is transferred to a new 15 mL centrifuge tube and treated with FC solutions A and B following the same protocol as for standard curve generation. Abs₆₀₀ is measured for each sample to reflect the keratinase activity expressed as the amount of tyrosine released.

Figure 12 | Experimental workflow of the Folin–Ciocalteu assay. The protocol begins with the establishment of a standard curve (Step 1), where the blue coloration corresponding to different L-tyrosine concentrations is quantified by Abs₆₆₀. In Step 2, the enzyme activity assay is performed to quantify keratinase activity based on the degradation of cat fur and the resulting release of tyrosine. Created by biorender.com.

Protocols for Folin-Ciocalteus assay

Here, we present the detailed experimental procedures of the Folin-Ciocalteus assay, adapted from iGEM 2021 SZ_SHD with several modifications. We extended the reaction time to 16h overnight since we used cat fur as the substrate, and clarified the enzyme concentration used and proper controls. We obtained Folin-Ciocalteus (FC) solution from Solarbio, which requires two FC solutions A and B, so we modified the procedure based on the protocol provided by Solarbio as well. The sources of most reagents and equipment are specified in the materials section to make our protocol more replicable. When the vendor is not critical (e.g., common laboratory chemicals such as NaCl), we indicate the source as "general" in the reagents and equipment list. A downloadable PDF file for this protocol is also available at the end of this section

Materials

Reagents

• ddH₂O (general, equivalent to Milli-Q water)

• Na₂CO₃ (general)

• NaOH (general)

• CuSO₄·5H₂O (general)

• Potassium sodium tartrate (general)

• Folin-Ciocalteus (FC) solution B (Solarbio cat. no. F8060)

• Tris-HCl (general)

• CaCl₂·2H₂O (general)

• L-tyrosine (Beyotime cat. no. ST1505-50g)

• Dulbecco's Phosphate-Buffered Saline (DPBS) (Servicebio cat. no. G4200)

• Cat fur (Obtained from pet shops directly)

Consumables

• Cuvettes (general)

• 5 mL centrifuge tubes (general)

• 15 mL centrifuge tubes (general)

Equipment

• Spectrophotometer (general)

Procedure

Stock solution preparation

1. Prepare 0.2 M NaOH solution as follows.
   Dissolve 0.8 g of NaOH into 100 mL of ddH₂O.

2. Prepare Folin-Ciocalteus solution A1 (FC) as follows.
   Dissolve 1 g of Na₂CO₃ in 50mL of 0.2 M NaOH solution.

3. Prepare 1% (w/v) potassium sodium tartrate solution as follows.
   Dissolve 2 g of potassium sodium tartrate into 200 mL ddH₂O.

4. Prepare Folin-Ciocalteus solution (FC) A2 as follows.
   Dissolve 0.5 g of CuSO₄·5H₂O in 100 mL of 1% (w/v) Potassium sodium tartrate.

5. Prepare Folin-Ciocalteus (FC) solution A as follows.
   Mix 50 mL of FC solution A1 with 1 mL of FC solution A2.

Note: FC solution A expires quickly, so only freshly prepared FC solution A should be used for the assay.

6. Prepare 1.1 mM L-tyrosine as follows.
   Dissolve 0.02 g of L-tyrosine into 100 mL ddH₂O. Gently heat the solution to facilitate dissolution.

7. Prepare the reaction buffer as follows.
   Dissolve 4.846 g of Tris-HCl and 0.147 g of CaCl₂·2H₂O in 400 mL of ddH₂O. Adjust the pH to 7.5 using 0.2 M NaOH. Then, add ddH₂O to bring the final volume to 500 mL.

Standard curve establishment

1. Prepare nine 15 mL centrifuge tubes and add 0, 25, 50, 100, 200, 250, 500, 750, and 1000 µL of 1.1 mM L-tyrosine stock solution to the respective tubes. Add reaction buffer to bring the total volume of each tube to 1 mL by following the table shown below. Label the tubes according to the final tyrosine concentrations: 0, 27.5, 55, 110, 220, 275, 550, 825 and 1100 µM.

2. Add 5 mL of Folin–Ciocalteu (FC) solution A to each tube and incubate at room temperature for 10 minutes.

3. Add 0.5 mL of FC solution B to each tube, mix thoroughly, and incubate at room temperature for 30 minutes. A gradual development of blue coloration should be observed over time.

4. Immediately after incubation, transfer 3.5 mL of each reaction mixture into clean cuvettes. Measure the absorbance at 660 nm wavelength using a spectrophotometer promptly, as the blue color was observed to fade over time. Record the readings for each sample.

5. Subtract the absorbance of the 0 µM sample as the blank from all other measurements to correct for the background value. Subtract Plot the standard curve by correlating all Abs₆₆₀ values with the corresponding L-tyrosine concentrations. Perform a linear regression with zero y-intercept (i.e., y = kx) to generate the standard curve, which should display a strong linear correlation between absorbance and tyrosine concentration.

Enzyme Activity Assay

Reaction

1. Quantify the protein concentration of the purified keratinase using a BCA assay, then dilute the enzyme to a standardized concentration of 2 mg/mL based on the measured values.

2. Prepare three 5 mL centrifuge tubes for each keratinase sample. In each tube, mix 0.5 mL of the diluted keratinase solution with 2.5 mL of reaction buffer and 25 mg of cat fur.

3. Prepare two 5 mL centrifuge tubes for controls.

a. Blank control: Add 0.5 mL of DPBS to 2.5 mL of reaction buffer containing 25 mg of cat fur.

b. Negative control: Add 0.5 mL DPBS to 2.5mL buffer without cat fur.

4. Incubate all samples at 37℃ for 16 h overnight.

Assay

1. The following morning, after 16 hours of incubation, stop the reaction by filtering each sample through a 0.22 μm membrane filter to remove residual cat fur.

2. Transfer 1 mL of each filtrate (including controls) into clean 15 mL centrifuge tubes.

3. Add 5 mL of Folin–Ciocalteu (FC) solution A to each tube and incubate at room temperature for 10 minutes.

4. Add 0.5 mL of FC solution B to each tube, mix thoroughly, and incubate at room temperature for 30 minutes.

5. Immediately transfer 3.5 mL of each reaction mixture into clean cuvettes. Measure the absorbance at 660 nm using a spectrophotometer and record the readings.

Data Analysis

1. Calculate the average Abs₆₆₀ value from the triplicate measurements for each keratinase sample.

2. Subtract the average Abs₆₆₀ of the blank control from that of each sample, then divide the result by 0.01 to obtain the total enzyme activity in units (U) for each keratinase.

3. Calculate the specific activity by dividing the total enzyme activity by the mass of keratinase used in the reaction. Under this protocol, each reaction contains 1 mg of keratinase, so the specific activity (U/mg) is numerically identical to the total activity (U).

4. Refer to the Folin-Ciocalteus standard curve, use the equation to convert the Abs₆₆₀ values into the amount of released tyrosine as the absolute units.

For the PDF document of the protocol, download at [Here].

Results

Calibration of the tyrosine standard curve

We first established a standard curve calibration for the Folin–Ciocalteu (FC) assay. As a qualitative demonstration, we tested L-tyrosine samples at concentrations of 0, 27.5, 55, 110, 165, 220, 275, 550, and 1100 µM, which exhibited a clear blue color gradient that intensified with increasing tyrosine concentration (Fig. 13a). For quantitative calibration, we refined the concentration range to 0, 27.5, 55, 110, 220, 275, 550, 825, and 1100 µM to ensure optimal curve fitting and dynamic range coverage. The absorbance at 660 nm of each sample was measured using a spectrophotometer, and the resulting standard curve showed a strong linear correlation between absorbance and tyrosine concentration (R² = 0.9959, Fig. 13b), confirming the reliability of the calibration for converting measured absorbance values into corresponding tyrosine concentrations.

Figure 13 | Calibration of the Folin–Ciocalteu assay using L-tyrosine standards. a, The color gradient of L-tyrosine solutions at concentrations of 0–1100 µM after reacting with the Folin–Ciocalteu reagent. Deeper blue coloration was produced as the tyrosine concentration increased. The 15 mL tubes were arranged across two racks and later combined into a single image for display. All samples were processed simultaneously in the same experiment. b, Standard curve generated by plotting Abs₆₀₀ against tyrosine concentration, showing a strong linear correlation (R² = 0.9959).

Characterization of keratinase activities

We then performed the Folin–Ciocalteu (FC) assay to evaluate the enzymatic activities of three purified keratinases: KrMKU3, KrMKU3 without signal peptide, and KrOcL9 without signal peptide. All three keratinases produced visibly darker blue coloration than the blank sample, indicating tyrosine release from cat hair substrates (Fig. 14a). Quantification of Abs₆₆₀ revealed comparable activities among the three enzymes, although KrOcL9 (no signal peptide) exhibited a noticeably larger standard deviation, primarily due to one abnormally low outlier measurement (Fig. 14b). The enzyme activity could be expressed as relative unit by simply dividing Abs₆₆₀ values to 0.01. To express enzymatic activity in absolute terms, Abs₆₆₀ values were further converted into the amount of tyrosine released (nmol). As shown in the equations below, the absorbance was first converted to tyrosine concentration using the equation derived from the standard curve, and then multiplied by the total 3 mL reaction volume to yield the total amount of released tyrosine. The resulting absolute activity is plotted on the right-hand vertical axis of Fig. 14b, which is directly proportional to the Abs₆₆₀ values shown on the left-hand axis.

Figure 14 | Keratinase activity characterization by Folin–Ciocalteu (FC) assay using cat hair as the substrate. a, Colorimetric results of FC assay reactions. All keratinase-treated samples displayed darker blue coloration than the blank control. The abbreviation w/o s.p. refers to "without signaling peptide". b, Quantification of keratinase activity based on Abs₆₆₀ (left axis) and corresponding the amount of released tyrosine in nmol (right axis). Bar heights represent the mean of triplicates. Error bars represent standard deviations of triplicate measurements, and individual data points are shown in black dots.

Discussion

In this study, we successfully established a linear standard curve for the Folin–Ciocalteu (FC) assay, confirming a strong correlation between Abs₆₆₀ and L-tyrosine concentration. This correlation validated Abs₆₆₀ as a reliable quantitative metric for determining tyrosine release in keratinase reactions. We then applied this calibration to evaluate the keratinase activity of three purified enzymes using cat fur as a natural keratin substrate. All three keratinases, KrMKU3, KrMKU3 without signal peptide, and KrOcL9 without signal peptide, demonstrated catalytic activity, as indicated by increased absorbance at 660 nm and corresponding tyrosine release. Furthermore, by reporting our results as the absolute amount of tyrosine released (nmol), rather than arbitrary absorbance values, we provide a standardized unit of measurement. This approach enables direct comparison of keratinase activity data across different iGEM teams and experimental setups, contributing to greater reproducibility and community-wide data integration.

Despite these promising results, several limitations remain. First, our current measurements were performed at a single enzyme concentration and a single time point, which restricts our understanding of the enzyme kinetics and potential concentration-dependent effects. Future experiments could include a concentration gradient or time-course analysis to more accurately determine specific activity and validate the reproducibility of our results. Additionally, the FC assay provides an indirect measurement of keratinase activity by detecting the concentration of the degradation product (tyrosine), rather than directly quantifying keratin breakdown. To further confirm enzyme performance, complementary direct measurements could be performed in parallel. For example, the Azokeratin assay monitors absorbance changes at 415 nm due to keratin-bound dye release (Gonzalo et al., 2020). This direct measurement of keratin degradation would strengthen the accuracy and robustness of enzyme activity assessment and help cross-validate the findings from the FC assay.

In conclusion, our application of the Folin–Ciocalteu assay to quantify cat fur degradation represents a practical example of heterogeneous catalysis measurement, where the enzyme acts on an insoluble natural substrate. This setup poses additional challenges compared to homogeneous systems, such as uneven substrate distribution, mass transfer limitations, and difficulties in accurately quantifying reaction progress. Nevertheless, by demonstrating that the FC assay can reliably detect and quantify soluble reaction products like tyrosine, we provide a versatile framework for studying other heterogeneous enzymatic reactions. Future iGEM teams can adapt the principles of this approach by linking measurable soluble products to the catalytic conversion of insoluble substrates to meet their own project needs. Beyond our specific case of keratin degradation, this workflow highlights a generalizable strategy for bridging biochemical assays with real-world, complex substrates, contributing to the foundation for enzyme activity measurements within the synthetic biology community.

References

↑ Askary, A., Chen, W., Choi, J., Du, L. Y., Elowitz, M. B., Gagnon, J. A., Schier, A. F., Seidel, S., Shendure, J., Stadler, T. & Tran, M. (2024). The lives of cells, recorded. Nature Reviews Genetics, 1–20. https://doi.org/10.1038/s41576-024-00788-w

↑ Beal, J., Farny, N. G., Haddock-Angelli, T., Selvarajah, V., Baldwin, G. S., Buckley-Taylor, R., Gershater, M., Kiga, D., Marken, J., Sanchania, V., Sison, A., Workman, C. T., Aachen, Pehlivan, M., Roige, B. B., Aalto-Helsinki, Aarnio, T., Kivisto, S., Koski, J., … Zhou, J. (2020). Robust estimation of bacterial cell count from optical density. Communications Biology, 3(1), 512. https://doi.org/10.1038/s42003-020-01127-5

↑ Bhattarai-Kline, S., Lear, S. K., Fishman, C. B., Lopez, S. C., Lockshin, E. R., Schubert, M. G., Nivala, J., Church, G. M. & Shipman, S. L. (2022). Recording gene expression order in DNA by CRISPR addition of retron barcodes. Nature, 608(7921), 217–225. https://doi.org/10.1038/s41586-022-04994-6

↑ Bhawsinghka, N., Glenn, K. F. & Schaaper, R. M. (2020). Complete Genome Sequence of Escherichia coli BL21-AI. Microbiology Resource Announcements, 9(10), 10.1128/mra.00009-20. https://doi.org/10.1128/mra.00009-20

↑ Cattle, M. A., Aguado, L. C., Sze, S., Venkittu, S., Wang, Y., Papagiannakopoulos, T., Smith, S., Rice, C. M., Schneider, W. M. & Poirier, J. T. (2025). An enhanced Eco1 retron editor enables precision genome engineering in human cells without double-strand breaks. Nucleic Acids Research, 53(14), gkaf716. https://doi.org/10.1093/nar/gkaf716

↑ Cong, L., Ran, F. A., Cox, D., Lin, S., Barretto, R., Habib, N., Hsu, P. D., Wu, X., Jiang, W., Marraffini, L. A. & Zhang, F. (2013). Multiplex Genome Engineering Using CRISPR/Cas Systems. Science, 339(6121), 819–823. https://doi.org/10.1126/science.1231143

↑ Gonzalo, M., Espersen, R., Al‐Soud, W. A., Falco, F. C., Hägglund, P., Sørensen, S. J., Svensson, B. & Jacquiod, S. (2020). Azo dying of α‐keratin material improves microbial keratinase screening and standardization. Microbial Biotechnology, 13(4), 984–996. https://doi.org/10.1111/1751-7915.13541

↑ Han, Y., Li, W., Filko, A., Li, J. & Zhang, F. (2023). Genome-wide promoter responses to CRISPR perturbations of regulators reveal regulatory networks in Escherichia coli. Nature Communications, 14(1), 5757. https://doi.org/10.1038/s41467-023-41572-4

↑ Horvath, P. & Barrangou, R. (2010). CRISPR/Cas, the Immune System of Bacteria and Archaea. Science, 327(5962), 167–170. https://doi.org/10.1126/science.1179555

↑ Lear, S. K., Lopez, S. C., González-Delgado, A., Bhattarai-Kline, S. & Shipman, S. L. (2023). Temporally resolved transcriptional recording in E. coli DNA using a Retro-Cascorder. Nature Protocols, 18(6), 1866–1892. https://doi.org/10.1038/s41596-023-00819-6

↑ Magoč, T. & Salzberg, S. L. (2011). FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics, 27(21), 2957–2963. https://doi.org/10.1093/bioinformatics/btr507

↑ Yosef, I., Mahata, T., Goren, M. G., Degany, O. J., Ben-Shem, A. & Qimron, U. (2023). Highly active CRISPR-adaptation proteins revealed by a robust enrichment technology. Nucleic Acids Research, 51(14), 7552–7562. https://doi.org/10.1093/nar/gkad510

Meyer, A. J., Segall-Shapiro, T. H., Glassey, E., Zhang, J. & Voigt, C. A. (2019). Escherichia coli “Marionette” strains with 12 highly optimized small-molecule sensors. Nature Chemical Biology, 15(2), 196–204. https://doi.org/10.1038/s41589-018-0168-3

↑ Pérez, M., Dominguez-López, I. & Lamuela-Raventós, R. M. (2023). The Chemistry Behind the Folin–Ciocalteu Method for the Estimation of (Poly)phenol Content in Food: Total Phenolic Intake in a Mediterranean Dietary Pattern. Journal of Agricultural and Food Chemistry, 71(46), 17543–17553. https://doi.org/10.1021/acs.jafc.3c04022

Yosef, I., Mahata, T., Goren, M. G., Degany, O. J., Ben-Shem, A. & Qimron, U. (2023). Highly active CRISPR-adaptation proteins revealed by a robust enrichment technology. Nucleic Acids Research, 51(14), 7552–7562. https://doi.org/10.1093/nar/gkad510