Secretome translation shaped by lysosomes and lunapark-marked ER junctions

Cell culture and electroporation

U-2 OS (ATCC) cells were cultured in phenol-red free DMEM supplemented with 10% fetal bovine serum (Corning), 2 mM L-glutamine (Corning), and 100 IU penicillin and 100 µg ml^{− 1} streptomycin (CellGro) at 37 °C and 5% CO₂. For the tet-inducible system, FBS that was tested free of tetracycline (Cytvia) was used. Cells were grown on either diluted fibronectin (1:100, Millipore) or Matrigel (1:100, Corning)-coated coverslip chambers or sterilized cover glasses (no. 1.5 Round (EMS)). Lonza Electroporation kit with the U-2 OS specific set-up suggested by the company was used for transient transfection, which was performed 16–20 h prior to imaging. In addition, HeLa (ATCC), COS-7 (ATCC), HEK293T (ATCC) and HT1080 (ATCC) cells were cultured and electroporated according to the manufacturer’s instructions. Cells were routinely tested for mycoplasma contamination.

Plasmids and cloning

A list of plasmids used in this study is provided in Supplementary Table 1.

Generation of stable cell line and doxycycline induction

To generate lentivirus particles, HEK293T cells were transfected with MCP-GFP, MCP-HaloTag or scFv-sfGFP plasmids, along with viral packaging plasmids. The supernatant was collected 48–72 h after transfection, filtered through a 0.22-µm syringe filter, and concentrated using Lenti-X Concentrator (Takara). U-2 OS cells were exposed to lentiviral particles, and MCP-HaloTag and scFv-sfGFP double positive U-2 OS cells were generated via sequential viral transduction and sorted via fluorescence-activated cell sorting.

To generate cells stably expressing the protein of interest with only MS2 binding sites mRNAs, lentiviral transduction was used on pre-sorted MCP-GFP or MCP-HaloTag positive U-2 OS cells. Cells stably expressing a tet-inducible cytERM–SunTag were generated using the PiggyBac system. MCP-HaloTag/scFv-sfGFP double positive U-2 OS cells were then electroporated with cytERM-SunTag plasmid with hyperactive piggyBac transposase (hyPBase). After 4 days post-electroporation, the cells were selected under 10 µg ml⁻¹ Blasticidin S for two weeks. The stable cells expressing MCP-HaloTag/scFv-sfGFP/cytERM–SunTag were induced with 100 ng ml^{− 1} doxycycline 3–5 h prior to imaging.

Labelling JF dyes with HaloTag ligand and SNAPTag ligand

For cells expressing HaloTag, they were incubated in complete media with 100 nM JF646 HaloTag ligand (JF646-HTL) at 37 °C for 30 min and then washed twice. The rinsed cells were then equilibrated in the media at 37 °C for 30 min prior to imaging. Similar methods were used for PA-JF646-HTL, PA-JF549-HTL, JF549-HTL, JF646-HTL and JF635-HTL.

Cells expressing SNAPTag were incubated in complete media with 250 nM JF549 with chloropyrimidine SNAPTag ligand (JF549-cpSNAP) for 1 h at 37 °C and then washed twice. The rinsed cells were then washed twice again after 30 min to remove residual dyes, and then equilibrated in fresh media at 37 °C for 30 min prior to imaging.

Reagents for condition screening

Rapamycin (Sigma, 2.5 mg ml⁻¹ in DMSO), torin-1 (Cell Signaling, 1 mM in DMSO), cycloheximide (Sigma, 100 mg ml^{− 1} in DMSO), chloroquine (Sigma, 10 mM in H₂O), thapsigargin (Sigma, 1 mM in DMSO), protease cocktail inhibitor (ThermoFisher) and puromycin (ThermoFisher, 10 mg ml^{− 1} in H₂O). Amino acid-free medium is composed of amino acid-free DMEM (USBiological), 10% dialysed FBS (ThermoFisher), 4.5 g l⁻¹ glucose (Sigma), 110 mg l⁻¹ sodium pyruvate (Sigma), 3.7 g l⁻¹ sodium bicarbonate (Sigma), 2 mM L-glutamine (Corning), 100 IU penicillin and 100 µg ml^{− 1} streptomycin (CellGro).

Single-molecule tracking with HILO microscopy

Single molecule tracking of single ribosomes was conducted using a customized Nikon TiE inverted microscope with a split port for simultaneous imaging onto two Andor iXON DU-897 EM-CCD cameras, equipped with a 100× Apo TIRF 1.49 oil-immersion objective (Nikon). U-2 OS cells were co-transfected with L10A-HaloTag plasmid and either mEmerald Sec61b or SNAPf-Sec61b plasmids and imaged after 16 h. These cells were labelled with either PA-JF549-HTL or PA-JF646-HTL/JF549-cpSNAP³⁶. A brief pulse of 405 nm laser was used to excite and activate the PA-JF dyes. The incident angle for excitation lasers was manipulated to achieve highly inclined and laminated optical sheet (HILO) illumination. The emission resulting from simultaneous excitation of two fluorophores was split using either T647LPXR (Chroma) or T565LPXR (Chroma) and collected in two Andor iXon EM-CCDs with emission filters: ET525/50 m (Chroma) for GFP, ET605/70 m (Chroma) for PA-JF549 or JF549, and ET700/75 m (Chroma) for PA-JF646.

Single-particle tracking with spinning-disk confocal microscopy

Single molecule tracking for single mRNAs was conducted using a customized Nikon TiE inverted microscope, outfitted with a Yokogawa spinning-disk scan head (CSU-X1, Yokogawa) and Andor iXON EM-CCD cameras. Fluorescence was collected using a 100× Apo TIRF 1.49 oil-immersion objective (Nikon). For live-cell imaging, cells were imaged in complete medium and incubated with a stage heater (Tokai Hit) at 37 °C and 5% CO₂. Two-colour imaging was performed similarly to single-molecule tracking, as described above. Three-colour imaging was performed using trigger mode. The emission from 561 and 640 nm excitation was collected sequentially into the same camera using a multi-bandpass emission filter (Chroma, ZET405/488/561/647 m).

Immunofluorescence and PLA

For immunofluorescence labelling, U-2 OS cells stably expressing MCP–HaloTag, scFv–sfGFP or cytERM–SunTag were plated on Matrigel-coated 12 mm coverslips (EMS) in sterile 24-well tissue-culture graded plates (Corning). The cells were induced with 100 ng ml^{− 1} doxycycline (Sigma) and labelled with JF dyes as described above. The cells were then fixed with 4% (w/v) paraformaldehyde (PFA) + 0.2% glutaraldehyde for 15 min at room temperature. Cells were permeabilized with 0.1% Triton X-100 for 10 min, then treated with 1% BSA (Sigma) in phosphate-buffered saline (PBS) plus 0.05% Tween-20 (PBS-T) for 1 h. The primary antibody (Rabbit Anti-LNPK antibody, Sigma) was added overnight at 4 °C and washed with 1% BSA + PBS-T. The coverslip was incubated with secondary antibodies conjugated with Alexa Fluor 594 (ThermoFisher) for 1 h at room temperature and rinsed with PBS. The coverslip was mounted on a slide and imaged using a Zeiss 980 AiryScan equipped with a 63× objective.

PLA was performed using the Duolink In situ Orange Starter kit Mouse/Rabbit (Sigma). U-2 OS cells were plated on Matrigel-coated 12 mm coverslips (EMS) in sterile 24-well tissue-culture graded plates (Corning). The manufacturer’s protocol was followed, using the following antibodies: rabbit LNPK antibody (Sigma, HPA014205-25), mouse EEA1 antibody (BD Biosciences, 610456), mouse LAMP1 antibody (Abcam, ab25630), mouse TOMM20 antibody (66777-1-Ig) and rabbit REEP5 antibody (ProteinTech, 14643-1-AP).

Immunoblotting

For immunoblotting, U-2 OS cells were collected by scraping the plate in lysis buffer (25 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS) supplemented with EDTA-free protease inhibitor cocktail (Roche) and PhoSTOP (Roche), followed by sonication. The lysate was quantified with a BCA kit (ThermoFisher), following the manufacturer’s protocol and diluted using 4× NuPage LDS Sample buffer (ThermoFisher) supplemented with 0.1 M DTT (Sigma Aldrich). The samples were heated in a 95 °C heat block for 10 min. Equal amounts of lysate (~5 μg) were run on 4–12% NuPage Bis-Tris Mini protein gels (ThermoFisher). After electrophoresis, the proteins were transferred onto 0.2 µm Nitrocellulose membranes using the Trans-Blot Turbo system. The membranes were blocked using EveryBlot blocking buffer (Bio-Rad). Primary antibodies were diluted 1:1,000 in the same blocking buffer and incubated overnight at 4 °C with gentle agitation. After washing 3 times with TBS-T, secondary antibody staining was performed for 1 h at room temperature. The membranes were washed with TBS-T and developed with SuperSignal West Pico PLUS Chemiluminescent Substrate (ThermoFisher). The membranes were imaged on ChemiDoc (Bio-Rad) with optimized exposure for signal detection. Primary antibodies used here are: rabbit ATF-4 antibody (Cell Signaling, 11815S), rabbit LNPK antibody (Sigma, HPA014205-25), mouse Tubulin antibody (Millipore, 05-829), rabbit eIF2S1 (phosphor S51) antibody (Abcam, ab32157), and rabbit eIF2S1 antibody (Atlas Antibodies, HPA064885). Secondary antibodies used here are: goat anti-mouse Ig H&L-HRP (Abcam, ab205719) and goat anti-rabbit Ig H&L-HRP (Abcam, ab205718).

HCR-smFISH

For HCR-smFISH, we designed and purchased oligonucleotides that hybridize to endogenous human CD9 mRNA from Molecular Instruments. U-2 OS cells transiently transfected with mEmerald–Sec61β were plated on Matrigel-coated 12 mm coverslips in sterile 24-well tissue-culture graded plates. Cells were fixed with 4% (w/v) PFA + 0.2% glutaraldehyde for 15 min at room temperature. Hybridization was performed based on the manufacturer’s protocol for mammalian cells on a coverslip for single-molecule detection. The coverslips were mounted on a glass slide and imaged immediately.

Single-particle tracking analysis

L10A-HaloTag and mRNA tracking was performed using TrackMate software (FIJI). The resulting trajectories were imported into Matlab using a custom Matlab code to store x, y and t. Displacement of each trajectory was calculated, and each trajectory was categorized based on the mask to classify cytosolic trajectories and remove nuclear localizations using custom Matlab code.

The MSD of the population was calculated by averaging squared displacement (SD) at a given lag time. The apparent diffusion coefficient (D_app) was calculated based on a linear fit of MSD versus lag time using the built-in linear fit module in Matlab, described by equation (1).

where D_app is apparent diffusion coefficient, τ is lag time, n is number of dimensions (2) and σ is localization error. The first four points were used to fit the curve to estimate the apparent diffusion coefficient as many of these trajectories showed a confined diffusion.

Mean square displacement from each trajectory (MSD_τ) was calculated by averaging the SD within the same trajectories. For determining the translationally active population, we analysed trajectories that are longer than 9 steps at 1 Hz and averaged 9 individual SDs to generate MSD_τ, as described in equation (2).

Monte Carlo simulation of Brownian motion

Monte Carlo simulation of random walk was performed using a custom-built Matlab code. For each time step, individual particles choose displacement in each of the three Cartesian directions, and its distribution is defined by the free, 3D diffusion coefficient D. The squared displacement of the corresponding Gaussian propagator is sqrt(2Dt).

Motion-based analysis of translating mRNA

Displacements of single trajectories of mRNAs in the cytosol are calculated based on their time progression. Each trajectory longer than 10 steps was used to calculate the mean square displacement in 1 s. Each trajectory was then categorized based on the MSD cut-off of 0.055 µm² based on equation (2). For determining the translation fraction, we calculated by dividing the number of trajectories that are classified as translation by the number of total classified trajectories.

To generate pseudocoloured single-molecule movies, the positions and time-mark from categorized trajectory was mapped onto two blank matrices (x,y,t) where one represents translationally active and the other represents translationally silent mRNAs. The resulted single pixel movies were saved as tagged image file (tif). The gaussian blurring of 50 nm was applied to mimic the localization accuracy of each molecule. The resulted movie was then overlaid on the images of corresponding ER (Sec61β).

SunTag intensity analysis and FRAP

Foci of scFv_GCN4-sfGFP signal were localized and quantified using TrackMate. The background signal was subtracted and each (x,y,t) coordinate was correlated to the trajectories of corresponding mRNAs so that the maximum distance away is less than or equal to two pixels. Then, only mRNAs that are in the cytoplasm were selected for analysis. We calculated the translating fraction of mRNA by dividing SunTag(+) mRNAs over all tracked mRNAs in the cytoplasm.

FRAP experiments were performed on a customized Nikon TiE inverted scope equipped with a Yokogawa spinning-disk scan head (CSU-X1, Yokogawa) and Andor iXON EM-CCD cameras. The excitation and emission was performed on 100× Apo TIRF 1.49NA oil-immersion objective (Nikon). Photobleaching was performed using a Bruker Mini-scanner that scans the region of interest. Sequential excitation of 488 and 640 nm laser was performed to collect scFv–sfGFP and MCP–HaloTag signal, respectively through a multi-bandpass emission filter (Chroma, ZET405/488/561/647 nm). The cells were imaged prior to bleaching for 10 consecutive frames and imaged post-bleaching at 5 or 10 s per frame for >5 min. Particles were tracked using Imaris spot tracker (Oxford Instrument) and corresponding bleached SunTag signal was analysed over time.

Lysosome–SunTag correlation

Lysosomes were tracked using TrackMate with the proper threshold to detect the centre of the lysosome. The distance from the centre of lysosome to each mRNA positions were calculated. The minimum distance from lysosome to each mRNA was then used for analysis. SunTag signals that were quantified during the SunTag intensity analysis was referenced to the minimum lysosome distance. The histogram of average SunTag intensity versus average minimum lysosome distance was quantified by binning the lysosome distance at 500 nm interval, starting from 750 nm to 4.25 µm. The SunTag intensity of the corresponding average lysosome distance binned at a 500-nm interval starting from 750 nm to 4.25 µm was used to generate the histogram.

Lysosome–ER recruitment and analysis

To optogenetically recruit lysosomes and the ER, we constructed LAMP1–iLID and Sec61β–SspB expression plasmids. Specifically, LAMP1 was fused at its C-terminus to the photosensitive improved light-inducible dimer (iLID) domain, whereas Sec61β (a subunit of the ER Sec61 translocon complex) was fused at its N-terminus to the iLID-binding partner SspB. Distinct fluorescent tags (HaloTag for LAMP1–iLID and mCherry for Sec61β–SspB) were inserted between each protein domain. Optogenetic activation was initiated by exposing U-2 OS cells—either wild-type or lacking endogenous LNPK (LNPK KO)—to 488 nm laser in the region of interest. Time-lapse images were captured every 1–2 s throughout the recruitment assay.

All image processing and analyses were performed using ImageJ/Fiji. Lysosomes were segmented and tracked using the TrackMate plugin with a threshold-based method. Custom MATLAB scripts were then used to measure ER fluorescence intensity within the tracked lysosomal regions over time, and the resulting intensities were normalized for comparative analysis.

siRNA knockdown and knockdown efficiency

Twenty-five picomoles of small interfering RNAs (siRNAs) targeting LNPK or CLIMP63 (Dharmacon) were transfected into U-2 OS cells using Lipofectamine RNAiMAX (ThermoFisher) in Opti-MEM (ThermoFisher) in a 6-well chamber (Corning). The transfected cells were incubated for 48 to 72 h before imaging. Prior to imaging, the total RNA of the cells from same-day transfection was collected using RNA extraction kits (NEB). The concentration of total RNA was measured using UV spectrometer (ThermoFisher) and performed One-Step quantitative PCR with reverse transcription (RT–qPCR) (NEB) on Roche LightCycler. Knockdown efficiency was calculated based on calculating ΔΔC_q using ACTB mRNA as a standard. The list of primers used for this is listed in Supplementary Table 2.

For quantification of XBP-1 splicing in control, LNPK-KD and LNPK-KO cells under different treatments, we extracted total RNA from the cells using Monarch Total RNA extraction kit (NEB). The concentration of total RNA was measured using Nanodrop (ThermoFisher) and performed One-Step RT–qPCR (NEB) on Roche LightCycler using the provided protocol. The extent of unspliced and spliced transcripts of XBP-1 was quantified based on calculating ΔC_q using total XBP-1 transcript as a standard. The ratio between the estimated fraction of spliced/unspliced XBP-1 was as 2^{−ΔCq,spliced}/2^{−ΔCq,unspliced}.

CRISPR knockout of LNPK

To generate cells with targeted knockout of human LNPK, single guide RNAs (sgRNAs) were designed against LNPK (Synthego). The sgRNA sequences (AGCAAAAAAUGGGAGUGUCA, UGUAAACAGAUAGAGAACUG, UCAAGCAUUGGAAGAAUUUA) were then assembled with SpCas9 2NLS Nuclease (Synthego) as per the manufacturer’s instructions. The resulting sgRNA–Cas9 complex was then electroporated into U-2 OS cells using a Lonza 4D nucleofector programmed for U-2 OS cells. The cells were expanded and cultured for single colonies by diluting them in 96-well plates (Corning). The genome of each expanded colony was then collected using the Monarch Genomic DNA purification kit, and PCR was performed on the purified DNA using primers provided by Synthego (listed in the Supplementary Table 3). The resulting Sanger sequencing data was analysed using Synthego software to identify positive colonies. These colonies were then checked for LNPK expression by collecting the lysate and performing immunofluorescence or immunoblotting as described above.

Membrane protein extraction and quantification

For whole-cell extract, U-2 OS cells were collected via scraping the plate in lysis buffer (25 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS) supplemented with EDTA-free protease inhibitor cocktail (Roche) and PhoSTOP (Roche), followed by sonication. Membrane proteins were extracted from one million U-2 OS control and LNPK-KO cells using the Mem-PER PLUS Membrane Protein Extraction Kit (ThermoFisher, 89842) following manufacturer’s protocol. Protein quantification was performed using the Pierce BCA Protein Assay Kit (ThermoFisher, 23225), with absorbance measured at 562 nm using a Tecan 96-well plate reader.

HPG incorporation assay

The Click-It HPG Alexa Fluor 488 protein synthesis kit (ThermoFisher) was used to perform the HPG incorporation assay. Prior to the addition of HPG (50 µM), U-2 OS cells were incubated with methionine-free DMEM for 1 h. HPG was incorporated into the cells over a period of 1.5 h. As HPG can also be incorporated into mitochondria, a cycloheximide incubation control was incorporated for all tested conditions. To monitor HPG incorporation into the membrane fraction only, HPG-incorporated cells were incubated with a solution containing 0.025% digitonin, 115 mM potassium acetate, 25 mM HEPES pH 7.4, 2.5 mM MgCl₂, 2 mM EGTA, and 150 mM sucrose for 3 min at 37 °C and proceeded with fixation. For the whole-cell fraction, the cells were fixed using 4% PFA and 0.2% glutaraldehyde for 15 min, and the washing and labelling protocol followed the manufacturer’s instructions. The labelled cells were mounted using VECTASHIELD antifade with DAPI (Vectorlabs), imaged using Zeiss 980 AiryScan, and quantified using CellProfiler. To remove the effect from mitochondrial translation, the quantified signal of each condition was subtracted with the corresponding cycloheximide treatment.

CDF curve fit

Cumulative distribution function (CDF) curves were fit with a three-component fit (equation (3)):

Where A_x is the fraction of molecules with the diffusion coefficient D_x, D_x is the diffusion coefficient of the xth diffusive species in μm² s⁻¹, t is the single lag time in seconds, and r is the single displacement in micron.

RNA extraction from cultured cells

U-2 OS control and LNPK-KO cells were plated in a 6-well dish, briefly washed with PBS, and collected in 1 ml Trizol (Invitrogen, 15596026). Total RNA was extracted following the manufacturer’s protocol (Invitrogen, 15596026), and both the quality and quantity of the RNA were subsequently assessed using an RNA Qubit (Invitrogen) and an Agilent Bioanalyzer.

RNA sequencing

cDNA was prepared from 1 ng of total RNA as previously described³⁷. Reverse transcription was performed using a barcoded 3′ polyT primer containing a unique molecular identifier (UMI), a 5′ template-switch oligo (Integrated DNA Technologies), and Maxima H Minus Reverse Transcriptase (ThermoFisher, EP0752), followed by PCR amplification and cDNA purification.

Tagmentation and library preparation were then conducted with 600 pg of cDNA per sample using a modified Nextera XT (Illumina) protocol with dual-indexed P5NEXTPT5 and i7 primers (IDT). The libraries were purified according to the Nextera XT protocol, quantified by qPCR using the Kapa Library Quantification Kit (Kapa Biosystems), and sequenced on a NextSeq 2000. Read 1 (26 bp) captured the sample barcode and UMI, while read 2 (50 bp) sequenced the 3′ cDNA fragment. A PhiX control library (Illumina) was spiked in at a final concentration of 15% to enhance colour balance in Read 1.

Sequence alignment was performed as described previously³⁸. Sequencing adapters were trimmed from the reads using Cutadapt v2.10 prior to alignment with STAR v2.7.5c against the Homo sapiens GRCh38 genome assembly from Ensembl. Gene counts were generated using the STARsolo algorithm and subsequently analysed using a version of DESeq2 on MatLab.

Solvents and chemicals

We used water (Optima, W6–4), acetonitrile (Optima, A9554), methanol (Optima, A454SK-4), formic acid (Pierce, PI28905), acetone (A949-1, Fisher) and RapiGest SF Surfactant (Waters,186001861). Dithiothreitol (DTT, R0861, Thermo scientific), trifluoroacetic acid (TFA, 302031–100 ML), iodoacetamide (IAA, 407710) and ammonium bicarbonate (40867) were purchased from Sigma Aldrich. Trypsin (V5113) was purchased from Promega.

Membrane protein digestion and desalting

Membrane proteins were precipitated with cold acetone at 1:4 (v:v) ratio, −20 °C overnight. Protein pellets were centrifuged and washed with cold acetone. Residue acetone was then air dried, and the proteins were re-dissolved using 1% RapiGest according to the manufacturer’s protocol. Afterwards, proteins were reduced with 5 mM DTT (65 °C, 30 min) and alkylated using 11 mM IAA at ambient temperature in the dark for 30 min. Trypsin was added at a protein-enzyme ratio of 50:1 for overnight digestion at 37 °C. The digestion was quenched by adding 10% TFA to pH of ~1. The digest was further incubated at 37 °C for 45 min to degrade the RapiGest. The solutions were then centrifuged at 14,000 rpm for 10 min and the supernatant peptides were collected. Desalting of the peptides was performed using C18 ZipTip (Millipore, ZTC18M096) and the eluents were dried using SpeedVac (Thermo Scientific). The samples were stored at –80 °C before being re-suspended in 0.1% formic acid for LC–MS/MS analysis.

LC–MS/MS analysis

Liquid chromatography separation was performed on a Vanquish Neo System (Thermo Scientific) with an IonOptik Aurora Ultimate C18 column (25 cm length × 75 μm inner diameter × 1.7 μm particle size) at 50 °C and 300 nl min⁻¹ flow rate. Mobile phase A consisted of 0.1% formic acid in water. Mobile phase B consisted of 0.1% formic acid in 80% acetonitrile. Eluting peptides were ionized by electrospray ionization and then analysed by an Orbitrap Ascend Tribrid mass spectrometer (tune v.4.1.4244, Thermo Scientific). Ion transfer tube temperature was set to 275 °C. Source positive ion voltage was set to 1,850 V. For single-shot proteomics with data independent acquisition (DIA), the MS1 scan resolution was set to 120,000 (at m/z 200). MS1 scan range was 400–900 m/z, AGC target was 225%, maximum injection time mode was set to auto. Precursors were isolated with an isolation width of 12 m/z covering 145–1,450 m/z. Precursors were fragmented by HCD at an NCE of 26%. MS2 scans were acquired by Orbitrap with a resolution of 15,000. Maximum injection time was 59 ms for the Ascend.

Mass spectrometry data processing

For the acquired DIA data, raw files were processed directly using Proteome Discoverer (v.3.1.0.638) with the H. sapiens proteome FASTA file (UP000005640). The analysis applied CHIMERYS intelligent search algorithm, allowed for a maximum of one missed cleavage, with cysteine carbamidomethylation (+57.0215 Da) set as a fixed modification, and oxidation (M), phosphorylation (S,T) were selected as variable modifications. The target false discovery rate (FDR) was maintained at 0.01. Fragment mass tolerance was set as 20 ppm.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.