Study approval

All mouse experiments were approved by the University of Southern California Institutional Animal Care and Use Committee (principal investigator Y.-K.H.) and the University of South Florida Institutional Animal Care and Use Committee (principal investigator J.S.).

Mice and related reagents

Mice were housed in air-filtered clear cages under a 12-h light/12-h dark cycle and were fed ad libitum. Mice were obtained from the following sources: C57BL/6J (The Jackson Laboratory); Prox1-eGFP47 and Prox1-tdTomato46 (Mutant Mouse Resource and Research Centers); ROSA26-LSL-tdTomato (The Jackson Laboratory); Prox1-CreERT2 (a gift from T. Mäkinen, Uppsala University)71; Piezo1fl/fl (Piezo1tm2.1Apat/J; The Jackson Laboratory)72; Piezo1-transgenic line (CAG-LSL-Piezo1; previously generated and reported by the authors)13 and Dp(16)1Yey/+ (B6.129S7-Dp(16Lipi-Zbtb21)1Yey/J, The Jackson Laboratory)59,73. The absence of the potential toxicity of Cre protein expression was confirmed by comparing lymphatic vessel density and thickness as well as drainage functionality of wild-type and Prox1-CreERT2 mice that do not carry Piezo1 floxed alleles (Supplementary Fig. 3). Lymphatic Piezo1-knockout mice (Piezo1dLEC) have Prox1-CreERT2 and Piezo1fl/fl alleles with or without a lymphatic reporter allele (Prox1-tdTomato or Prox1-eGFP). Lymphatic Piezo1 overexpression mice (Piezo1TG_LEC) harbor Prox1-CreERT2 and Piezo1-transgenic alleles (CAG-LSL-Piezo1) with or without a lymphatic reporter allele. DS and Piezo1TG_LEC compound mice have Prox1-tdTomato, Dp(16)1Yey/+, Prox1-CreERT2 and Piezo1TG alleles. Males and females were nonselectively used for each experiment as we did not find sex-specific differences in phenotype. Tamoxifen (MP Biomedicals) was dissolved in dimethyl sulfoxide (DMSO) and diluted in sunflower seed oil (1:9 (vol/vol) DMSO:sunflower seed oil). Diluted tamoxifen was injected into pups (once) and adults (twice) at 50 mg per kg (body weight) per day to induce Cre recombinase activity. Yoda1 (Sigma-Aldrich) was dissolved in DMSO (710 µg ml–1) and diluted in phosphate-buffered saline (PBS; 3:100 (vol/vol) DMSO:PBS) before injection at a final concentration of 213 μg per kg (body weight) per day, which was previously established13,14. Supplementary Table 1 provides the strain, sex, number and age information of all mice used in each experiment. Mice of the indicated ages in Supplementary Table 1 were randomly selected for each experiment.

Cell culture-related assays

Human dLECs were isolated from deidentified human foreskin samples and cultured as previously described74,75 with approval by the University of Southern California Institutional Review Board (principal investigator Y.-K.H.). LECs lower than six to seven population passages were used for all experiments. Yoda1 was dissolved in DMSO (2 mM) and added to cell culture medium at a final concentration of 0.3–2 µM. For siRNA transfection, LECs were seeded in T75 flasks and incubated for 24 h. Control siRNA or Piezo1 siRNA (Dharmacon, L-020870-03-0005) was mixed with Lipofectamine RNAiMAX transfection reagent (Invitrogen, 13778150). The siRNA–reagent complex was added to the cell culture 24 h before performing western blotting or drainage tests using a lymphatics-on-chip (see below). A cell death detection enzyme-linked immunosorbent assay (ELISA) kit (Roche Molecular Biochemicals, 11544675001) was used to evaluate the potential cytotoxicity of Yoda1. LECs were seeded on six-well plates and incubated for 24 h. The cells were treated with 0.5, 1 or 2 μM Yoda1 or vehicle (DMSO) for 8 h before the cell death assay. For cell proliferation assays, LECs were seeded and incubated for 24 h and treated with vehicle, Yoda1 or Yoda1 + inhibitor (l-NAME (Selleckchem, S2877), capivasertib (Selleckchem, S8019), axitinib (Selleckchem, S1005), SAR131675 (Selleckchem, S2842) and cabozantinib malate (Selleckchem, S4001)). After 2 days, the cells were detached and counted with a Cellometer Auto T4 (Nexcelom Bioscience).

Western blotting

Western blotting assays were performed as described previously13. Antibodies were acquired from the following sources: β-actin (Sigma-Aldrich, A5441), p-CDH5 (Try 658; Thermo Fisher Scientific, 44-1144G), p-CDH5 (Try 685; ECM Biosciences, CP1981), CDH5 (Santa Cruz Biotechnology, sc-6458), p-AKT(Ser 473; Cell Signaling Technology, 4060), AKT (Cell Signaling Technology, 9272), eNOS (Cell Signaling Technology, 32027), p-eNOS (Ser 1177; Cell Signaling Technology, 9571), VEGFR2 (R&D Systems, AF357), p-VEGFR2 (Tyr 1054/Tyr 1059; Invitrogen, 44-1047G), VEGFR3 (Santa Cruz Biotechnology, sc-321), p-VEGFR3 (Tyr 1230/Tyr 1231; Cell Applications, CY1115), Prox1 (DSHB, AB2619013) and LYVE1 (R&D Systems, AF2089). Rabbit polyclonal anti-Piezo1 was generated by the authors (GenScript).

Immunofluorescence staining

mLVs from Prox1-tdTomato, Prox1-eGFP, Piezo1TG_LEC and Piezo1dLEC mice were stained using anti-LYVE1 (Angiobio, 11-034), anti-Pdpn (DSHB, AB531893), anti-VEGFR3 (R&D Systems, AF743) and/or anti-Piezo1 (generated by the authors) by following the standard whole-mount immunostaining protocol or frozen section immunostaining protocol12,13,75. Specificity of rabbit anti-Piezo1 generated by the authors was verified using meningeal tissues from wild-type, Piezo1TG_LEC and Piezo1dLEC mice as well as western blotting analyses (Extended Data Fig. 1f and Supplementary Fig. 2). Alexa Fluor fluorescent secondary antibodies (Invitrogen, A21206, A21113, A21207 and A11058) were used to detect protein signal. Fluorescence images were captured with a Zeiss LSM 800 AxioObserver.M2 confocal microscope system or a Zeiss Apotome microscope (AxioZoom V16) controlled by Zen 2.6 software. Randomly selected images were used for statistical analysis.

RNAscope

The RNAscope analysis was performed using an RNAscope Multiplex Fluorescent Reagent kit v2 with TSA Vivid Dyes (Advanced Cell Diagnostics, 323270) and following the manual (Advanced Cell Diagnostics, UM323100) provided by Advanced Cell Diagnostics.

Preparation and pretreatment of samples

Control, Piezo1TG_LEC and Piezo1dLEC mice were i.p. injected with tamoxifen (MP Biomedicals; 50 mg per kg (body weight) twice, 3 days apart) at the age of 6 weeks. After 6 days from the first tamoxifen administration, skulls were collected and fixed in 4% (wt/vol) paraformaldehyde (PFA; Sigma, 252549-1L) at 4 °C for 1 day. The meninges were isolated from the bond and frozen in Tissue-Tek optimum cutting temperature medium (VWR, 25608-930). Frozen blocks were sectioned by cutting 10-μm-thick sections to mount on Superfrost Plus slides (VWR, 48311-703) before air drying for 2 h at –20 °C. The slides were washed with PBS, baked for 30 min at 60 °C in a HybEZ II Hybridization System (Advanced Cell Diagnostics, 321721) and postfixed with 4% (wt/vol) PFA (Sigma, 252549-1L) at 4 °C for 15 min. The tissues were serially dehydrated with ethanol, and RNAscope hydrogen peroxide was applied to the samples for 10 min at room temperature before the samples were boiled in RNAscope 1× Target Retrieval Reagent. RNAscope Protease III was added to the samples after creating a barrier around the tissue section on the slide with an ImmEdge hydrophobic barrier pen (Vector Laboratory, H-4000). The samples were then incubated for 30 min at 40 °C in a HybEZ II Hybridization System.

RNAscope fluorescent assay and imaging

The RNAscope probes Mm-Piezo1 (Advanced Cell Diagnostics, 400181) and Mm-Prox1-C2 (Advanced Cell Diagnostics, 488591-C2) were mixed, added to the samples and incubated for 2 h at 40 °C in a HybEZ II Hybridization System. RNAscope 3-plex Positive Control Probe (Advanced Cell Diagnostics, 320881) and RNAscope 3-plex Negative Control Probe (Advanced Cell Diagnostics, 320871) were used as the controls for the RNAscope reactions, respectively. RNAscope Multiplex FL v2 AMP1 was applied to the samples and hybridized by incubating for 30 min at 40 °C in a HybEZ II Hybridization System. In the same way, RNAscope Multiplex FL v2 AMP2 and AMP3 were hybridized sequentially. To develop horseradish peroxidase (HRP) channel signal for Piezo1 mRNA, RNAscope Multiplex FL v2 HRP-C1 was added to the samples before incubation for 15 min at 40 °C in a HybEZ II Hybridization System. TSA Vivid Fluorophore 520 was then applied and incubated for 30 min at 40 °C in a HybEZ II Hybridization System. The reaction was terminated by adding RNAscope Multiplex FL v2 HRP blocker. Similarly, RNAscope Multiplex FL v2 HRP-C1 and TSA Vivid Fluorophore 570 were applied to develop HRP channel signal for Prox1 mRNA. To label the nuclei, DAPI was added to the samples before incubation for 30 s at room temperature. ProLong Gold Antifade Mountant (Thermo Fisher Scientific, P36930) was added on the slides before a coverslip was placed over the tissue. Fluorescence images were captured with a Zeiss Apotome microscope (AxioZoom V16). For statistical analysis, 16 of the images acquired from three to five mice were randomly selected.

VEGF-C and VEGF-D ELISA

Adult wild-type mice were i.p. injected with vehicle or Yoda1 (213 μg per kg (body weight)) at 18 h and 90 min before collecting the brains. Piezo1TG_LEC mice were i.p. injected with tamoxifen (50 mg per kg (body weight) twice, 3 days apart), and the brains were collected after 6 days from the first injection. Brains were homogenized in PBS (1 ml PBS per 100 mg of brain tissue), and whole-brain extracts were prepared. Concentrations of VEGF-C and VEGF-D in the brain extracts were quantified using a mouse VEGF-C ELISA kit and a mouse VEGF-D ELISA kit (CUSABIO, CSB-E07361m and CSB-E07357m, respectively).

i.c.m. injection and imaging

Tracers were injected into the cisterna magna as described previously28,76. Mice were anesthetized with isoflurane (Kent Scientific Corporation), and the head was tightly placed with the head adaptors of a stereotaxic instrument (World Precision Instruments, 505213). A surgical incision was made to separate the subcutaneous tissue and muscles of the nuchal. Mice were placed so that the head formed a nearly 135° angle with the body to expose the dura mater on the cisterna magna. The following fluorescent tracers were used: ALB-Red (0.5 mg ml–1, Molecular Probes, A13101), OVA-Green (0.5 mg ml–1, Molecular Probes, 034781) and ICG (2.5 mg ml–1, MP Biomedicals, 155020). Tracers were loaded in a syringe (World Precision Instruments, NANOFIL) that was connected with a 33-gauge needle (World Precision Instruments, NF33BV-2) and Microinjection Syringe Pump system (World Precision Instruments, UMP3T-1) and injected (3.5 μl) into the cisterna magna at a rate of 2.5 µl min–1. The incision was immediately sutured. For ICG time-lapse imaging, the tracer was i.c.m. injected as described earlier into Prox1-eGFP mice. The mandibular skin was surgically removed for imaging under a stereomicroscope. The green fluorescence images were first taken to reveal the location of LNs, and near-infrared fluorescence images were obtained every 5 min to record the arrival of ICG to the mandibular LNs. For the brain tracer drainage experiments, cervical and mandibular LNs were surgically collected and imaged for the fluorescent tracer signals at 45–60 min after injection. Images were captured with a Leica stereomicroscope (Leica, M165 FC) controlled by Leica Application Suite X 3.7.6 software, and tracer intensity was measured using ImageJ software (National Institutes of Health). The relative amount of outflowed tracer was quantified by combining the tracer intensities in LNs on both sides.

Mouse hydrocephalus model

The mouse hydrocephalus model was induced as previously described76. A sterile solution of 15% kaolin (suspended in PBS; Sigma-Aldrich, 795453) was loaded in a syringe (Hamilton, 80201) connected with a sterile 31-gauge needle. Under a dissection microscope, the needle was moved into the cisterna magna 3–5 mm deep from the skin, and 10 μl of kaolin was injected using a Microinjection Syringe Pump system (World Precision Instruments, UMP3T-1). The muscles and skin incision were sutured (Redilene, P8698-SP), and mice were monitored over several days for recovery and health.

ICP measurement

ICP measurements were performed as previously described77 with minor modifications. Mice were anesthetized, and an incision was made as described in i.c.m. injection and imaging. A sterile 31-gauge needle connected with a pressure monitor (SYS-BP1, World Precision Instruments) was inserted into the cisterna magna under a dissection microscope. The needle was held for 1 min until an ICP value was obtained.

MRI

Mice were anesthetized using 2% isoflurane in 90% oxygen delivered through a precision vaporizer (Kent Scientific Corporation) and transferred to a 7-T, 24-cm-bore horizontal magnetic resonance scanner (MR Solutions). Two-dimensional fast spin echo (FSE) T2-weighted magnetic resonance images were acquired in coronal and transverse orientations to identify anatomy. The transverse FSE T2-weighted imaging parameters were repetition time = 4,500 ms, echo time = 45 ms, number of averages = 3, echo train length = 7, slice thickness = 0.40 mm, field of view = 16 mm × 16 mm, matrix size = 256 × 238, in-plane resolution = 0.0625 × 0.0672 mm2 per pixel and number of slices = 32. The coronal FSE T2-weighted images had the following parameters: repetition time = 4,500 ms, echo time = 45 ms, number of averages = 3, echo train length = 7, slice thickness = 0.40 mm, field of view = 18 mm × 18 mm, matrix size = 238 × 256, in-plane resolution = 0.0756 × 0.0703 mm2 per pixel and number of slices = 32. Vital signs (respiration in breaths per min) were continuously monitored using a pneumatic pillow (SAII) placed over the abdominal region of the mouse. Body temperature was monitored using a rectal temperature probe (SAII) and maintained at 37 °C using a heated animal holder with a temperature controller unit (Minerve). Brain ventricle volumes were analyzed for all magnetic resonance images by 3D rendering using Multi-image Analysis GUI (Mango)78.

Transcardiac perfusion

Transcardiac perfusion was performed following a previously described method79. Hydrocephalus was induced by i.c.m. injection of kaolin 5 days before transcardiac perfusion. Mice were anesthetized with a combination of ketamine (100 mg per kg (body weight), i.p.) and xylazine (20 mg ml–1, i.p.), and the heart was fully exposed by incision of the skin, diaphragm and ribs. An infusion 25-gauge needle (Terumo, SV25BLK) was inserted into the left ventricle, and a small incision was made in the right atrium. PBS containing heparin (10 U ml–1, final) was infused into the heart at a constant speed of 150 μl s–1 until the fluid outflow was clear of blood, and the liver turned pale. Importantly, our initial examination did not find significant differences in brain fluid content with versus without transcardiac perfusion with PBS before brain collection (Supplementary Fig. 8). We thus did not routinely perform transcardiac perfusion for brain fluid content measurement.

Brain fluid content measurement

Brain fluid content was measured following a previously described method80 with minor modifications. The skin and skull were removed after decapitation of the mouse. The brain was isolated and weighed immediately (wet weight). For dehydration, the brain was placed on a hot plate at 110 °C for 24 h and reweighed (dry weight). Fluid content in the brain was calculated according to the following equation:

$$}}\,}}\,\left( \% \right)=\frac}}\,}}-}}\,}}}}}\,}}}\,\times\,100$$

Lymphatic drainage test in PDMS chips

A lymphatics-on-chip was microfabricated as previously described50,81,82,83. Briefly, the PDMS chip was assembled by bonding a PDMS chip on top of a glass coverslip. After treatment with poly-l-lysine and glutaraldehyde, the PDMS chip was injected with 2.5 mg ml–1 rat tail collagen 1 (Corning) to surround two 250-μm acupuncture needles (Hwato). After the collagen polymerized, acupuncture needles were removed, and primary human dLECs were seeded into one of the channels, while the other channel was left as an acellular channel. To measure lymphatic drainage in the lymphatics-on-chip, 300 μl and 20 μl of cell culture medium was added to the acellular and lymphatic channels, respectively. The volume of medium in the two channels was measured after 12 h, and the drainage score was calculated according to the following equation:

$$}}=\frac_}}_1}}_}}_0}+_}}_0})/2},$$

where \(_}}_0}\) and \(_}}_1}\) are the volumes of medium in the lymphatic channel before and after 12 h, respectively, and \(_}}_0}\) is the volume in the acellular channel before 12 h. After drainage score measurement, the lymphatics-on-chips were stained as previously described82. Briefly, the chips were fixed with 4% PFA, permeated with 0.3% PBST (0.3% Triton X-100 in phosphate-buffered saline) and blocked with 3% bovine serum albumin overnight at 4 °C on a shaking platform. Anti-VE-cadherin (F11, Santa Cruz Biotechnology; 1:100) in blocking buffer was added and incubated overnight at 4 °C. Primary antibodies were washed overnight using PBS at 4 °C. Secondary antibody (Invitrogen, 1:500), phalloidin (actin, 1:200) and DAPI (Millipore Sigma, 1:500) were subsequently incubated in blocking buffer overnight at 4 °C in the dark. The chips were washed to remove fluorescent background before confocal microscopy. Confocal images were acquired with a Leica SP8 confocal microscope. Image quantification was performed with ImageJ software84.

Excised lymphatic collector physiology assays

Supplemental Videos 1–6 show changes in the contractile parameters of an excised collecting lymphatic after exposure to vehicle, Yoda1 and Yoda1 + l-NAME.

Vessel isolation and cannulation

Mice were anesthetized with a combination of ketamine (100 mg per kg (body weight), i.p.) and inactin (100 mg per kg (body weight), i.p.) and placed on a plexiglass board next to a pillar made from Sylgard 184 (Dow Corning). The inguinal–axillary lymphatic vessel was exposed by making a 2-cm-long midline incision extended dorsally with two more incisions at the shoulder and groin to create a rectangular skin flap. The skin flap was gently pulled away from the body and pinned into the Sylgard pillar. Connective tissue was gently dissected to enable the skin to be pulled further, revealing both the inguinal and axillary LNs. The inguinal–axillary vessel was identified as the collecting lymphatic vessel connecting these two LNs, as previously reported85. Using fine microscissors, the axillary collecting lymphatic vessel was carefully excised along with adipose and connective tissues and transferred to a custom chamber coated with Sylgard 170 (Dow Corning) and containing Krebs buffer. The axillary vessel was then pinned to the Sylgard layer with 40-µm stainless steel wire. Adipose and connective tissues were removed from the vessel by careful microdissection at room temperature. The cleaned axillary lymphatic vessel was transferred to a 3-ml acrylic chamber filled with Krebs buffer that was mounted onto a custom breadboard (Thorlabs) onto which a pair of micromanipulators (LBM-7, Scientifica) was mounted. The micromanipulators were used to position two micropipettes (80 µm) at each end of the 3-ml chamber. The axillary lymphatic vessel was then cannulated on each micropipette and secured with a single knot of unbraided 4-0 silk suture.

Pressure system and diameter tracking

Once cannulated, the isolated vessel board was transferred to an inverted microscope (Observer Z1, Zeiss) and connected to a custom pressure control system. The pressure was controlled using LabView software that ran a pressure pump connected to the vessel via polyethylene tubing. The pressure was constantly measured at each end of the vessel by low-pressure transducers. The vessel was allowed to equilibrate to 37 °C for 1 h at a pressure of 2–3 cmH2O, while the temperature was controlled by a heat exchanger pumping warm water into the water jacket of the 3-ml acrylic chamber. The pressure transducer signals were amplified and recorded by a custom-written LabView program86. The same LabView program tracked and recorded the inner diameter of the vessels from a video image obtained by a FireWire camera (Basler) at 30 Hz.

Isolated vessel protocol

Isolated and cannulated collecting lymphatic vessels from the axillary region of mice were subjected to the following protocol to evaluate the actions of Yoda1 on lymphatic endothelium. To enable pairwise comparisons, collecting lymphatic vessels were allowed to contract spontaneously in the same vehicle used for diluting Yoda1 for vehicle control while the diameter was recorded. Pressure at each end of the vessel stepped from 0.5 cmH2O to 1, 2, 3, 5, 7 and 10 cmH2O. At the end of the pressure steps, the pressure was lowered to 3 cmH2O before adding Yoda1. Yoda1 (1 µM) was then added to the superfusion bath solution outside the vessel and incubated for 30 min before repeating the same pressure steps from 0.5 to 10 cmH2O. Finally, l-NAME (100 µM) and Yoda1 were combined and added to the bath for a 30-min incubation period at 3 cmH2O. After the incubation, the same pressure steps were repeated. At the end of all experiments, the superfusion buffer was replaced with a Ca2+-free Krebs buffer for 30 min. The intraluminal pressure was then lowered to 0.1 cmH2O, and the pressure was stepped to 0.5, 1, 2, 3, 5, 7 and 10 cmH2O to obtain the maximal vessel diameter at each pressure to calculate the basal tone.

Solutions and chemicals

Krebs buffer contained the following components: 141.4 mM NaCl, 4.7 mM KCl, 2 mM CaCl2•2H2O, 1.2 mM MgSO4, 1.2 mM NaH2PO4•H2O, 3 mM NaHCO3, 1.5 mM NaHEPES, 5 mM d-glucose and 0.1% bovine serum albumin (pH 7.4 at 37 °C). The Krebs buffer was sterile filtered and used within 1 week. This buffer was used for all dissection, cannulation and superfusion solutions. During the experiment, the superfusate was gradually supplemented by the continuous addition of Krebs buffer at a rate of 0.5 ml min–1. The Ca2+-free Krebs buffer was identical, except that CaCl2 was replaced with 3 mM EGTA. All chemicals except for bovine serum albumin (Affymetrix) were purchased from Sigma. l-NAME was dissolved in Krebs buffer at 100 mM and stored at –20 °C until the day of use.

Data analysis

After the experiments, separate custom-written LabView programs were used to detect and record the end-diastolic diameter, end-systolic diameter and contraction frequency. Data were copied into a Microsoft Excel file that was used to calculate the following contractile function parameters: amplitude = EDD – ESD and percent tone = [(maxD – EDD)/maxD] × 100, where EDD is the end-diastolic diameter, ESD is the end-systolic diameter, and maxD is the maximal passive diameter at each pressure, which was obtained during the Ca2+-free portion of the experiment.

Physical activity analyses

The open field test was performed following a published protocol87 with minor modifications. Mice were carried to a testing room isolated from sound and unintentional interruptions at least 1 h before the test. Mice were then placed in the center of an opaque white plastic open field chamber (45 cm × 25 cm), and free and uninterrupted movement of each mouse was recorded for 5 min. Using video tracking software (Tracker 6.0.10)88, the movement of the mouse was traced, and total distance, velocity and time spent in the center (27 cm × 5 cm) were quantified. The rotarod test was performed following a published protocol89 with minor modifications. Mice were placed on a rotarod cylinder (Ajanta-96, AJANTA) 25 mm in diameter rotating at 4 rpm. The rotational speed was gradually increased to 20 rpm min–1 until 40 rpm (maximum rotational speed) was reached. The mice were trained twice on the day before the experiment and on the morning of the experimental day to reduce variability between animals. The length of time the mouse continuously walked forward to keep from falling off the rotating rod was measured to determine the ability to maintain balance. The mesh hanging test was performed as previously described. We generated a mesh screen by cutting a 30-cm square of wire mesh (1-cm squares of 1-mm-diameter wire). Mice were placed in the center of the mesh screen, and the screen was rotated to an inverted position while starting the stopwatch. The screen was steadily held at the height of 50 cm above a pad until the mouse fell off (up to 3 min).

Statistical analysis

Sample size was predicted and chosen by power analysis using G*Power (hhu). Parameters for the analysis were effect size (Cohen’s d or Cohen’s f) = 0.8–2, α error probability (significance level) = 0.05, power (1 – β error probability) = 0.8 and allocation ratio N2/N1 = 1. Effect sizes were calculated based on pilot experiments. The experiments were randomized. Data collection and analysis were not performed blind to the conditions of the experiment, but no animals or data points were excluded. Normality of the data was tested by Shapiro–Wilk or Kolmogorov–Smirnov test, and homoscedasticity was tested by Levene’s test or Box’s M-test. Sphericity was tested by Mauchly’s sphericity test, and the F critical value was adjusted by Greenhouse–Geisser correction. Parametric or nonparametric statistics were selected based on the result of the tests. A two-tailed, unpaired Student’s t-test or Mann–Whitney U-test was used to determine the statistical significance between two groups. A one-way ANOVA or Kruskal–Wallis H-test was performed to compare differences between multiple groups. A two-way repeated measures ANOVA was used to compare the time or pressure series data between two or three groups. Post hoc tests following one-way ANOVAs and two-way repeated measures ANOVAs were performed by Tukey’s multiple comparison test, Bonferroni’s multiple comparison test or Dunnett’s multiple comparison test, and statistical significance (P) was set at less than 0.05. The Bonferroni correction method was used as a post hoc test for the Kruskal–Wallis H-test. Statistical significance of the Bonferroni correction method was determined based on the number of groups and is indicated in the figure legends. Statistical analyses were performed using SPSS 12 (IBM) and GraphPad Prism 8 (GraphPad Software).

Reporting summary

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