A novel GABA_A receptor ligand MIDD0301 with limited blood-brain barrier penetration relaxes airway smooth muscle ex vivo and in vivo

Abstract

Airway smooth muscle (ASM) cells express GABA A receptors (GABA_ARs), and previous reports have demonstrated that GABA_AR activators relax ASM. However, given the activity of GABA_ARs in central nervous system inhibitory neurotransmission, concern exists that these activators may lead to undesirable sedation. MIDD0301 is a novel imidazobenzodiazepine and positive allosteric modulator of the GABA_AR with limited brain distribution, thus eliminating the potential for sedation. Here, we demonstrate that MIDD0301 relaxes histamine-contracted guinea pig (P < 0.05, n = 6–9) and human (P < 0.05, n = 6–10) tracheal smooth muscle ex vivo in organ bath experiments, dilates mouse peripheral airways ex vivo in precision-cut lung-slice experiments (P < 0.001, n = 16 airways from three mice), and alleviates bronchoconstriction in vivo in mice, as assessed by the forced-oscillation technique (P < 0.05, n = 6 mice). Only trace concentrations of the compound were detected in the brains of mice after inhalation of nebulized 5 mM MIDD0301. Given its favorable pharmacokinetic properties and demonstrated ability to relax ASM in a number of clinically relevant experimental paradigms, MIDD0301 is a promising drug candidate for bronchoconstrictive diseases, such as asthma.

INTRODUCTION

Asthma is a common disorder associated with variable airflow obstruction, airway smooth muscle (ASM) hyperresponsiveness, and lung inflammation that continues to be a significant contributor to morbidity and mortality worldwide. In the United States, the Centers for Disease Control reports that the prevalence of asthma is 8.3%. This number is even higher in certain minority groups and for those living below the poverty level (17). For many years, inhaled corticosteroids and β-agonists have been the mainstays of therapy, yet nearly 40% of asthmatics have inadequate control of their symptoms (4, 19). Emerging biologic therapies are limited to small subsets of carefully phenotyped patients, are expensive, and require administration in a health care setting (18, 21, 22, 25). Moreover, no new therapies have been introduced for decades that directly target ASM relaxation.

Several years ago, our group first described the expression of GABA_A receptors (GABA_ARs) on ASM cells (20). These receptors, which are heteropentameric ligand-gated chloride channels, best known for their role in inhibitory neurotransmission, are the site of action of many anesthetics and sedative medications, including benzodiazepines. Independent of these actions in the central nervous system (CNS), we (8, 10, 31) and others (23) have shown that pharmacologic activators of GABA_ARs lead to direct ASM relaxation. However, given the likely CNS effects of such drugs, the use of GABA_AR activators to treat bronchoconstriction has the potential to cause undesired sedation.

In an attempt to circumvent these sedative side effects, we have previously developed and studied novel benzodiazepine derivatives that are selective for GABA_AR subtypes that are predominately expressed on ASM (α4 and α5 subunit-containing receptors) and not the GABA_AR variants thought to mediate primarily sedation in the CNS (those containing α1– α3 subunits) (10, 31). Recently, we have developed a tissue-selective imidazobenzodiazepine (GABA_AR-positive allosteric modulator), MIDD0301 (6), thus eliminating the potential for CNS effects.

In the present study, we demonstrate that MIDD0301 relaxes ASM in a number of clinically relevant experimental paradigms. These include relaxation of human and guinea pig tracheal ASM in organ bath experiments, dilation of mouse peripheral airways in precision-cut lung-slice (PCLS) experiments, and alleviation of in vivo bronchoconstriction in mouse airway resistance experiments.

MATERIALS AND METHODS

Animals, tissues, and reagents.

All animal experimental protocols were approved by the Columbia University Medical Center Institutional Animal Care and Use Committee. Male Hartley guinea pigs (∼400 g) or 8- to 10-wk-old A/J mice (Jackson, Bar Harbor, ME) were used for animal experiments. Human ASM samples were obtained from the tracheal trimmings of de-identified, healthy human lung-transplantation donor organs at the time of surgery. These tissues were deemed to be nonhuman research material by the Columbia University Medical Center Institutional Review Board.

MIDD0301 {(R)-8-bromo-6-(2-fluorophenyl)-4-methyl-4H-benzo[f]imidazo[1,5-a][1,4]diazepine-3-carboxylic acid} was synthesized in the laboratory of Dr. James M. Cook at the University of Wisconsin, Milwaukee. Identity was confirmed by ¹H and ¹³C NMR and high-resolution mass spectrometry (MS). Purity was determined by HPLC to be >98%. Tetrodotoxin was obtained from Calbiochem (San Diego, CA). MK571 and capsaicin were obtained from Tocris Bioscience (Bristol, UK), and all other reagents were obtained from Sigma-Aldrich (St. Louis, MO).

Ex vivo guinea pig tracheal ring organ bath experiments.

Guinea pig organ bath experiments were performed as previously described (30). Briefly, guinea pigs were deeply anesthetized with pentobarbital sodium (100 mg/kg ip), and their tracheas were removed. Under a dissecting microscope, each trachea was transected into eight closed segments consisting of two cartilaginous rings. These segments were subsequently denuded of epithelium with a cotton swab and suspended on two strings in water-jacketed (37°C) organ baths (Radnoti Glass Technology, Monrovia, CA). One string was fixed to a stationary hook, and the other was fixed to a FT-03 force transducer (Grass Telefactor, West Warwick, RI), coupled to a computer via Biopac hardware and AcqKnowledge 7.3.3 software (Biopac Systems, Goleta, CA). The baths contained Krebs-Henseleit (KH) buffer, consisting of (in mM) 118 NaCl, 5.6 KCl, 0.5 CaCl₂, 0.24 MgSO₄, 1.3 NaH₂PO₄, 25 NaHCO₃, and 5.6 glucose (pH 7.4) and were bubbled with 95% O₂-5% CO₂. Tissues were equilibrated at 1 g isotonic force for 1 h with fresh KH buffer changes every 15 min before experiments.

All tracheal rings were exposed to 10 µM capsaicin to deplete nonadrenergic, noncholinergic nerves to avoid confounding neural contribution to subsequent ASM contraction. The rings were then subjected to two cycles of increasing cumulative concentrations of acetylcholine (0.1 μM–0.1 mM) to determine the acetylcholine EC₅₀ concentration for each ring. Tissues with similar acetylcholine EC₅₀ values were randomly assigned to treatment groups within individual experiments. Following six KH buffer changes, tissues were allowed to stabilize at a resting tension of 1.0 g. Tetrodotoxin (1 µM) (neural transmission inhibitor), 10 µM indomethacin (inhibitor of prostaglandin synthesis), and 10 µM pyrilamine (antihistamine) were also added to the bath to avoid confounders.

Each ring was then contracted with 10 µM histamine. After a peak contraction was reached, the rings were exposed to concentration ranges of MIDD0301 or vehicle control (0.1% DMSO) during continuous contractile-force monitoring. The degree of relaxation of the tracheal rings at 15, 30, and 45 min was compared between MIDD0301 and vehicle control groups.

Ex vivo human tracheal ASM strip organ bath experiments.

ASM from human trachea was dissected into strips, denuded of epithelium, and suspended in organ baths, as described above (although resting tension was set at 1.5 g). The acetylcholine EC₅₀ for each strip was determined, and the strips were assigned to treatment groups as described above. Tetrodotoxin (1 µM), 10 µM indomethacin, and 10 µM MK571 (leukotriene D4 receptor inverse agonist) were added to the baths to avoid potential confounding contributions to contraction. The strips were then contracted with 10 µM histamine and after stable contraction was achieved, exposed to 100 µM MIDD0301 or vehicle control (0.1% DMSO). The degrees of relaxation at 15, 30, and 45 min were compared between MIDD0301 and vehicle control groups.

Ex vivo mouse PCLS experiments.

PCLS peripheral airway diameter studies were performed as previously described (3, 10, 24, 30). Briefly, male A/J mice were euthanized, their thoracic cavities opened, and their lungs inflated with warm 2% agarose. After the agarose was allowed to cool at 4°C for 20 min, the lungs were removed and held in ice-cold Hanks’ balanced salt solution (HBSS) for 15 min. The lung lobes were separated and then fully covered with 6% gelatin to form a block. The block was cut into serial 130 μm sections with a vibratome (VF-300; Precisionary Instruments, Greenville, NC). Lung slices, containing small peripheral airways, were incubated overnight in low-glucose DMEM, supplemented with 1× antibiotic solution containing l-glutamine, penicillin, and streptomycin (Thermo Fisher Scientific, Waltham, MA) at 37°C and 10% CO₂.

Lung slices containing airways with a lumen diameter of 100–300 μm with active (beating), ciliated epithelial cells were used. Lung slices were mounted in a custom-made perfusion chamber, and peripheral airways were visualized using a ×10 objective. The slices were superfused with HBSS, supplemented with 20 mM HEPES, and adjusted to pH 7.4 during imaging at room temperature. Digital images were recorded at 0.5 Hz, using a charge-coupled device camera (KP-M1A; Hitachi, Tokyo, Japan), a frame grabber (Picolo; Euresys, San Juan Capistrano, CA), and image-acquisition software (Video Savant; IO Industries, London, ON, Canada). Airway contraction was induced by superfusion of 100 nM methacholine (MCh) in HBSS. Following attainment of stable airway contraction, the PCLS was exposed to 100 µM MIDD0301 in the continuous presence of MCh. The lumenal area was calculated from each recorded image using custom-written macros and normalized to the area before stimulation with MCh.

In vivo mouse airway resistance.

In vivo airway resistances were assessed in A/J mice using a flexiVent (Scireq, Montreal, QC, Canada) with an FX1 module and an in-line nebulizer, as previously described (26, 31). Briefly, the mice were anesthetized with pentobarbital sodium (50 mg/kg ip), paralyzed with succinylcholine (10 mg/kg ip), and mechanically ventilated via a tracheostomy (tidal volume, 10 mg/kg; 150 breaths/min; positive end-expiratory pressure, 3 mmHg). The mice received nebulized 5 mM MIDD0301 (10 s nebulization, 50% duty cycle; ~25 µl delivered) or vehicle (25% ethanol in PBS), 10 min before measurement of the central lung resistances (R_n), by the forced oscillatory technique during a graded, nebulized MCh challenge (0–25 mg/ml). Electrocardiography and temperature monitoring were performed throughout the experiment. Resistance values for each mouse at each MCh dose represent an average of three measurements.

Following in vivo airway resistance testing, brain, lung, and serum samples were collected from the MIDD0301-treated mice (10 and 30 min after nebulization of MIDD0301). The brain and lung samples were immediately snap frozen with liquid nitrogen, and the serum samples were frozen at −20°C. Tissue concentrations of MIDD0301 were subsequently determined by liquid chromatography-MS/MS, as previously described (6).

Statistical analyses.

Each experimental procedure included internal controls. For organ bath and in vivo airway resistance experiments, two-way ANOVAs with repeated measures and Bonferroni’s post-test comparisons were used. For PCLS airway diameter comparisons, Student’s t-test was used. Analyses were performed using Prism 4 software (GraphPad Software, La Jolla, CA). Data are presented as means ± SE, and P < 0.05 was considered significant.

RESULTS

MIDD0301 relaxes histamine-contracted guinea pig and human tracheal ASM ex vivo.

Guinea pig tracheal rings suspended in organ baths were contracted with 10 µM histamine. After a plateau in increased muscle force contraction was achieved, the rings were exposed once to MIDD0301 (10–100 µM). After 15 min of exposure, MIDD0301 led to a significant dose-dependent ASM relaxation at 25, 50, and 100 µM compared with DMSO control (Fig. 1; two-way ANOVA with Bonferroni’s post hoc test, P < 0.05, n = 6–9 rings from six guinea pigs). All concentrations of the MIDD0301 tested led to significant relaxation compared with the vehicle control by 60 min (data not shown; P < 0.05, n = 6–9 rings from six guinea pigs).

Fig. 1.

MIDD0301 relaxes guinea pig tracheal muscle ex vivo. A: representative tracing of guinea pig tracheal ring contractile force versus time in organ bath experiments. The rings were exposed to 10 µM histamine (His), which led to a rapid increase in contractile force. After peak contraction was achieved, MIDD030 or the DMSO vehicle control was added to the baths (at the point marked Compound). B: MIDD0301 led to a dose-dependent decrease in tracheal ring contractile force, with 25, 50, and 100 µM significantly decreased compared with DMSO control (0 µM MIDD0301) after 15 min of drug exposure. ANOVA with Bonferroni’s post hoc tests, *P < 0.05, **P < 0.01, n = 6–9 rings from 6 guinea pigs.

Human tracheal ASM strips were contracted with 10 µM histamine in similar organ bath experiments. MIDD0301 (100 µM) also led to significant relaxation of human ASM, with a mean relaxation to less than baseline tension by 30 min (Fig. 2; ANOVA with Bonferroni’s post hoc test, P < 0.05, n = 6–10 strips from five human donors).

Fig. 2.

MIDD0301 relaxes human tracheal smooth muscle strips ex vivo. A: representative tracing of human airway smooth muscle (ASM) strip contractile force versus time in organ bath experiments. The strips were exposed to 10 µM histamine (His), which led to a rapid increase in contractile force. After stable contraction was achieved, 100 µM MIDD0301 or the DMSO vehicle control was added to the baths (at the point marked Compound). B: MIDD0301 100 µM led to a significant decrease in ASM contractile force by 30 min of drug exposure. Two-way ANOVA with Bonferroni’s post hoc tests, *P < 0.05, **P < 0.01 compared with vehicle control, n = 6–10 strips from 5 human donors.

MIDD0301 relaxes MCh-contracted peripheral airways in mouse PCLS.

In A/J mouse PCLS experiments, 100 nM MCh caused a 34.0 ± 3.6% decrease in the peripheral airway luminal area. The subsequent addition of MIDD0301 relaxed this contraction by a mean of 33.4 ± 3.5% (Fig. 3; P < 0.001, n = 16 airways from three mice). Vehicle control (0.01% DMSO) led to no change in airway area (data not shown).

Fig. 3.

MIDD0301 relaxes mouse peripheral airways in precision-cut lung slices ex vivo. The peripheral airways of A/J mouse lung slices were contracted with 100 nM methacholine (MCh) and subsequently exposed to 100 µM MIDD0301 (in the continued presence of MCh) during video microscopy imaging of airway luminal area. A: representative still photos from video recordings at the time points indicated in B. B: representative tracing of airway luminal area over time. C: MIDD0301 led to a significant decrease in the degree of MCh-mediated airway contraction by Student’s t-test, ***P < 0.001, n = 16 airways from 3 mice.

Inhaled MIDD0301 reduces MCh-induced bronchoconstriction in vivo in mice with hyperresponsive airways.

Nebulized MIDD0301 (5 mM, 10 s nebulization, 50% duty cycle) inhibited MCh-induced increases in R_n during MCh challenge tests in A/J mice compared with vehicle control (Fig. 4; two-way ANOVA with Bonferroni’s post hoc test, *P < 0.05, n = 6 mice). There was no significant difference in baseline R_n values between groups.

Fig. 4.

MIDD0301 alleviates bronchoconstriction in mice in vivo. Before an inhaled, graded methacholine (MCh) challenge, anesthetized and mechanically ventilated A/J mice received nebulized MIDD0301 (5 mM, 10 s nebulization, 50% duty cycle) or vehicle control (25% ethanol in PBS). Mice receiving MIDD0301 had a significantly diminished increase in central airway resistance (R_n) in response to a MCh challenge (two-way ANOVA with repeated measures, *P < 0.05, n = 6 mice).

Concentrations of MIDD0301 in perfused tissues were determined 10 min (Fig. 5A) and 30 min (Fig. 5B) after a 10-s nebulization of 5 mM MIDD0301 in 25% ethanol/PBS. The mean serum concentrations after 10 and 30 min were 86 and 47 ng/g, respectively; the lung mean concentrations were 1,334 and 412 ng/g, respectively; whereas the mean brain concentrations were 58 and 9.6 ng/g in perfused tissue, respectively. Thus inhaled MIDD0301 rapidly reached a high concentration in the lung, whereas the concentrations in serum and brain were significantly lower. The reduction of ~70% of the lung concentration of MIDD0301 over 20 min demonstrated fast lung clearance, whereas the serum concentration of MIDD0301 was reduced 45% during the same time interval.

Fig. 5.

Tissue concentrations of MIDD0301 in mice after inhalation. MIDD0301 was administered to anesthetized, mechanically ventilated mice by nebulization using a flexiVent (5 mM, 10 s nebulization, 50% duty cycle, 25% ethanol/PBS vehicle). Serum, lung, and brain tissues were collected 10 min (A) and 30 min (B) after inhalation (5 mM MIDD0301 in 25% ethanol/PBS for 10 s) via nebulizer. Values are given as means ± SE (n = 6).

DISCUSSION

In the current study, we demonstrate that MIDD0301, a novel, positive allosteric modulator of the GABA_AR that exhibits very low concentrations in the CNS after systemic administration, relaxes ASM from several species in a variety of clinically relevant experimental paradigms. Previous work showed that MIDD0301 relaxed substance P-contracted guinea pig tracheal rings ex vivo and reduced airway reactivity in spontaneously breathing ova-sensitized and ova-challenged mice (as measured by whole-body plethysmography) (6). The current work significantly advances these findings by demonstrating the following: 1) relaxation of ASM contracted with an alternative clinically relevant agonist (histamine), 2) relaxation of human ASM, 3) relaxation of mouse peripheral ASM in PCLS, and 4) alleviation of in vivo mouse bronchoconstriction using the forced oscillation technique, a widely accepted measure of in vivo airway resistance.

Previous studies have demonstrated significant levels of MIDD0301 in the blood [maximum concentration (C_max): 8.24 μg/g; half-life (t_1/2): 836 min] and lungs (C_max: 4.39 μg/g; t_1/2: 234 min) of mice after oral administration of 25 mg/kg. However, concentrations reached a mean C_max of only 0.48 μg/g in nonperfused brains (6). Consistent with this, MIDD0301 led to no sensorimotor impairment in mice on rotarod experiments with oral doses of 1,000 mg/kg. In the current study, we administered MIDD0301 by inhalation before the measurement of in vivo airway resistance and measured serum, lung, and brain concentrations of MIDD0301 in each mice. As expected, high concentrations of MIDD0301 were detected in the perfused mouse lungs shortly after nebulization. Similar to other inhaled drugs, serum concentrations of MIDD0301 were low (6–11%) compared with its lung concentrations (16). The average MIDD0301 lung concentration changed from 1,334 ng/g (~3.2 μM) at 10 min to 412 ng/g (~1 μM) at 30 min, which is significantly lower than the buffer concentrations used in this study to relax ASM ex vivo (Figs. 1–3). However, we reported that only 10% of the organ bath compound concentration penetrated into the tracheal tissue (7). Thus it can be expected that at least for the period of 30 min, the concentration of MIDD0301, when nebulized as a 5-mM solution for 10 s, was sufficient to mediate a bronchorelaxing effect. This duration of action is similar to short-acting bronchodilators, such albuterol, terbutaline, and isoprenaline (28). In contrast to almost all other bronchodilators, MIDD0301 is also effective orally (6). MIDD0301 has a significantly longer t_1/2 in the lung when given orally and is therefore compared with long-acting β agonists, such as salmeterol and formoterol (1). However, long-acting β-agonists are only effective when inhaled and due to long-term adverse effects, can only be used in conjunction with inhaled corticosteroids (2). Thus MIDD0301 is a unique asthma drug candidate that is short acting when inhaled and long acting when administrated orally.

Previous reports have examined the effects of GABA_AR activators on ASM cell-signaling processes in an attempt to determine their mechanism of relaxation. GABA_AR-mediated currents have been demonstrated in human ASM cells (9), and several chemically distinct GABA_AR ligands relax ASM, including benzodiazepines (10, 23, 31), propofol (8), and muscimol (11). Furthermore, classic GABA_AR antagonists (i.e., gabazine and picrotoxin) have been shown to block these effects in several studies (11, 20), and gabazine enhances acetylcholine-mediated ASM contraction directly (8). This evidence suggests that the mechanism of relaxation of GABA_AR activators is consistent with “on-target” (GABA_AR-mediated) effects. Given that the GABA_AR is a chloride channel, much of the previous work has hypothesized that changes in membrane potential are altering procontractile cell-signaling processes. For example, GABA_AR activators have been shown to inhibit ASM cell calcium signaling events that occur upon exposure to classic, Gq-coupled contractile agonists (i.e., MCh, bradykinin). These include intracellular calcium oscillations measured in PCLS (10) and population-based calcium responses measured in cultured primary human ASM cells (31).

Of note, GABA_AR-mediated chloride currents are thought to be depolarizing in resting ASM cells, unlike mature neurons, due to their relatively high internal chloride concentration (chloride efflux upon channel opening). At first glance, it seems counterintuitive that a depolarizing stimulus would be associated with decreased calcium responses. Although it is beyond the scope of this manuscript, several hypotheses have been suggested to reconcile this apparent contradiction. First, upon acetylcholine-mediated contraction, the membrane of ASM cells depolarizes (29), potentially beyond the reversal potential for chloride. Thus in the contracted state, the flux of chloride via the GABA_AR may reverse, resulting in a relative hyperpolarization and thus inhibit calcium influx via voltage-gated channels (8). Second, previous reports have suggested a GABA_AR activator may inhibit ASM cell store-operated calcium entry (SOCE) (10). SOCE is extracellular calcium entry triggered by low sarcoplasmic/endoplasmic reticulum calcium concentrations to refill empty intracellular calcium stores. This process, which is mediated by the stromal interaction molecule-Orai1 complex, is key in the maintenance of ASM cell intercellular calcium oscillations and contraction (5). Although not voltage gated, calcium entry via the Orai1 pore is inhibited by membrane depolarization, secondary to changes in the calcium electrochemical gradient (12). Furthermore, it has been demonstrated that calcium entry via voltage-gated calcium channels does not contribute to ASM calcium oscillations in the peripheral airways of lung slices (24). Therefore, if ASM GABA_AR activation is in fact depolarizing, then it may be inhibiting calcium entry and calcium oscillations by the inhibition of SOCE. Alternatively, in addition to chloride, GABA_ARs are permeant to bicarbonate. Thus activation of GABA_AR decreases intracellular pH (14, 15). The pore region of Orai1 is highly basic (a glutamate serves as selectivity filter) (13), and calcium conductance through Orai1 is significantly hindered by decreases in intra- and extracellular pH (27). Therefore, GABA_AR activation may inhibit SOCE, not by depolarizing membrane potential but by intracellular acidification. Further studies on these potential mechanisms are underway.

A recent study by Pera et al. (23) suggested an interesting potential “off-target” mechanism for benzodiazepine-mediated ASM relaxation. The authors report that multiple benzodiazepines activate the proton-sensing ovarian cancer G protein-coupled receptor (OGR1/GPR68) and that the subsequent biased agonistic signaling events are benzodiazepine-variant dependent. For example, OGR1 activation with lorazepam led to Gq, Gi, and Gs signaling events (with a pH dependence), whereas OGR1 activation with sulazepam seemed only to signal via the Gs pathway. Consistent with this, sulazepam increased intracellular cAMP concentration in primary ASM cells and reduced cell stiffness in magnetic twisting cytometry assays, whereas lorazepam did not at pH 7.4. It is not known how these very interesting results would apply to nonbenzodiazepine GABA_AR activators, such as propofol and muscimol.

In summary, the GABA_AR-positive allosteric modulator MIDD0301 relaxes ASM in a number of clinically relevant experimental paradigms. Because this benzodiazepine derivative was not found to cross the blood-brain barrier significantly, following inhalation or oral administration in mice, it offers the potential to mediate bronchodilation without CNS suppression.

GRANTS

This work was supported by National Institutes of Health Grants HL-140102 (to G. T. Yocum), HL-122340 (to C. W. Emala Sr.), HL-118561 (to J. M. Cook and L. A. Arnold), GM-065281 (to C. W. Emala Sr.), MH-096463 (to J. M. Cook and L. A. Arnold), and DA-031090 (to L. A. Arnold); Louis V. Gerstner Jr. Scholars Program (to G. T. Yocum); Lynde and Harry Bradley-Herzfeld Foundation (to L. A. Arnold); Richard and Ethel Herzfeld Foundation (to L. A. Arnold), and University of Wisconsin-Milwaukee Research Foundation (to L. A. Arnold and J. M. Cook).

DISCLOSURES

G. Li, R. Kodali, D. C. Stafford, J. M. Cook, L. A. Arnold, and C. W. Emala have financial interests as inventors on patents covering MIDD0301 and its uses as described in the report. D. C. Stafford, L. A. Arnold, and C. W. Emala have equity interests in Pantherics Incorporated, which has certain intellectual property rights to these patents.

AUTHOR CONTRIBUTIONS

G.T.Y., D.C.S., L.A.A., J.M.C., and C.W.E. conceived and designed research; G.T.Y., J.F.P.-Z., J.D., A.S.K., Y.Z., G.L., M.S.R.R., and R.K. performed experiments; G.T.Y., J.F.P.-Z., J.D., A.S.K., Y.Z., M.S.R.R., R.K., L.A.A., J.M.C., and C.W.E. analyzed data; G.T.Y., J.F.P.-Z., J.D., L.A.A., J.M.C., and C.W.E. interpreted results of experiments; G.T.Y. and J.F.P.-Z. prepared figures; G.T.Y. drafted manuscript; G.T.Y., J.F.P.-Z., J.D., A.S.K., Y.Z., G.L., M.S.R.R., R.K., D.C.S., L.A.A., J.M.C., and C.W.E. edited and revised manuscript; G.T.Y., J.F.P.-Z., J.D., A.S.K., Y.Z., G.L., M.S.R.R., R.K., D.C.S., L.A.A., J.M.C., and C.W.E. approved final version of manuscript.