Hyperpolarized gas diffusion MRI of biphasic lung inflation in short- and long-term emphysema models

Abstract

During lung inflation, airspace dimensions are affected nonlinearly by both alveolar expansion and recruitment, potentially confounding the identification of emphysematous lung by hyperpolarized helium-3 diffusion magnetic resonance imaging (HP MRI). This study aimed to characterize lung inflation over a broad range of inflation volume and pressure values in two different models of emphysema, as well as in normal lungs. Elastase-treated rats (n = 7) and healthy controls (n = 7) were imaged with HP MRI. Gradual inflation was achieved by incremental changes to both inflation volume and airway pressure. The apparent diffusion coefficient (ADC) was measured at each level of inflation and fitted to the corresponding airway pressures as the second-order response equation, with minimizing residue (χ² < 0.001). A biphasic ADC response was detected, with an initial ADC increase followed by a decrease at airway pressures >18 cmH₂O. Discrimination between treated and control rats was optimal when airway pressure was intermediate (between 10 and 11 cmH₂O). Similar findings were confirmed in mice following long-term exposure to cigarette smoke, where optimal discrimination between treated and healthy mice occurred at a similar airway pressure as in the rats. We subsequently explored the evolution of ADC measured at the intermediate inflation level in mice after prolonged smoke exposure and found a significant increase (P < 0.01) in ADC over time. Our results demonstrate that measuring ADC at intermediate inflation enhances the distinction between healthy and diseased lungs, thereby establishing a model that may improve the diagnostic accuracy of future HP gas diffusion studies.

chronic obstructive pulmonary disease (COPD) is characterized by various respiratory ailments, including, but not limited to, emphysema and chronic bronchitis (1). COPD patients typically suffer from “air trapping,” or dynamic overaccumulation and retention of lung gases (29), with consequent airspace enlargement causing impaired gas exchange and hyperinflation of the lungs (17). In cases of emphysema more specifically, degradation of the alveolar walls (19) alters the micromechanical properties of lung tissue and leads to decreased elastic recoil (21).

The most commonly used method for evaluating lung microstructure and micromechanics is computed tomography (CT): CT is readily accessible and relatively noninvasive, and it has the ability to provide local in vivo measurements of tissue density (2, 8, 13, 28, 33). However, it does not provide sufficient resolution to reveal the structure of acinar airways. Over the past decade, hyperpolarized (HP) gas MRI has emerged as a methodology capable of quantitatively describing the diffusion limitation of inhaled gas molecules within the acini and, thereby, indirectly quantifying airspace dimensions (32). This quantification is accomplished by measuring the apparent diffusion coefficient (ADC) of hyperpolarized ³He gas introduced into the lungs during inhalation; the regional microstructure of the lung can then be delineated by mapping ADC, creating a noninvasive index of airspace enlargement (12, 31).

Increased ADC in emphysematous lungs has been demonstrated in both animal and human studies (19, 20, 31). In these studies, ADC is typically acquired only at end inspiration. Yet, ADC values have been shown to vary considerably according to inflation level (4–6, 14). Indeed, ADC’s response to lung inflation is a complex one, as previous studies have demonstrated that it responds to both alveolar distension and alveolar recruitment, which have opposing effects (6, 14). This biphasic behavior is poorly characterized by linear and exponential models of inflation (30). Accurately describing ADC within a broader range of inflation may enhance our ability to discriminate between healthy and diseased lungs. For this purpose, we characterized airspace inflation by measuring ADC at incremental airway pressure values in two different models of emphysema: elastase treatment in rats, and cigarette smoke exposure in mice. We hypothesized that 1) there is an optimal inflation range within which separation between normal and diseased ADC values is greatest, 2) this inflation range is consistent among animal species and disease models, and 3) imaging lungs at optimal inflation improves discrimination between experimental groups when investigating disease progression.

METHODS

Animal preparation.

All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. In the rat studies, 6- to 8-mo-old healthy (n = 7) and emphysematous (n = 7) Sprague-Dawley rats (500 ± 50 g body wt) were used. Emphysema was induced through a single intratracheal instillation of porcine pancreatic elastase (22 U/100 g), and animals were studied ~3 mo after induction (11). For imaging, rats were anesthetized, intubated, temporarily paralyzed, and mechanically ventilated by a custom small-animal MR-compatible ventilator with a delivery accuracy of ± 100 μl/breath. Rats were ventilated with a mixture of ⁴He:O₂ (4:1) (3) at 60 beats/min and I:E = 1:2, at a nominal V_T = 10 ml/kg. Animals’ airway pressure tracing was continuously monitored and recorded by a high-precision, MR-compatible optical pressure transducer (Samba Sensors, Gothenburg, Sweden). Peak inspiratory pressure (PIP) was measured at each incremental inhalation level. Before imaging began, rats underwent an alveolar recruitment maneuver using a stepwise sequence of positive end expiratory pressure (PEEP) levels to minimize the effect of atelectasis (4).

In the mouse studies, 4-mo-old C57BL/6 mice (Jackson, Bar Harbor, ME) received nose-only exposure to 4% cigarette smoke from 3R4F cigarettes (College of Agriculture, Reference Cigarette Program, University of Kentucky) for 2 h/day, 5 days/week, for 6 mo (n = 6) and 18 mo (n = 6). Smoke was generated by a Baumgartner-Jaeger CSM 2070i Smoking Machine (CH Technologies, Westwood, NJ). The 4% smoke gas was produced via a 2-s, 35-ml puff from each cigarette once per minute; it was then mixed with air and delivered to the exposure tower. During smoke exposure, mice were maintained in restraining tubes containing stainless-steel nose cone inserts. Control mice were similarly restrained and exposed to air with the same frequency as smoke exposure cohorts for 6 mo (n = 8) and 18 mo (n = 7). To rule out the effect of aging and smoking machinery on lung structure, two additional groups of untreated (naïve) mice were studied at the time points of 6 mo (n = 7) and 18 mo (n = 7). All six mice cohorts were provided by GlaxoSmithKline (King of Prussia, PA), and further details on cigarette smoke model induction can be found in Ref. 27. Smoke exposure studies were conducted in accordance with the GlaxoSmithKline policy on the care, welfare, and treatment of laboratory animals and reviewed and approved by the Institutional Animal Care and Use Committee at GlaxoSmithKline. All mice were imaged in random order during a 1-wk period after transfer to the animal housing facility at University of Pennsylvania. During the imaging session, mice were anesthetized, intubated with a 1.5-mm endotracheal tube, and mechanically ventilated. The mice received a mixture of ⁴He:O₂ (4:1), V_T = 10 ml/kg body wt; 110 beats/min; I:E = 1:2. Three out of 24 mice in the 6-mo cohort died during intubation, anesthesia, or mechanical ventilation, as did 4 out of 24 mice in the 18-mo cohort.

ADC imaging protocol.

In all rats and in the 18-mo mouse cohort, images were obtained during a 3-s breath hold at incremented inflation levels, which were controlled by varying the inhalation time (0 to 2,000 ms) at a fixed inspiratory flow rate. This resulted in PIP levels ranging from 0 to 35 cmH₂O. Only a single inflation level (V_T = 10 ml/kg) was assessed in the 6-mo mouse cohorts. Immediately before image acquisition, animals received a sigh breath (20 ml/kg) to fill the lung with HP ³He:O₂ (4:1), followed by two identical breaths at the designated inflation level with the same HP ³He:O₂ (4:1) gas mixture, as shown in Fig. 1.

Fig. 1.

A: schematic diagram of the ventilation protocol used to measure apparent diffusion coefficient (ADC) at end inspiration. Prior to the imaging sequence, lungs were ventilated with ⁴He:O₂ at a 4:1 ratio. Then a single high volume “sigh” breath was delivered with hyperpolarized ³He:O₂ at the same ratio. This was followed by three hyperpolarized ³He:O₂ breaths at the desired inflation volume, with an inspiratory hold after the last inspiration during which ADC was measured. B: graded inflation levels were reached by increasing inspiratory time (IT), while maintaining a constant inspiratory flow. PIP, peak inspiratory pressure.

ADC was measured during a breath hold, using an interleaved diffusion-weighted gradient echo imaging pulse sequence with centric phase-encoding in a 50-cm bore 4.7 T MRI scanner (Varian, Palo Alto, CA) equipped with a 12-cm and 25 G/cm gradient and a 2.75″ ID quadrature 8-leg birdcage body coil (Stark Contrast, Erlangen Germany) for rats, as well as a 1.5″ ID quadrature 8-leg birdcage body coil (Stark Contrast) for mice. The bipolar diffusion-sensitizing gradients were applied along the phase-encoding (L–R) direction, and the b values are given by b(j) = [2⋅π⋅γ⋅G(j)]²⋅[δ²⋅(Δ − δ/³) + τ³/30 − δ⋅τ²/6], where the gyromagnetic ratio of ³He γ = 32.43 mHz/T, diffusion gradient amplitude G(j) = 0.00, 20.00, 16.25, 11.50, 0.00 G/cm, gradient ramp time τ = 180 μs, diffusion gradient duration (from the beginning of the ramp to the end of the gradient flat top) δ = 200 μs, and the diffusion time (from the beginning of the first to the beginning of the second gradient lobe) Δ = 1 ms. The corresponding b values were 0, 6, 4, 2, 0 s/cm². The initial and final zero b values were used to estimate the RF pulse flip angle. For all animals, values of α = 4–5°, TR = 6.6 ms, and TE = 4 ms were utilized. The ADC value for each pixel was fitted to S(j) = S₀⋅exp(–b(j)/ADC), where S(j) is the pixel intensity at the corresponding b value, and S₀ is the intensity with b value equal to 0. Images were acquired in the middle coronal slice of the lung with the following imaging parameters: field of view (FOV) = 6 × 6 cm², slice thickness (ST) = 6 mm, matrix size (MS) = 64 × 64 for rats; for mice, FOV = 3 × 3 cm², ST = 5 mm, MS = 64 × 64.

Image analysis.

An exponential model was used to evaluate ADC response to lung inflation. In each individual rat and mouse, the PIP for all inflation volumes and the corresponding mean ADC value as a function of PIP in the imaged slice were fitted to a second-order response equation (9, 15), which yielded ADC_TLC, ADC_FRC, and the changing rate (ω).

Tissue fixation and histology analysis.

Tissue analysis was performed in the rats. After imaging, each animal was euthanized using a lethal dose of pentobarbital sodium. Rat lungs were inflated with 10% formalin solution for 24 h at a transpulmonary pressure of 20 cmH₂O. Three to five lung slides with 5-μm thickness, including the whole coronal lung field, were obtained in positions that approximately matched the MRI image slices. Slides were scanned using a scanner (Aperio-Technologies, Vista, CA) with ×20 magnification. The mean and second-order moments of airspace diameters (D₀, D₂) (20, 24) were analyzed using a custom MATLAB code (17) and compared with the mean ADC results of the whole slice.

Statistical analysis.

Studentʼs t-test was used to compare between cohorts. A linear regression model was used to study the effect of smoking and age on ADC. The least squares method was used for curve fitting. P values < 0.05 (for two-tailed testing) were considered significant.

RESULTS

Elastase-treated rats: analysis of lung inflation.

Figure 2 shows examples of regional ADC maps obtained at the examined inflation levels with corresponding measured PIP values for one elastase-treated and one control rat. Mean slice values of ADC vs. PIP are plotted in Fig. 3 for each individual rat. As in our previous studies, ADC decreased slightly at higher pressures after attaining its maximum value. The second-order response equation was able to capture this change and provide a reasonable fit between ADC and corresponding PIP values. The fitting residual was satisfactory in both control and elastase-treated rats (χ² < 0.001).

Fig. 2.

Representative images of ³He spin density maps and corresponding ADC maps obtained at increasing inflation volume and corresponding PIP in healthy vs. elastase-treated rats.

Fig. 3.

Plots of ADC vs. PIP in each individual control rat (top) and elastase-treated (bottom) included in the study. Each ADC vs. PIP plot was fitted with a second-order response system.

Elastase-treated rats: discrimination between healthy and emphysematous lungs.

Group statistics of ADC measurements are shown in Fig. 4. ADC was consistently higher in the elastase-treated rats than in controls at all tested inflation levels. The difference between the two groups reached its highest statistical significance at PIP = 10 to 11 cmH₂O (Fig. 4). This was identified as the point of optimal discrimination between emphysematous and healthy lungs (optimal PIP). ADC values at PIP = 0 cmH₂O, optimal PIP, and at the point of maximum ADC PIP (maximum PIP) are listed in Table 1. In the elastase-treated rats, ADC reached its maximum value at a smaller PIP (18.1 ± 1.4 vs. 20.4 ± 1.9 cmH₂O) than in controls (P = 0.014), but the slope of the ADC rise vs. PIP was greater (0.21 ± 0.02 vs. 0.19 ± 0.02) in the treated group (P = 0.039).

Fig. 4.

A: summary statistics (median, interquartile range, and extremes of the distribution) of fitted ADC values in healthy (black) vs. elastase-treated (red) rats. B: P values of the comparisons between the two groups are shown in at all tested incremental PIP levels. C: higher slope of the ADC vs. PIP curve. D: lower PIP corresponding to the maximum ADC suggest higher compliance in elastase-treated vs. normal rats.

Table 1.

List of ADC values measured at PIP corresponding to no inflation (PIP 0), optimal discrimination between emphysema and control rats (PIP optimal), and at the level of maximum inflation explored (PIP max) in all elastase-treated and control rats

	Control ADC Values, cm2/s			Elastase ADC Values, cm2/s
Rat No.	PIP 0	PIP optimal	PIP max	PIP 0	PIP optimal	PIP max
1	0.112	0.157	0.169	0.154	0.211	0.227
2	0.117	0.164	0.196	0.124	0.184	0.193
3	0.100	0.146	0.155	0.135	0.201	0.216
4	0.102	0.156	0.170	0.148	0.206	0.213
5	0.102	0.154	0.164	0.135	0.206	0.211
6	0.124	0.190	0.203	0.121	0.182	0.188
7	0.109	0.175	0.190	0.126	0.210	0.214

ADC, apparent diffusion coefficient; PIP, peak inspiratory pressure.

Comparison of histology and MRI data.

While absent in controls, obvious tissue destruction was found in emphysematous rats (Fig. 5, A and B). At PIP = 20 cmH₂O, which is also the tissue fixation pressure, we found strong correlations between global ADC and morphometric values of airspace geometry (Fig. 5, C and D), including mean airspace diameter (D₀, R = 0.79) and second-order moments of airspace diameter D₂ (R = 0.89). Similar to mean ADC values, the histologically derived values of airspace dimensions were noticeably different for emphysematous rats compared with controls. Mean airspace diameter, a standard marker of airspace enlargement due to emphysema, showed a strong correlation with ADC.

Fig. 5.

Representative histology slices of elastase-treated (A) and control (B) rats. Correlation between ADC values and histological metrics or airspace morphometry: mean airspace diameter D₀ (C), and second-order moments D₂ (D) of the diameter distribution.

ADC measurement in mice.

Inflation response in 18-mo smoking mice was comparable to those observed in elastase-treated rats. As in the rats, ADC values in smoke-exposed mice were consistently higher than in controls at all inflation levels (Fig. 6). Similar to the rats, optimal PIP was identified at PIP = 10 to 11 cmH₂O. ADC values at PIP = 0 cmH₂O, optimal PIP, and maximum PIP are listed in Table 2. No significant difference was found between naïve and control groups at any inflation level (P = 0.558). Figure 7 shows ADC values obtained at tidal inflation (PIP = 10 to 11 cmH₂O): smoke exposure caused a significant increase of ADC vs. control and naïve groups (P < 0.001) at both 6 and 18 mo. Although no significant interaction between age and smoking was shown by the data (P = 0.427), age as a separate factor shows a significant increase in ADC when compared with 6-mo mice (P < 0.001). A post hoc Studentʼs t-test showed that the smoke exposure group displayed higher ADC (P < 0.001) at 18 mo than at 6 mo.

Fig. 6.

Left: summary statistics (median, interquartile range, and extremes of the distribution) of fitted ADC values in naïve (green), control (blue) vs. smoke exposure (red). Right: panels show P values of the comparisons between mice after smoke exposure vs. control and naïve mice, plotted against incremental inflation levels.

Table 2.

List of ADC values measured at PIP 0, PIP optimal, and PIP max in all mice with 18-mo smoke exposure, controls, and naïve mice

	6-mo Smoked Mice ADC Values, cm2/s			18-mo Smoked Mice ADC Values, cm2/s
	Naïve	Control	Smoked	Naïve			Control			Smoked
Mouse No.	PIP optimal	PIP optimal	PIP optimal	PIP 0	PIP optimal	PIP max	PIP 0	PIP optimal	PIP max	PIP 0	PIP optimal	PIP max
1	0.090	0.102	0.095	0.089	0.129	0.136	0.077	0.118	0.120	0.105	0.141	0.139
2	0.088	0.083	0.118	0.054	0.117	0.113	0.074	0.124	0.127	0.073	0.135	0.129
3	0.112	0.094	0.115	0.082	0.127	0.116	0.072	0.126	0.132	0.093	0.145	0.128
4	0.102	0.112	0.111	0.072	0.106	0.112	0.091	0.130	0.129	0.095	0.146	0.156
5	0.114	0.112	0.125	0.052	0.115	0.115	0.090	0.118	0.119	0.097	0.147	0.139
6	0.102	0.110	0.125	0.099	0.119	0.116	0.083	0.127	0.117	0.110	0.129	0.131
7	0.123	0.116	0.149	0.061	0.125	0.128	0.084	0.120	0.113
8		0.099		0.071	0.112	0.113

For 6-mo mice cohorts, ADC was measured only at optimal PIP.

Fig. 7.

Summary statistics (median, interquartile range, and extremes of the distribution) of ADC value in 6-mo naïve (n = 7), 6-mo control (n = 8), 6-mo smoke (n = 6), 18-mo naïve (n = 8), 18-mo control (n = 7), and 18-mo smoke (n = 6) mouse cohorts.

DISCUSSION

In this study, we investigated ADC response, an established metric of airspace dimensions, to graded lung inflation in two animal models of emphysema. In addition to confirming higher ADC values in both rats and mice with emphysema compared with controls, we found that diseased lungs could be identified more accurately when imaging at intermediate inflation levels. This optimal inflation range was comparable between the two tested models of disease, and its use helped to measure the effects of smoking duration in mice.

In both disease models (rats and mice), the higher ADC vs. healthy controls was a consequence of airspace enlargement due to the destruction of interstitial structures. This pathogenesis has been confirmed by other authors (1a, 10), as well as by the current study, in which we report high correlations between tissue morphometric variables—the gold standard of emphysema quantification—and ADC (Fig. 5). Correlations with ADC were higher for D₂ (the second moment of the airspace diameters), a metric of abnormal distribution of airspace dimensions and an established histological marker of early emphysema (24).

Assessing ADC response across a broad range of inflation levels, rather than measuring ADC at a single standardized inspired volume (e.g., tidal volume), provided a more detailed characterization of diseased vs. normal lungs, and identified the optimal conditions for diagnostic accuracy. We achieved the latter by plotting the statistical significance of the intergroup ADC comparisons against all tested inflation levels (Fig. 4B). In both elastase-treated rats and mice exposed to smoke for 18 mo, we found the highest diagnostic accuracy at inflation pressures between 10 and 11 cmH₂O, corresponding to a tidal volume of ~8–10 ml/kg.

Our analysis of airspace inflation yielded several additional observations. First, unlike in the typical pressure-volume compliance curves used by other authors (30), we found a good fit between ADC and PIP using the second-order response equation, due to both the nonlinear relationship between the two variables and the biphasic effect of inflation volume on airway expansion. Second, ADC reached a distinct peak, indicating the onset of maximum airspace expansion, and then decreased at higher pressure. We have observed this phenomenon in previous studies on atelectasis and lung injury and are now reporting it for the first time in emphysema. Decreased ADC at high airway pressure is likely caused by recruitment (reopening) of atelectasis (4) or accessory alveoli (22), with a consequent redistribution of gas from previously inflated airspaces to newly opened units. This redistribution is not captured by whole lung measurements of respiratory mechanics and inflation (6). At lower airway pressure, alveolar expansion dominates inflation and results in rising ADC values. At higher inflation, however, ADC is affected by variable recruitment, as well as lung disease, and abnormal lung microstructure. Although the state of recruitment is highly variable and depends on lung ventilation history (16), the finding that HP MRI is best able to discriminate between diseased and normal lung tissue at intermediate inflation levels (i.e., before recruitment processes dominate the small airway expansion) is consistent with previously observed dynamic behavior (5).

Emphysema in both animal models was characterized by a steeper increase in ADC at lower inflation pressure, likely due to higher airspace compliance resulting from tissue disruption. This effect was particularly evident in the middle range of inflation, indicating an airway pressure span where lung mechanics are optimal. However, peak ADC occurred at lower pressure in emphysematous lungs, perhaps indicating that previously open airspaces are maximally expanded (before recruitment) at lower inflation pressure than normal. Further attempts at increasing airspace dimensions were unsuccessful and did not improve the diagnostic accuracy of ADC.

Strikingly, we not only found that both animal models displayed similar changes in ADC as pressure increased, but also that the transition from distension at low inflation pressure to recruitment at high-inflation pressure occurred at the same inspiration pressure. This occurred despite the obvious differences between species, as well as the fact that elastase and smoking exposure induce lung disease with different severities and time courses: pancreatic elastase induces severe alveolar destruction in the short term, while cigarette smoke exposure generates a milder and slower-developing disease. In a study comparing ADC values of elastase-induced, smoking-induced, and healthy mice, Dugas et al. (10) found a statistically significant difference between healthy and elastase-induced mice; the difference observed between smoke-exposed and healthy mice was smaller, however, likely indicating a closer replication of the initial slow progression of alveolar destruction and disease development. Leveraging the reproducibility of our findings between models, as well as our discovery of the optimal diagnostic range of inflation, we followed the evolution of emphysema during prolonged smoking exposure in mice and confirmed the gradual development of emphysematous changes captured by ADC. These changes were distinct from the effects of aging observed in naïve and control mice (Fig. 7). Nevertheless, it is important to note that ADC measures structural changes in the lungs after they are established, while the inflammatory responses to cigarette smoking may be tangible earlier in the course of smoke-related disease (1a).

Our study has several limitations. First, because of the small scale of the microstructures under examination, we cannot rule out the possibility that our ADC measures were affected by diffusion along branching airways. However, we believe that ADC values shown in this study and in our published works (4, 5) were mostly reflective of the geometry of subacinar airspaces—an assumption that is supported by the correlation with histology metrics (Fig. 5), by the plausibility of the responses to lung injury and treatment [e.g., airway pressure and recruitment (6)], and by their similarity to values reported in the literature (7, 19, 26). Unfortunately, given the large diffusivity of ³He and the limitations imposed by our imaging gradients, true discrimination between lung structures (e.g., alveolar ducts vs. sacs) was impossible with our approach. Additionally, we fitted our diffusion measurements to a monoexponential model instead of the models that have been used in microgeometric studies (23, 25, 32), a choice that was motivated by the difference of selected diffusion timings and b values. Nevertheless, we are convinced that this simple monoexponential model was adequate for the aim of the present study, which was to discriminate ADC in disease models vs. healthy tissue rather than to provide a comprehensive in vivo micromorphometric assessment. The use of the same diffusion time for both rats and mice constitutes another limitation of this study. Although we did not measure ex vivo morphometry in mice, we extrapolated the linear regression line in Fig. 5 to the values of ADC that we observed in this model (~0.10 to 0.14 cm²/s). This yielded estimated D₀ values that were similar to those reported in the literature for mice (~23 to 45 μm) (1a, 18), thereby reassuring us about the plausibility of our measurements in this species.

Conclusions.

We demonstrated the first HP gas MRI study in rodent models of emphysema by imaging ³He ADC at different lung inflation levels. Measuring ADC at optimal inflation pressure allowed us to differentiate between healthy and diseased lungs with greater accuracy, independent of the species and method of disease induction. These findings have the potential to improve the diagnostic accuracy of future HP gas diffusion studies in relevant disease populations.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants R01-HL-129805 and R01-HL-124986.

DISCLOSURES

K. Emami is employed by Polarean, Inc. B. Bolognese, J. Foley, and P. Podolin are employed by Glaxosmithkline, Inc.

AUTHOR CONTRIBUTIONS

Y.X., S.K., K.E., P.L.P., and R.R.R. conceived and designed research; Y.X., K.E., H.P., B.J.B., and J.P.F. performed experiments; Y.X., M.C., S.K., K.E., and H.H. analyzed data; Y.X., M.C., S.K., K.E., H.H., P.L.P., and R.R.R. interpreted results of experiments; Y.X. prepared figures; Y.X., M.C., and L.H. drafted manuscript; Y.X., M.C., S.K., K.E., H.H., I.D., J.R., L.H., N.M., J.N., B.J.B., J.P.F., P.L.P., and R.R.R. edited and revised manuscript; Y.X., M.C., S.K., K.E., H.H., I.D., J.R., L.H., N.M., J.N., B.J.B., J.P.F., P.L.P., and R.R.R. approved final version of manuscript.