Dynamic respiratory muscle function in late-onset Pompe disease

Maximal inspiratory pressure (PIMAX) reflects inspiratory weakness in late-onset Pompe disease (LOPD). However, static pressure tests may not reveal specific respiratory muscle adaptations to disruptions in breathing. We hypothesized that dynamic respiratory muscle functional tests reflect distinct ventilatory compensations in LOPD. We evaluated LOPD (n = 7) and healthy controls (CON, n = 7) during pulmonary function tests, inspiratory endurance testing, dynamic kinematic MRI of the thorax, and ventilatory adjustments to single-breath inspiratory loads (inspiratory load compensation, ILC). We observed significantly lower static and dynamic respiratory function in LOPD. PIMAX, spirometry, endurance time, and maximal diaphragm descent were significantly correlated. During single-breath inspiratory loads, inspiratory time and airflow acceleration increased to preserve volume, and in LOPD, the response magnitudes correlated to maximal chest wall kinematics. The results indicate that changes in diaphragmatic motor function and strength among LOPD subjects could be detected through dynamic respiratory testing. We concluded that neuromuscular function significantly influenced breathing endurance, timing and loading compensations.

Dynamic MRi of the thorax. Table 2 summarizes the C-C and A-P distances at end-inspiration and end-expiration during tidal breathing and IC, as well as the kinematic excursion. One patient (P3) required use of BiPAP during the MRI test; therefore, IC kinematics were measured in only six LOPD subjects. There was a significant interaction between the group and direction of movement for kinematic reserve (Fig. 1e). While kinematic reserve was similar between patients and controls for A-P excursion, the C-C kinematic reserve was significantly smaller in LOPD (F = 3.571, p < 0.05).
Single-breath ILC. Significant single-breath ILC responses are depicted in Fig. 2. There was a significant interaction effect between the breath type, group, and load magnitude for PIF (3- p < 0.05). During loaded breaths, PIF decreased progressively in LOPD patients as the load magnitude increased, and it was significantly lower than controls at the highest load magnitude (60% of MIP, p < 0.005). In both LOPD and control subjects, PIF was highest during the recovery breaths and lowest during the loaded breaths (breath main effect, p < 0.001). No differences were found for dP/dt. On exhalation, the sample generated a significantly higher PEF during the recovery breaths, following single-breath load magnitudes>0% (breath order-load interaction, F = 6.463, p < 0.005). There were significant breath-load interaction effects for VTI (F = 19.403, p < 0.005) and VTE (F = 7.013, p < 0.005) during single-breath ILC. Neither group preserved breath volume during the 40% and 60% load magnitudes, when compared to lower loads, unloaded control breaths, or recovery breaths (p < 0.005). Despite a lower PIF during loaded breaths, both Pompe and control subjects minimized the loss of volume by tending to lengthen TI (F = 4.186, p = 0.06). Additionally, both groups significantly shortened TE during loaded breaths (F = 5.767, p < 0.05), as compared to the unloaded control and recovery breaths. endurance time. Figure 3 depicts the endurance time of the LOPD and CON subjects. Endurance tests of two control subjects (C1 and C3) were terminated early by the investigators, despite excellent inspiratory valve opening, for clinically significant elevations of blood pressure. As a group, control subjects were able to continue breathing with a 40% inspiratory load for 7.7 ± 3.1 minutes, while LOPD subjects managed the inspiratory load for 2.9 ± 2.8 minutes to failure (t = 2.983, p < 0.05).

Discussion
The key findings of these research tests of dynamic inspiratory muscle function indicated kinematic evidence of diaphragm paresis, accompanied by altered inspiratory endurance and strength in patients with LOPD. The ILC tests indicate an impaired ability for LOPD patients to generate inspiratory flow with an unanticipated inspiratory load. However, a compensatory prolongation of TI helped to preserve volume during the loaded breath. We also observed flow and volume ILC were associated with MRI tests of maximal chest wall expansion, but not with diaphragm kinematics, of patients. While likely more useful as research tools than in daily clinical practice, these tests appear to capture distinct aspects of dynamic respiratory muscle function beyond standard pulmonary function tests, which could help clarify how patients maintain minute ventilation despite gradual LOPD disease progression.
We used dynamic MRI to evaluate kinematics of breathing at rest and during inspiratory capacity. The C-C kinematic reserve represented the difference between resting diaphragm excursion and maximal diaphragm excursion, and it indicated the presence of diaphragmatic paresis in LOPD. On the other hand, A-P resting and reserve excursion did not differ significantly between patients and controls, indicating relative preservation of chest wall expansion in patients.  Table 2. Anterior-posterior (AP) and cranio-caudal (CC) kinematic excursion measured in LOPD and CON subjects revealed that at peak inspiration, the CC height of the lung was significantly reduced in LOPD, and this reduced inspiratory height resulted in a lower CC excursion during inspiratory capacity (mean, SD www.nature.com/scientificreports www.nature.com/scientificreports/ Existing studies in adult onset Pompe also contrasted MRI to conventional lung function tests 3,7 , but many prior efforts focused on static breath-holds. Our findings of early diaphragmatic paresis in LOPD were also in agreement with existing respiratory studies using opto-electronic plethysmography (OEP) 21 and static MRI 3,6 . Additional advantages to using dynamic MR modes include the ability to average the kinematics over multiple breaths in a short period of time and the removal of possible confounding effects of elastic recoil on breath-holding 22 . MRI may be suitable for multi-site research studies that cannot accommodate the substantial equipment costs and data analyses required with OEP. For example, children with infantile-onset Pompe disease tolerated dynamic respiratory MRI in a clinical trial of AAV1 gene therapy 11 , while on ventilatory support and for brief periods of CPAP or independent breathing.
Thoracic MRI may also offer scientists insights regarding the presence of fat or fibrotic incursions in the respiratory muscles 23,24 . However, there are limitations to MRI, particularly in the context of clinical care. In the current study, one LOPD subject with inspiratory muscle weakness and absent C-C excursion required the use of BiPAP to image tidal breathing. Therefore, while ultrasonography can yield lower tissue resolution and an insufficient window in some patients, it may be preferred in clinical settings or to monitor moderate to severe diaphragmatic involvement 25,26 . Additionally, sniff nasal inspiratory pressure (SNIP) reflects pressure associated with a maximal effort sniff. While the current study focused on MIP due to its common inclusion as a clinical study endpoint, the SNIP test is well tolerated in neuromuscular diseases 27 with potential predictive value for respiratory morbidity 28 .
Tlim was restricted in LOPD and correlated strongly with conventional lung function tests as well as maximum diaphragm excursion during MRI. In contrast to the MRI tests, all LOPD subjects were able to complete Tlim. Advantages of endurance tests are their ability to discern changes in pressure generation that do not involve a static hold and become less effort-dependent with approaching task failure 29 . Since the respiratory Figure 1. Kinematic reserve in LOPD and control subjects. Dynamic MRI was used to evaluate cranio-caudal lung height, representative of diaphragm excursion, and anterior-posterior (A-P) chest wall expansion. We did not find significant group differences for C -C excursion (a) or A-P expansion (b) during tidal breathing. Inspiratory capacity C-C excursion (c) was significantly lower for LOPD patients. In contrast, A-P expansion during inspiratory capacity (d) remained similar between groups. Kinematic reserve is the difference between MRI measurements of diaphragm (C-C) and chest wall (A-P) excursion at rest and during inspiratory capacity maneuvers. While the C-C reserve of the diaphragm (e) was significantly smaller for LOPD subjects, we observed a similar difference in the A-P reserve (f) of the chest wall muscles (*Significant group-direction interaction, F = 3.571, p < 0.05). (2019) 9:19006 | https://doi.org/10.1038/s41598-019-54314-8 www.nature.com/scientificreports www.nature.com/scientificreports/ pattern and mouth pressure were standardized, the endurance test has been considered less susceptible to learning effects, in healthy subjects and patients with neuromuscular disease 30,31 . As loads increase in magnitude or are sustained to fatigue, inspiratory flow and inspiratory time progressively diminish 32 . Subjects reached task failure (Tlim) when they could not generate sufficient pressure to continue to open the inspiratory valve and receive airflow 32 . Longitudinal changes in Tlim may indicate an external factor (e.g. disease, training, medication) altered the endurance Tlim, and further study is needed on whether tests of threshold loading can longitudinally track changes respiratory muscle function with either progressive disease or in response to therapeutic interventions such as inspiratory muscle training. Indeed, recent reports suggest inspiratory muscle training can lead to strength gains, in some individuals with LOPD 33 , We previously observed that difficult to wean ICU patients who failed to wean after inspiratory muscle training had lower flow/volume ILC that failed to improve with training 34 . We speculate that Tlim and/or ILC tests could help scientists further identify inspiratory flow and volume signatures associated with an incomplete strengthening or functional benefit from inspiratory muscle training 16,31 .
Compensatory responses to a single-breath threshold load include reductions in inspiratory flow and volume, with prolongation of TI and reduction of TE 15,35 . We observed these modifications during single-breath and load magnitude. During loaded breaths, peak inspiratory flow (a) progressively decreased in LOPD as compared to control subjects (3-way interaction, F = 2.794, p < 0.05). Compared to recovery breaths, peak expiratory flow (b) was lower during loaded breaths, but neither group assignment nor load magnitude affected the loaded peak expiratory flow. The inspired volume of loaded breaths (c) was similar for both groups at the lowest load magnitude, but in LOPD subjects, volume progressively decreased with larger load magnitudes (breath-load interaction, F = 19.403, p < 0.005). Inspiratory time (d) did not vary by breath type, load magnitude, or group, largely due to variability of breath timing in the control subjects. Expiratory time (e) was lower for loaded breaths, compared to unloaded controls and recovery breaths (F = 5.767, p < 0.05). Despite the group differences in flow and volume ILC, the maximal rate of inspiratory pressure development (dP/dt) of the loaded breaths (f), an estimate of neuromuscular activation, did not differ between the groups. (Mean, SEM depicted). ILC, with some distinctions in the LOPD group. Consistent with other neuromuscular diseases 36 , both TI and TE were slightly shorter in LOPD subjects than CON. As the magnitude of a single breath load increased, both PIF and volume dropped significantly in LOPD subjects. With inspiratory challenges, the diaphragm responds by increasing the firing frequency of active motor units and by recruiting previously inactive diaphragm motor units 37 . Additionally, accessory muscle motor units that are largely inactive at rest are recruited 38,39 . In healthy individuals, recruitment of chest wall muscles is thought to offset diaphragm fatigue during sustained inspiratory loading 40 . In LOPD, we noted a strong association between increased chest wall kinematic motion and the PIF and VTI generated during single-breath ILCs. We speculate that chest wall respiratory recruitment may influence single-breath ILC responses in LOPD, because early, severe diaphragm paresis minimizes the ability of the diaphragm to compensate to the transient, unexpected load.
Fifty-seven percent of the LOPD subjects required nighttime non-invasive respiratory support, which itself indicates a loss of respiratory neuromuscular reserve 41 . Both neurological and muscular contributions to respiratory dysfunction in LOPD have been reported in the literature. Diaphragm dysfunction is a distinguishing characteristic of LOPD identified even among mildly affected LOPD patients 42 , but an enhanced inspiratory neural drive may be able to preserve independent breathing in early diaphragm dysfunction 43 . Neuromuscular activity, as estimated by dP/dt, was similar for Pompe and control subjects during ILC 16,44 . ILC loading was administered as a proportion of MIP, rather than a fixed value, indicating that relative proportional loads elicited a similar proportional neuromuscular activation. While the small sample did not permit a more rigorous comparison of neuromuscular activation in the patients who used nighttime support 43,[45][46][47][48] , the use of nighttime ventilation in LOPD improves both ventilation and oxygenation during sleep, without an appreciable change in respiratory muscle strength 49 . In the current study, the similar dP/dt and prolongation of TI with threshold loading in both LOPD and control subjects indicates that both groups used a similar motor control strategy to overcome inspiratory loads, and we speculate this could indicate a loss of respiratory motor units in Pompe, as opposed to an altered functioning of the motor neurons 43,[45][46][47][48] .
We acknowledge some limitations to this study. FVC and respiratory pressures were not tested in supine. Subject fatigue was the primary consideration for this decision, and IC in the MRI was computed instead. While the requirement to maintain a consistent respiratory rate and duty cycle during the endurance test has been criticized by some, it reduces variability in the breathing pattern, namely an early prolongation of exhalation, which could artificially prolong Tlim 50 . It could be argued that both the Tlim test could be considered as effort-dependent as lung function and maximal respiratory pressure tests 51 . A distinguishing factor is the binary nature of the threshold valve, which blocks inspiratory flow for inspiratory pressures below the threshold. Additionally, the threshold valve mechanics and limited compensatory strategies for clinical populations reduce effort and learning effects on ILC and Tlim. Finally, the finite time and energy of LOPD patients required us to limit the range of respiratory tests. Specifically, invasive tests of evoked diaphragmatic pressure would provide additional mechanistic insights on non-voluntary activation of the diaphragm, while SNIP and diaphragm ultrasound tests offer clinical insights into dynamic diaphragm function.
In summary, our findings provide evidence of early dynamic diaphragmatic dysfunction in LOPD, compared with controls. This provides critical evidence that diaphragm dysfunction in Pompe alters not only static force generation, but also impairs the ability to preserve minute ventilation during transient or sustained inspiratory loads. In turn, a lowered dynamic inspiratory muscle reserve may ultimately affect a patient's ability to resist physiological stressors such as bronchoconstriction or a narrowed airway during sleep, placing them at higher risk for failure. Future studies should distinguish how dynamic respiratory muscle function changes with disease progression.

Methods
Subjects. Fourteen subjects volunteered to participate. Seven subjects had a diagnosis of LOPD confirmed by genetic testing or enzyme activity assays (age 29-63 years, 5 women), and the remaining seven were age-and gender-matched unaffected control subjects (age 20-61 years). Based upon a power analysis of preliminary data on the correlation between PI MAX and the volume ILC, at least 6 subjects per group would be needed for α = 0.05, (1-β) = 0.9, (ρ = 0.68). The study design and procedures were approved by the University of Florida Institutional Review Board and conducted in accordance with principles established in the 1964 Declaration of Helsinki and www.nature.com/scientificreports www.nature.com/scientificreports/ its subsequent amendments. All participants provided their informed consent prior to inclusion in the study. The study was registered in February 2015 (NCT02354664).
Exclusionary characteristics for patient participation included: pre-existing obstructive lung disease or asthma, a forced vital capacity (FVC) < 30% of age/gender predicted values, requirement for mechanical ventilation while awake and upright, inability to travel to the study site, pregnancy, or any other chronic medical condition that would interfere with study participation. experimental protocol. Each subject participated in a single 6-hour study visit, divided into two days.
Clinical pulmonary function and maximal respiratory pressure tests. Upright forced vital capacity (FVC) was completed per American Thoracic Society guidelines 52 and compared to references for age, gender and ethnicity 53 . Tests were repeated until three technically-acceptable efforts were acquired within 5% variability. In addition, subjects performed PI MAX maneuvers from residual volume and maximal expiratory pressure (PE MAX ) from total lung capacity, per American Thoracic Society testing guidelines 52 . Each trial was repeated after a>30 second break, until <10% variability was achieved between three trials (typically achieved within 3-7 trials).
Dynamic MRI of the thorax. Patients were positioned supine in a 3.0T research MR scanner (Philips Achieva) using a 32-channel cardiac coil. We customized a dynamic MR sequence, based on techniques validated in pulmonary disease 54,55 . Triplanar localizer images were acquired during a maximal inspiratory breath-hold, to establish the optimum field of view for the dynamic scans. A T2-weighted, balanced, gradient echo sequence with short repetition time and retrospective gating was employed to obtain fast imagery with excellent tissue resolution 55,56 , (TR = 2.97 ms, TE = 1.62ms, slice thickness: 8mm, ~5 frames/sec). Separate dynamic scans captured resting tidal breathing and inspiratory capacity maneuvers, in the coronal and right sagittal planes. Each tidal breathing acquisition lasted one minute and captured 12-15 tidal breaths.
During each MRI scan, respiratory parameters were collected simultaneously, using a flowmeter (Capnostat, Philips Respironics) connected to the patient via facemask and placed in series with a respiratory monitor (Nico, Philips Respironics) and laptop computer. Respiratory data acquisition occurred at 100 Hz frequency. Extended tubing was required to maintain the data acquisition equipment behind the 5 Gauss line, which added 24 mL of dead space to the respiratory circuit.
Resting breathing pattern. Tidal breathing pattern was assessed with subjects seated wearing a sealed facemask and in a relaxed state listening to music. Approximately 10 minutes of steady-state, resting data was sampled at 100 Hz with a respiratory monitor (CO2SMO, Philips Respironics) attached to a laptop. The resting respiratory rate was used as the breath rate for the endurance test.
Single-breath ILC. Single-breath ILC was tested using a custom apparatus that contained 4 pressure-threshold devices (Threshold IMT, Philips-Respironics) connected to the patient circuit, which consisted of a pneumotachograph (HR1000, Hans Rudolph) with phalange mouthpiece and nose clip. Some of the pressure-threshold devices contained a modified spring to provide higher threshold loads (14-82 cm H 2 O). Inspiratory loads of either 10, 20, 40, or 60% of PI MAX (or a no-load control) were applied during exhalation, for a single inspiratory effort. Subjects were separated from the testing apparatus by a screen. Six presentations of each load were applied in random order, and a minimum of 4 and maximum of 12 unloaded breaths separated each ILC breath.
Respiratory muscle endurance test. Endurance was evaluated by identifying the time limit (Tlim) that a participant could maintain breathing with a submaximal inspiratory threshold load, as described previously in Duchenne muscular dystrophy 31 . Subjects rested a minimum of 30 minutes prior to the test. Subjects used a mouthpiece and nose clip connected to pneumotachograph and pressure transducer. Respiratory parameters and breathing kinematics were sampled at 100 Hz (PowerLab S35/16, AD Instruments). A pressure-threshold inspiratory load equivalent to 40% of PI MAX was placed on the inspiratory port of the mouthpiece. The respiratory rate was set to each subject's self-selected resting breath rate with a 50% duty cycle, and a two-tone metronome timed inhalation and exhalation. Oxygen saturation, end-tidal CO 2 (ETCO 2 ), heart rate, tidal volume, and respiratory rate were continuously monitored, with blood pressure readings every 2 minutes. Subjects received encouragement to maintain the target rate and mouth pressure. The test ended when one of the following conditions was met: (1) the inspiratory port did not open for 3 continuous breaths, (2) significant changes in heart rate or blood pressure outside of conventional exercise testing limits, (3) the subject reported intolerance or inability to continue, or (4) the subject reached 10 minutes of continuous loaded breathing.
Data analysis. Dynamic MRI of the thorax. MRI images were uploaded to a secure database and then analyzed offline with an open-source DICOM viewer (Osirix, Pixmeo) by a trained assistant blinded to subject group. We evaluated the right sagittal plane, since height of the diaphragm differs by ventral-dorsal position (See Fig. 4) 57 . The first ten uninterrupted breaths in each breathing sequence were evaluated. The cranio-caudal length (C-C) and anterior-posterior width (A-P) of the thorax were measured in each breath at end-inhalation and end-exhalation and represented the positions of the diaphragm and chest wall, respectively. Kinematic excursion was calculated as the difference between inspiration and expiration distance. Figure 4 illustrates sagittal plane images in representative unaffected and LOPD subjects, during tidal breathing and inspiratory capacity (IC) www.nature.com/scientificreports www.nature.com/scientificreports/ maneuvers. MRI images were time-matched to the respiratory measures. The respiratory measurements were averaged for tidal breathing and for 3 IC maneuvers.
Single-breath ILC. At each load magnitude, the breaths with the highest and lowest volume were excluded. The remaining 4 presentations were averaged for peak inspiratory and expiratory flow (PIF, PEF), inhaled and exhaled volume (VTI, VTE), inspiratory and expiratory times (TI, TE), and maximal date of inspiratory pressure development (dP/dt). The dP/dt was the positive peak of the pressure-time derivative 44 . In addition, we averaged the breaths before (unloaded control) and after (recovery) the loaded breath.
Statistical analysis. Analyses were performed with the IBM SPSS Statistics 21.0 statistical package (IBM Corp., Armonk, NY, USA). Group demographics were compared using 2-sample t-tests. For MRI studies, we used 3-way repeated measures ANOVAs to evaluate the kinematic excursion of the respiratory muscles (2 subject groups, 2 directions of movement, 2 conditions -rest vs IC), and 2-way ANOVA to compare the volumetrics (2 groups, 2 respiratory conditions). Single-breath ILC was assessed using a 3-way ANOVA (2 groups, 3 breath types, 5 load magnitudes). A 2-way ANOVA tested dP/dt of the loaded breaths. With ANOVA testing, we corrected violations of sphericity with Greenhouse-Geiser adjustments. Tukey's HSD post hoc test was used to assess individual ANOVA differences. Two-sample t-tests were used to compare the Tlim of the Pompe subjects and controls. Pearson's correlations were used to evaluate the linear relationships between the pulmonary function, imaging, ILC, and endurance variables. The alpha level was 0.05, with the exception of correlations, which used a more stringent threshold of p < 0.005 to correct for multiple correlations.

Data availability
The datasets generated during the current study are available from the corresponding author on reasonable request.