Hypoxic pulmonary vasoconstriction (HPV) is thought to actively regulate ventilation-perfusion (V̇a/Q̇) matching, reducing perfusion in regions of alveolar hypoxia. We assessed the extent of HPV in the healthy human lung using inhaled nitric oxide (iNO) under inspired oxygen fractions (FiO2) of 0.125, 0.21, and 0.30 (a hyperoxic stimulus designed to abolish HPV without the development of atelectasis). Dynamic measures of blood flow were made in a single sagittal slice of the right lung of five healthy male subjects using an arterial spin labeling (ASL) MRI sequence, following a block stimulus pattern (3 × 60 breaths) with 40 ppm iNO administered in the central block. The overall spatial heterogeneity, spatiotemporal variability, and regional pattern of pulmonary blood flow was quantified as a function of condition (FiO2 × iNO state). While spatial heterogeneity did not change significantly with iNO administration or FiO2, there were statistically significant increases in Global Fluctuation Dispersion, (a marker of spatiotemporal flow variability) when iNO was administered during hypoxia (5.4 percentage point increase, P = 0.003). iNO had an effect on regional blood flow that was FiO2 dependent (P = 0.02), with regional changes in the pattern of blood flow occurring in hypoxia (P = 0.007) and normoxia (P = 0.008) tending to increase flow to dependent lung at the expense of nondependent lung. These findings indicate that inhaled nitric oxide significantly alters the distribution of blood flow in both hypoxic and normoxic healthy subjects, and suggests that some baseline HPV may indeed be present in the normoxic lung.
- arterial spin labeling
- hypoxic pulmonary vasoconstriction
- nitric oxide
- pulmonary blood flow
although the first documented evidence of hypoxic pulmonary vasoconstriction (HPV) in humans dates back to the work of Motley et al. (33) in 1947, some 65 years later there remains considerable disagreement as to its role in healthy lung function. While it has been postulated that pulmonary vascular smooth muscle responsiveness to alveolar Po2 (PaO2) serves to mitigate ventilation-perfusion mismatch throughout life (33, 35), it has also been suggested that HPV in the adult is a vestigial reflex response left over from intrauterine development, when high pulmonary vascular resistance was required (47).
If HPV in the adult is a vestigial response, one would expect it to be governed by a fairly simple regulatory scheme that is chiefly operative only when the lung is exposed to frank hypoxia. However, while there is sufficient evidence to suggest that alveolar oxygen tension directly triggers a redox sensor in pulmonary vascular smooth muscle cells as part of the HPV response (34, 45), endothelial production of nitric oxide (NO), a potent vasodilator, is believed to play a significant role in modulating the strength of the hypoxic response (4, 28). It is therefore also plausible that this complex balance of vasoconstrictive and vasodilatory forces supports a mechanism by which gas exchange efficiency can be regulated in healthy adults through local matching of ventilation (V̇a) to perfusion (Q̇), quantified as the V̇a/Q̇ ratio. If it were the case that HPV subserves active V̇a/Q̇ regulation, however, then it would be expected to be active under normoxic conditions to be effective.
We have previously reasoned that changes in vascular tone as a result of changing FiO2 might manifest as alterations in pulmonary blood flow heterogeneity. When assessed in a strictly spatial sense (as measured by relative dispersion, RD, of images of pulmonary blood flow), only very modest changes with exposure to hypoxia were detectable, with no discernable difference between normoxic and hyperoxic conditions (2). More recently, however, we showed that the inclusion of temporal information in the assessment of blood flow heterogeneity revealed marked differences not previously seen. The key finding of that work was that while spatial heterogeneity was unchanged (2), changes in FiO2 led to shifts in the underlying spatial flow pattern, and in the case of hypoxia, increased temporal flow variability as well (5).
As a key endogenous regulator of HPV, administration of nitric oxide via inhalation (iNO) is another method that has been employed in the assessment of vascular tone in healthy subjects. As opposed to other vasodilatory agents, iNO shows selectivity for the pulmonary vasculature due to its rapid inactivation by hemoglobin binding and thus avoids potential confounding influences on the systemic circulation (32, 37). Studies to date have demonstrated that iNO reduces pulmonary vascular pressures under conditions of hypoxia, but not normoxia (14, 37), supporting the view that HPV is normally inactive in healthy lungs under normoxic conditions. However, pulmonary vascular pressure is expected to respond to net changes in pulmonary vascular resistance, and may be insufficient to detect subtle changes in blood flow distribution. The remarkable ability of the lungs to maintain low vascular pressures despite five-fold increases in cardiac output during exercise (22, 29) speaks to a large “silent” zone whereby flow changes, undetectable from pulmonary arterial pressure alone, may occur.
In this study, we hypothesized that if iNO mediated relaxation of vascular tone were occurring under normoxic conditions at a regional level, then changes in spatiotemporal flow distribution should coincide with the addition of iNO. Further, any changes should be enhanced under hypoxic conditions in which HPV is most strongly engaged, and lessened or absent under hyperoxic conditions, when HPV should be least active. To address this question we measured the spatial-temporal distribution of pulmonary blood flow under conditions of hypoxia, normoxia, and hyperoxia (separate studies), including a challenge period in each study in which 40 ppm inhaled nitric oxide was added to the inspired gas mixture.
An MRI technique called arterial spin labeling (ASL) was used to make repeated measurements of pulmonary blood flow in a single sagittal slice of the right lung under conditions of hypoxia, normoxia, or hyperoxia, either with or without the addition of inhaled nitric oxide. The approach used was similar to that previously performed by us in Ref. 5, with a synopsis and specific differences presented below.
We studied a population of six healthy young men between 23 and 36 yr of age, with normal lung function and no history of cardiovascular or pulmonary disease. Subjects were screened for health history, pulmonary function, and potential contraindications for MRI prior to testing. The University of California San Diego Human Research Protection Program approved the study, and subjects provided written, informed consent prior to participation. Technical difficulties ultimately resulted in exclusion of one subject from the set (discussed in Potential for cardiac and respiratory gating failure under Limitations and Technical Considerations). Descriptive data for the remaining five subjects are given in Table 1.
Physiological Challenge and Monitoring
Setup and monitoring.
Subjects were fitted with a full-face mask (7400 series Oro-Nasal Mask, Hans Rudolph) and equipped with a nonrebreathing T-valve (Hans Rudolph). The inspiratory port was connected to a gas-tight bag, fillable from a gas cylinder in the control room, while the outlet was connected to a 6-m low-resistance expiratory line that permitted measurement of expired tidal volume with a ParvoMedics Metabolic Measurement System (Sandy, UT). Subjects then laid supine on the posterior elements of an eight-channel MR cardiac coil, with the anterior elements placed directly on the anterior chest wall before advancing into the MRI scanner. An electrocardiogram (lead II) was used for gating the MRI scanner and a fingertip pulse oximeter (7500FO, Nonin, MN) was used to monitor oxygen saturation (SpO2) and heart rate (HR), which was recorded every 2 min.
Respiratory and cardiac gating.
Subjects were instructed to voluntarily gate their respiration using audible cues, so that all images were acquired during the normal respiratory pauses following tidal breathing. Images were acquired with the glottis open while subjects were at functional residual capacity (FRC). The repetition time for each successive image was set to match the measured breathing rate of the subject with a value of ∼5 s being typical. Although subjects were asked to return to FRC before the next image acquisition, they were fully allowed to vary their tidal volume as they required. Thus subjects could sigh or take deep breaths as necessary. Image acquisition and preceding tagging pulses were cardiac gated to occur during diastole using an MRI patient monitoring system (Magnitude 3150M, Invivo, Orlando, FL).
Subjects underwent three consecutive trials on the same day, each at a different fixed level of FiO2, with images for a given trial acquired during short apneas after each of 180 consecutive breaths (spaced ∼5 s apart, ∼15 min total). Subjects remained in the scanner for the duration of the study or as long as was possible to minimize repositioning, helping to ensure correct identification of the same lung slice from one trial to the next. Conditions tested were: 1) normoxia (FiO2 = 0.21, PiO2 = 150 mmHg); 2) hypoxia (FiO2 = 0.125, PiO2 = 90 mmHg, 12,500 ft altitude equivalent); and 3) hyperoxia sufficient to completely abolish HPV, but preclude the development of absorption atelectasis (FiO2 = 0.30, PiO2 = 214 mmHg). A computer simulation by Joyce et al. (27) predicts that the time required to develop atelectasis at an FiO2 of 0.30 is on the order of 200 min, an order of magnitude larger than the trial length. The order of gas presentation was balanced between subjects so that any potential influence of presentation order on the experimental results would be minimized (e.g., some were presented with hypoxia first, others normoxia first, and a third group hyperoxia first). Also note that all three gases were delivered from gas cylinders containing an oxygen/balance nitrogen mixture, and thus lacked humidity.
A trial was divided into three equal length parts; an initial “baseline” block, a middle “challenge” block during which NO was added to the breathing gas at 40 ppm, and a final “recovery” block after cessation of iNO delivery. NO was added to the inspired line by flowing a carrier gas through a clinical iNO delivery system (INOMAX DSIR, Ikaria, Hampton, NJ), and directing the output into a port mounted on the face mask. The carrier gas was switched between runs to match that used to fill the breathing bag (e.g., hypoxic carrier gas during the hypoxic run). To ensure that subjects received the full 40 ppm NO concentration, carrier gas was added to excess during the challenge block such that the flow rate through the NO delivery system was well above the measured minute ventilation of the subject during the baseline period. This configuration allowed for rapid (1–2 breaths) activation and deactivation of inspired NO, but precluded accurate measurement of expired minute ventilation during NO administration due to the flow of NO containing gas into the expiratory limb of the breathing circuit.
The delivery system included safety monitors (for gas sampled at the mouth) to ensure accurate NO delivery and detect the generation of NO2. With the high carrier gas flow rates used, measured NO only reflected the inspired gas concentration. As any excess NO was released into the scanner console room after passing down the expiratory line, the delivery system monitor was also used to periodically check ambient NO and NO2 accumulation.
The ASL method has been successfully and extensively used by us (2, 5, 20, 21, 23, 24, 38) in prior studies, and we employed the same technique here. Briefly, we capture two images of the same lung slice, in which the magnetization of protons outside the slice are differentially modulated while magnetization of protons within the slice are treated identically. Subtraction of these two image types, one which reflects both static tissue signal recovery and signal enhancement from inflowing blood (a “blood-bright” or “control” image), and the other which reflects primarily static tissue recovery (a “blood-dark” or “tag” image), generates a single difference (or ASL) image with signal proportional to blood flow. As in prior studies, we used the FAIRER (flow-sensitive alternating inversion recovery with an extra radiofrequency pulse) variation of ASL (7).
All images were acquired in a single sagittal slice of the right lung positioned approximately where the apical-basal dimension of the lung was maximal. We referenced the location of the slice to the spine on the localizer images to ensure comparable scan locations after any subject repositioning. Parameters for each image were as described in Ref. 5 and resulted in images at a resolution of 256 × 128 on a 40 cm field of view, with a 10 mm slice thickness. Thus a single voxel before processing (see below) was ∼45 mm3 in volume. The temporal acquisition protocol was based on one used in our prior study (5) updated to achieve an effective sample rate of ∼1 image per breath (compared with the prior sample rate of 1 image every 2 breaths), achieved by increasing the ratio of blood-bright to blood-dark images from 1:1 to 9:1. Experiments using previously collected data showed that variation in blood-dark images collected during the same state (e.g., hypoxia) is within the level of measurement noise, making a 9:1 collection ratio more efficient while maintaining sufficient redundancy in case of a mistimed image acquisition.
Data Processing Path
Our data processing path was previously described in detail in Ref. 5 using custom algorithms written in Matlab (The Mathworks, Natick, MA). For ease of reference, an overview is provided below. The end result of these preprocessing steps is a time series of blood flow maps, each normalized such that the mean map intensity equals 1, with a spatial resolution of ∼1 cm3 and a temporal resolution of ∼5 s.
Image quality control.
A quality control process was implemented to remove images in which considerable deviation from FRC or evidence of cardiac gating failure was observed. The criteria for exclusion were qualitative and included the appearance of significant MR artifact related to motion, features that are easily visually distinguishable. For example, motion often leads to image phase shifts that displace the subclavian and hepatic vessels from their normal anatomical location, or substantially smear focal areas of previously high signal intensity. Those responsible for image quality control were not blinded to subject or condition, as it was not practical to do so. Images that did not pass inspection were removed from the time series, and not included in any subsequent calculations.
Image registration and subtraction.
An image registration algorithm was then applied to the data to bring image voxels into alignment over the course of a particular FiO2 run (3 per subject). After passing initial inspection, remaining images for each run were registered to a reference lung volume using the boundaries of the lung in each image. The reference volume was trial specific (i.e., was not the same between subjects) and was constructed by temporally averaging the baseline (pre-NO) portion of an experimental run. Areas contained within the lung field were then moved as if they lay on a linearly deformable isotropic rubber-sheet that had been stretched between the lung boundaries. After registration, ASL image series were constructed by taking each blood-bright image and subtracting a representative blood-dark image from within the same test condition (in this case FiO2 and NO level). This representative blood-dark image was computed by temporally averaging all blood-dark images acquired during the respective test block that had passed visual inspection (those acquired at approximately FRC).
Removing signal from large conduit vessels.
The ASL image contains contributions from both blood that is or will be delivered to the capillary bed within the image plane, and from larger “conduit” vessels that traverse the image plane, but which contain blood destined for other regions of the lung. As blood in these conduit vessels is not reflective of slice perfusion, we employed the technique described in Burrowes et al. (8) to reduce their contribution to measures of PBF. In short, very high flow voxels in the lung field, defined as those with signal exceeding 35% of the maximum signal measured, were masked out. The maximum signal in the slice corresponds to those voxels completely filled with tagged blood from outside the image plane, and whose signal cannot be assumed to represent actual flow. To ensure consistent masking within a FiO2 condition, all of the registered ASL images for that trial were averaged to determine a mean ASL image, from which the mask was then constructed and applied to the entire series.
A smoothing procedure was implemented to help mitigate any uncorrected small misregistration between images. After large vessel masking, the ASL images were smoothed using the same algorithm as in our previous study (5) and a Gaussian kernel with a full-width half-maximum of 7 voxels. A correction was then applied to remove the effect of smoothing across mask boundaries. The effective spatial resolution after processing was ∼1 cm3, or on the order of a few acini.
Normalizing each ASL image.
In generating measures of fluctuation heterogeneity (see Data Metrics below) all masked and smoothed ASL images were normalized so that the average value of each image was 1. This normalization is given by Eq. 1, where S represents the signal in the image, the sum is conducted over voxels in the lung field, and n is the number of lung voxels. The rationale for this normalization was discussed at length in (5). In essence, this normalization procedure has a twofold beneficial effect. First, any global change in blood delivery (e.g., from a change in cardiac stroke volume) over time is essentially eliminated, leaving measures of spatiotemporal heterogeneity that reflect regional vascular activity. Note that iNO is not expected to cause large changes in cardiac output or HR (19, 40), and this normalization may help to reduce any remaining stroke to stroke variation between images. Second, any changes in the gross MRI signal that would arise from T1 or T2 effects as FiO2 changes, or from electronic drift in the scanner itself, are removed, making value comparisons between different physiologic states possible. (1)
For purposes of comparison, we have chosen to focus on the principal metrics used in our prior publication (5). A synopsis of each is presented below. It should be noted that unless otherwise specified these metrics represent a time course over each experimental run. In such instances, the value assigned to a particular block was the temporal average over the last 2/3 of that period, with the first third used to reach an approximate steady state.
Spatial relative dispersion.
Relative dispersion was calculated for each ASL image. As ASL images were normalized to a mean of 1.0, the spatial RD at any time point is then simply the standard deviation across the lung field, and is given by Eq. 2 where NS is the normalized ASL signal at time t located within the mask at coordinates (x,y), the sum is over voxels in the lung field, and n is the number of lung voxels. (2)
Fluctuation dispersion (FD).
We previously demonstrated that while hypoxia does not have a discernable effect on overall spatial blood flow heterogeneity (spatial RD) for healthy subjects, it does induce changes in the heterogeneity of blood flow fluctuations (5). Fluctuation heterogeneity was determined by removing a baseline (or reference) flow pattern from the normalized ASL series, and computing the relative dispersion on these fluctuations. Note that due to the normalization, the relative dispersion is again simply the spatial standard deviation of fluctuations (and the fluctuation images have a mean of zero). (3)
Two different protocols were employed to select the reference pattern (NSref in Eq. 3) and generate fluctuation time series, leading to two different values of FD at every time point. For the first protocol, identified by the subscript Global, the initial or baseline flow in the slice was removed from the rest of the time series, leaving maps that characterized how much flow changed from the start of a trial. The initial flow was calculated for each trial by averaging together images 11–20, the period in which subjects typically begin to reach a steady state end-tidal volume. Using this initial flow pattern, FDGlobal was then calculated image by image for the entire time series, before finally producing three block averages (one each for baseline, challenge, and recovery periods) utilizing the last 40 breaths of each block (a quasi steady-state) to compute the respective averages. In the second protocol, identified by the subscript Local, the mean flow of each block was removed from that block's own time series, leaving maps of fluctuations about the steady states achieved under each of the baseline, challenge, and recovery periods. The values of NSref for FDGlobal and FDLocal are given by Eqs. 4 and 5, respectively, where time t is measured in terms of the image number in the series (approximately one image every 5 s). (4) (5)
Long-range trends in FDGlobal time series were analyzed as a function of FiO2. For each trial (3 per subject), the challenge block was masked out and a straight line was fit through the remaining data points, with the fit slope taken as a measurement of the linear trend present.
Effects on Regional Pulmonary Blood Flow Patterns
Changes in the underlying pulmonary blood flow pattern were analyzed on a group level by segmenting the lung, and comparing block average flow signal (using the normalized series) for the same segment across subjects. Two different methods of segmentation were used, referred to as “gravitational” and “lobar” segmentation. In the first case, the lung was divided into 9 regions by indexing the maximum anterior-posterior and cranial-caudal dimensions by thirds, effectively superimposing a 3 × 3 grid onto the lung field. In the second case, the lung was anatomically divided into lobes by segmenting along the horizontal and oblique fissures visible in the unprocessed images. For visualization purposes, group change maps for each condition were constructed by taking the mean change value for the same segment across subjects, and plotting that value back into a single representative lung slice.
For each data condition the average values of RD and FD and regional relative pulmonary blood flow were calculated. After testing for normality across subjects using the Shapiro-Wilks procedure (implemented in R), an ANOVA for repeated measures was used (Statview) to statistically compare changes in these dependent variables. As the Shapiro-Wilks procedure requires the use of multiple comparisons, the family-wise error rate for deviations from normality was set to P = 0.05 on a per metric level (FDGlobal, FDLocal, Spatial RD and regional flow), and was controlled using the Bonferonni-Holm method. Main effects tested were the inspired oxygen concentration used in each run (FiO2; three levels: 0.125 hypoxia, 0.21 normoxia, 0.3 hyperoxia) and the phase of the experiment during which the condition occurred (two levels: −NO and +NO). For the pulmonary blood flow pattern maps, an additional main effect corresponding to lung region was added to the analysis (9 regions arranged in a 3 × 3 grid, or 3 lobes: upper, middle, and lower). Interactions between the main effects indicated differences in the degree to which dependent variables changed with NO between the three levels of FiO2. If a significant omnibus F was found, post hoc testing was performed using the Fisher's protected least significant difference test unless otherwise specified. Significant differences were accepted at P < 0.05, two-tailed.
Physiological responses and the responses of pulmonary blood flow are shown in Table 2. The Shapiro-Wilks normality test did not reveal any significant deviations from the normal distribution for subject mean data. The 95% confidence interval for lung volume variation was within ±6% of a subject's average lung volume and was the same across the conditions tested. HR and SpO2 were unaffected by the addition of iNO, and the time courses of both remained stable over the length of each trial (P = 0.65 and P = 0.69). Overall, changes in spatial RD and FDLocal were nonsignificant, whereas FDGlobal varied as a function of FiO2 and iNO, signifying a change in the spatial pattern of blood flow relative to baseline (P = 0.0001). Regional analysis indicated that iNO increased flow to the dependent lung and decreased flow to the nondependent lung while subjects were either hypoxic (P = 0.007) or normoxic (P = 0.008), but not while hyperoxic (P = 0.64). Statistically significant deviations from normality were not detected in any of the tested metrics. Detailed findings are given below.
FDGlobal demonstrated an upward drift over time of 0.73% per minute during the non-iNO portions of the experiment. As this effect was not dependent upon FiO2 (P = 0.98), a linear trend fit was removed from the data prior to further analysis, facilitating baseline vs. challenge and baseline vs. recovery comparisons. Analysis of variance revealed significant effects of FiO2 (P = 0.005), experimental phase (i.e., baseline, challenge, recovery, P = 0.001), and FiO2 × phase interactions (P = 0.0001), the latter suggesting a differential effect of FiO2 on response to iNO. FDGlobal values for baseline and recovery blocks did not differ from one another (P = 0.86), but there were significant differences between the baseline and challenge blocks (P = 0.0009), owing to a 5.4 percentage point increase in FDGlobal over baseline when iNO was administered during hypoxia (P = 0.003). This difference over baseline in hypoxia was significantly different from the differences over baseline observed in either normoxia (P = 0.0006) or hyperoxia (P = 0.0008).
iNO administration during normoxic and hyperoxic stimuli did not elicit a measurable response in FDGlobal (P = 0.76 and P = 0.84, respectively). Figure 1, A–C, shows individual subject responses of FDGlobal (using block averages) at each oxygen level and serves to highlight the effect of iNO during hypoxia. Subjects exhibited a similar response in FDGlobal to that represented by the group average, and subject responses were comparable to one another. The group average time course of FDGlobal is displayed in Fig. 2, at each of the three levels of FiO2.
In contrast to FDGlobal, FDLocal (Fig. 1, D–F) showed no effect from the addition of iNO under any of the FiO2 conditions studied. Further, the individual responses varied widely, with some subjects displaying an increase in temporal variability in response to iNO, whereas others showed a decrease. As a consequence, no statistically significant effect on temporal variability (via FDLocal) was detected (P = 0.99 for FiO2 × phase interaction). Similarly, changes in spatial RD as a result of iNO administration were small, varied among subjects, and were not statistically significant (P = 0.50).
The maps displayed in Fig. 3 (for gravitational segmentation) and Fig. 4 (for lobar segmentation) show the spatial changes in temporal mean slice blood flow between the iNO and baseline blocks at each of the three FiO2 values. When segmented gravitationally, three-factor repeated measures ANOVA was significant for between-regions main effect (P < 0.0001), a significant phase by region interaction (P = 0.004), and there was a significant FiO2 by phase by region interaction (P = 0.02). These significant three-factor interactions indicate regional differences in the response to iNO that varied as a function of FiO2.
Post hoc analysis was performed by considering each FiO2 level separately to determine significance of redistribution from baseline, and in pairs to determine whether the degree of redistribution differed as a function of FiO2. Whereas both hypoxia (P = 0.007) and normoxia (P = 0.008) showed regional differences in the PBF response to iNO, no such differences were seen in hyperoxia (P = 0.64). Regional redistribution with iNO in settings of hypoxia and normoxia were not significantly different from one another (P = 0.26); however, both showed significant differences from the pattern of redistribution in hyperoxia (hypoxia vs. hyperoxia P = 0.02; normoxia vs. hyperoxia P = 0.02). Further, after cessation of iNO the pattern of blood flow returned to its baseline state, with no significant differences in regional flow between the recovery and baseline periods (P = 0.98, Fig. 3B). Analysis of the same data using a lobar segmentation approach (Fig. 4) did not show any statistically significant dependence of lobar flow on iNO or FiO2 status (FiO2 by lobe P = 0.28, phase by lobe P = 0.31, and FiO2 by phase by lobe P = 0.15).
The principal findings of this study are that iNO administration affects the pattern of pulmonary blood flow while breathing 12.5% and 21% O2. This is most clearly evident in the hypoxic setting, where spatiotemporal variability (measured by FDGlobal) was markedly and rapidly elevated after the start of the challenge block. As a reminder, the difference between FDGlobal and the more familiar relative dispersion (RD, alternately coefficient of variation) is that FDGlobal focuses on changes in regional flow relative to a baseline pattern (or initial time), and therefore removes the static heterogeneity that may otherwise obscure small changes. Although the change in normoxia does not appear statistically different from that in hyperoxia, as both follow a similar trend over time, further regional analysis supports the assertion that both hypoxia and normoxia show a pattern of spatial redistribution in response to iNO that is different than that observed in hyperoxia (no apparent change), where the potential for hypoxic pulmonary vasoconstriction would be expected to be minimized.
Neither within-block temporal variability (measured with FDLocal) nor overall spatial heterogeneity (measured with relative dispersion) was significantly altered by the administration of iNO or by differences in FiO2. That overall spatial heterogeneity was unaffected by vasodilation with iNO is not at all surprising given the similar lack of effect with onset of HPV in our previous study (5). Yet, that no effect on temporal variability was observed with iNO was unexpected, given that we had previously observed an increase in temporal variability with onset of HPV (5). This could suggest that a third as yet uncontrolled factor may be responsible for controlling temporal changes, for example local Pco2. Increased ventilation as a result of the switch from air to hypoxia in the previous study would be expected to reduce end-tidal Pco2 by ∼6 mmHg. Due to the method of iNO delivery we were unable to determine how ventilation may have been affected by iNO, if at all, although meaningful ventilatory changes were not expected. If the pulmonary circulation were to rely on local Pco2 to fine-tune alveolar V̇a/Q̇, then one could imagine that changing Pco2 might manifest as the increased temporal variability previously observed, and explain why temporal differences would not be seen in this experiment. While this theory is clearly speculative at this time, the role of CO2 in regulating vascular spatiotemporal dynamics would be an interesting area for future study.
The constellation of findings in this study are consistent with an overall shift in the pattern of blood flow as a consequence of iNO administration in hypoxic and normoxic conditions. In hypoxia and normoxia, administration of iNO appears to increase blood flow to the dependent lung at the expense of nondependent lung, with these changes in flow occurring at a regional (sublobar) level. As was the case in our prior studies, simply relying on overall spatial metrics of pulmonary blood flow distribution such as relative dispersion was not adequate to capture changes in flow resulting from iNO, as overall heterogeneity (RD) was maintained despite substantial underlying differences in the way blood flow was distributed throughout the lung slice. Rather, it appears a more subtle approach is required.
FDGlobal elevation without concomitant increases in FDLocal suggests that iNO elicits a change in spatiotemporal variability that does not result from a temporal process. Thus spatiotemporal dissimilarity detected by FDGlobal must result from a spatial redistribution of blood flow within the imaged lung slice, with iNO altering the steady-state flow pattern during administration. The inclusion of temporal information allows one to specifically remove static components of heterogeneity that may otherwise mask changes in flow distribution, and perhaps ultimately in the future better characterize the time constants of that response. Indeed there is the suggestion from Fig. 2 that the response to iNO on average occurred rapidly, over just a few seconds, whereas the recovery after iNO cessation required approximately 2–3 min for subjects to return to baseline, and presumably HPV to reassert itself after having been disengaged. However, data loss in this study precluded making any more detailed measurements of the rise/decay time of the response on a per-subject basis.
That FD is capable of distinguishing changes to which RD is insensitive certainly appears promising, but it should be noted that as of yet these parameters do not themselves give direct insight into the mechanisms underlying blood flow control, nor can they clearly distinguish in what manner the distribution of flow itself changes, instead only signifying that a change occurred. Until such time as more advanced statistical models can be developed to supplant FD and offer additional insight and sensitivity, we must instead turn to other analytic techniques to further identify the way in which the flow pattern is altered by iNO.
Figure 3 highlights the pattern of flow changes we observed with iNO administration as a function of FiO2, averaged across subjects, and plotted back into a representative lung slice for demonstration purposes. Segmentation into predefined regions was necessary to account for normal anatomic variation in lung shape between subjects. In this case, a 3 × 3 grid was superimposed by using the largest cranial-caudal extent of the lung slice as the long axis, and the largest antero-posterior dimension as the short axis, dividing by thirds along each. When analyzed in this fashion, the differential effect of FiO2 on the ability of iNO to redistribute flow is striking. Flow is clearly redistributed in hypoxia, whereas in the hyperoxic case almost no redistribution occurs. In addition, moderate and statistically significant (P = 0.008) redistribution also appears to occur in normoxia, which may be indicative that there is in fact some normoxic vascular tone governing blood flow in the lung. Further, that the regional redistribution in normoxia and hypoxia does not appear different from one another (P = 0.26), but both are quite different from hyperoxia (hypoxia vs. hyperoxia P = 0.02; normoxia vs. hyperoxia P = 0.02) suggests more similarity in vascular tone between the normoxic and hypoxic lung than between either of these and the fully relaxed, hyperoxic lung.
These findings are consistent with the premise that spatiotemporal changes observed in this study stem from the presence or absence of HPV-induced vasomotor tone at baseline (the pre-NO period). Although the specific range of PaO2 over which HPV is active in humans is still an area of active discussion, it is generally accepted that local PaO2 values below 60 mmHg are sufficient to trigger demonstrable hemodynamic changes consistent with HPV (17). Using a similar 12.5% inspired oxygen challenge in human subjects, Mélot et al. (31) showed a significant increase in mean PA pressures from 13.3 mmHg breathing air to 18.6 mmHg during hypoxia. Similar changes were noted by Frostell et al. (14) using 12% O2, who additionally showed that the administration of iNO at 40 ppm was sufficient to return PA pressures to their normoxic levels.
By implementing a compartmental pulmonary gas exchange computer model described by West and Wagner (48) we can calculate the expected range of PaO2 for a hypothetical healthy lung at each FiO2 employed in this study. We used as input to the model typical normal values for physiological parameters (V̇o2 = 300 ml/min, V̇co2 = 250 ml/min, Hct = 40, Hb = 15 g/dl, etc.) and defined healthy V̇a/Q̇ dispersion in the supine lung (logSDV̇a and logSDQ̇) to have a value of 0.35, approximately equal to supine V̇a/Q̇ heterogeneity measured by Henderson et al. (21). All of these parameters were held constant across each of the three simulations, with only the FiO2 varied between them. The mixed venous point however was not an independent variable, but was calculated based upon the input V̇o2 and V̇co2, the resulting respiratory gas exchange for each condition tested, and conservation of mass. Figure 5 shows the model output in each condition, plotted as a cumulative distribution function on the basis of pulmonary blood flow, and it is clear that in the hypoxic lung, nearly all of pulmonary blood flow goes to compartments with PaO2 values less than 60 mmHg, in the agreed upon realm of active HPV.
The data strongly suggest that iNO has the potential to alter pulmonary blood flow in hypoxia owing to underlying HPV activity. This is not entirely surprising given iNO's established role as a pulmonary vasodilator (1, 3, 42). What is of considerably more interest, however, is that iNO appeared to have a reduced (compared with hypoxia) but still statistically significant effect on the spatial distribution of blood flow in the normoxic lung despite the conventional thinking that, in healthy individuals, HPV should not be active (14, 16, 36). Such a finding does, in fact, have some basis in HPV literature. Rahn and Bahnson (39) demonstrated that unilateral right lung hyperoxia in dogs (with 30% O2) caused a decrease in flow through the left lung from ∼40% to ∼30% of total cardiac output, suggestive of a baseline level of normoxic vascular tone. They then went on to measure the HPV stimulus-response curve by altering the FiO2 of one lung, and in contrast to defining a particular threshold oxygen tension for HPV, found the greatest changes in blood flow occurred between PaO2 60 and 120 mmHg, the normal physiological range (39).
Barer et al. (6) later did a similar stimulus-response study in both dogs and cats, finding sharp decreases in blood flow in both species below 100 mmHg at a rate of 15.7% per 20 mmHg in cats and 11.8% per 20 mmHg in dogs. Utilizing the techniques of control theory, Grant et al. (18) calculated that HPV in the coatimundi (a raccoon-like mammal from south America) was most effective at controlling V̇a/Q̇ mismatch within the range of 65–85 mmHg, and Mélot et al. (31) calculated that in humans HPV provided feedback effects on V̇a/Q̇ control between V̇a/Q̇ ratios of 0.1 to 1 (corresponding to PaO2 values of ∼40–100 mmHg breathing air) with a maximum gain at V̇a/Q̇ = 0.4 (PaO2 of ∼60 mmHg). If we take from these studies a conservative PaO2 figure of 85 mmHg as the realm in which constriction begins in response to lowered oxygen tension, we find from Fig. 5 that even in normoxia, 20% of pulmonary perfusion may be directed at lung regions experiencing active HPV. In contrast, the use of 30% O2 raises PaO2 above ∼140 mmHg in ∼95% of the lung, well outside the range of HPV sensitivity.
That flow redistribution appears to occur in not just hypoxia, but normoxia as well, suggests that vasomotor tone present in healthy normoxic adults could be contributing to V̇a/Q̇ matching on a routine basis. It has been well established that there exists gravitational heterogeneity in V̇a/Q̇ ratio (46), and that V̇a/Q̇ ratio ultimately determines PaO2 (48). Utilizing MR based measures of ventilation and perfusion in the lung, Henderson et al. (21) showed that such gravitationally induced V̇a/Q̇ heterogeneity remains present in supine posture, with the lowest V̇a/Q̇ ratios found toward the dependent, with the correspondingly lowest PaO2 values. If present, HPV would be expected to be most active in areas of low Po2 and administration of iNO would then lead to flow increases in dependent lung at the expense of nondependent lung. Therefore, it does seem plausible that active V̇a/Q̇ matching, mediated by HPV, could be occurring. The net result appears that vasomotor tone driven by HPV corrects some of the gravitationally induced V̇a/Q̇ heterogeneity in normal subjects.
The assertion that HPV-mediated vascular tone is driving iNO-mediated flow redistribution is bolstered by the comparably greater effect in hypoxia, and apparent absence of this effect while breathing 30% O2. This suggests that hyperoxia led to the least amount of vascular tone during the pre-NO period, such that vessels were already near fully dilated before iNO administration, and that vascular tone is not completely abolished in normoxic subjects at baseline. This observation also appears consistent with our previous study (5). In that study, switching subjects from breathing room air to 100% oxygen induced a change in the pattern of blood flow (measured with FDGlobal), which one would expect if the normoxic lung still possessed some vascular tone no longer present in hyperoxia. When taken together, these data point to the similar action of oxygen and a known chemical vasodilator (NO) on air breathing subjects, bringing HPV activity in normoxia to the forefront as the common mechanism linking these observations.
An argument could be made that the flow redistribution observed might have more to do with the anatomy of the vascular tree than with gravitational effects on V̇a/Q̇ ratio. As a second and parallel analysis approach we utilized visible lobar boundaries corresponding to the horizontal and oblique fissures, easily seen in unsmoothed blood flow images, to segment the lung into three lobar regions and analyze flow pattern changes across subjects. The results of this approach are demonstrated in Fig. 4. When segmented anatomically, neither FiO2 nor NO were found to have any significant influence on pulmonary blood flow. We posit that this indicates flow redistribution with iNO occurs on a smaller functional scale than is captured by gross lobar measurements, and hence large-scale anatomy alone cannot fully explain the changes observed. This is consistent with the generally accepted view that HPV occurs at the level of the small precapillary arterioles (13).
Limitations and Technical Considerations
Normality of the underlying data.
Analysis of variance relies on the assumption that the data to be tested are normally distributed. Data generated from ratios may risk violation of that assumption. Thus we employed the Shapiro-Wilk test in R as a means of attempting to detect large deviations from normality in the data that would make the use of an ANOVA an inappropriate choice. No such deviations from normality were found in our data, but this may not be entirely surprising given the limited sample size (N = 5). Hence, as an additional cautionary measure we employed a bootstrapping procedure to generate a statistical test for changes in FDGlobal, an approach that does not require any assumption about the underlying distribution.
Bootstrapping provides a nonparametric means to generate confidence intervals and/or conduct hypothesis tests. The general framework is described in detail in Ref. 30. In brief, the underlying distribution of the test statistic is approximated by simulating repeated resampling of the data, where each simulated data set (of equal size to the original) is constructed by randomly drawing from the observed data with replacement. The statistic of interest is then computed on each simulated resampling, and the resulting distribution of that statistic from among all the resamples is used to estimate the true sampling distribution. The bootstrap distribution is then used to estimate the probability of observing a deviation from the null hypothesis as large as seen in the original data.
Specifically, each detrended individual subject time course of FDGlobal was randomly resampled with replacement to produce 10,000 simulated sets of subject data (at each FiO2). To estimate the distribution for the null hypothesis of no difference in FDGlobal between blocks in the same trial (or no difference in the change between blocks as a function of FiO2), the block structure of the data was not preserved during the resampling of a given time series. Note that the null hypothesis of no block differences implies that a particular value of FDGlobal occurred in a particular block by chance and not as a result of the experimental conditions associated with that block. Thus, for example, a valid resample may consist of data points in the baseline block that were originally measured in any of the three blocks (baseline, challenge, or recovery). The subject and trial structure of the original data were both maintained during resampling (e.g., subject A's data did not appear in a simulated data set for subject B, and data acquired in the hypoxic run did not appear in a simulated normoxic run).
The mean changes over baseline in FDGlobal induced by iNO, and the mean difference in these changes as a function of FiO2, were calculated from each of the 10,000 simulated data sets, producing an estimate of the probability distribution corresponding to no change. P values were then assigned based on the proportion of simulations resulting in an observation equal to or greater than that found in the original data. The change in FDGlobal over baseline with iNO during hypoxia was statistically significant, as was the difference between this change and those observed in normoxia and hyperoxia (all at the P < 0.0001 level). As was the case with the parametric statistical results, changes over baseline in normoxia and hyperoxia were not significant (P = 0.75 and P = 0.83, respectively).
Effects of altered CO2.
In interpreting these data, it is important to bear in mind that these experiments were performed in a poikilocapnic fashion, with Pco2 allowed to freely vary. Both Pco2 and Po2 are known to modulate pulmonary vascular tone (12), and Croft et al. (10) have shown that these effects are additive, with hypercapnia and hypoxia both promoting vasoconstriction. Therefore, it is possible that some of the effects noted may be an indirect result of ventilation-induced changes in Pco2 as a function of FiO2, or that the full magnitude of any O2-mediated effects may be blunted, as hypocapnia would be expected to accompany hypoxia in this protocol, and vice versa. Although we did not explicitly measure arterial Pco2 in this study, unpublished data from a prior study by our group in which subjects lay supine and were given the same gas mixtures over similar duration suggests that Pco2 is only minimally altered by our protocol [FiO2 = 0.125, PaCO2 = 36.4 (1.7) mmHg; FiO2= 0.21, PaCO2 = 38.7 (1.5) mmHg; FiO2 = 0.3, PaCO2 = 38.4 (2.9) mmHg]. Given that the vasomotor response to CO2 is thought to be considerably less than the response to O2 (6), and that the expected alterations in CO2 are small, the potential for these changes to significantly influence the conclusions from this protocol would be expected to be minimal.
Possible effects of iNO on bronchomotor tone.
Although it is true that NO has a bronchodilator effect on constricted airway smooth muscle, it is unlikely that it had much if any impact on our results. Subjects for this study were chosen from a population of healthy young, athletic adults with no known history of pulmonary disease, and the literature supports the conclusion that iNO, even at twice the dose used in this study, does not have a measurable effect on airway conductance in healthy subjects (25).
Comparison of measurements between different values of FiO2.
In interpreting these results it should be pointed out that the experimental design utilized here focuses on detecting changes from a baseline (pre-iNO) condition. In the case of FDGlobal in particular it is the change in the parameter, not the value itself, which is indicative of spatiotemporal variability. This is a direct result of the calculation method, which uses a temporal average flow map from the initial period of each run to remove static spatial heterogeneity, and makes FDGlobal a function of both the current and initial state. Therefore, it is difficult to make comparisons of the absolute value of FDGlobal reported across different oxygenation levels in this study, as we instead sought to capture the effect of iNO while FiO2 was held constant. This is in contrast to our prior study, in which changes in oxygenation were used as the challenge stimulus, permitting such comparisons between the challenge FiO2 and the baseline normoxic state.
Potential for cardiac and respiratory gating failure.
Data collection depends on proper timing of preparatory RF pulses and image acquisition relative to the cardiac cycle. Mistiming of the sequence generally leads to poor image quality and substantial artifact effects that hinder flow measurement. The two major contributors to sequence mistiming in this study were ECG signal quality and subject heart rate variability. Reliable ECG collection within an MRI scanner can be difficult, as even tiny motions of the connecting cables within such a large magnetic field can induce spurious signals that dwarf cardiac surface potentials, and may provide false triggers. Additionally, the activation of the gradient coils along with scanner activity makes reliable detection of QRS waves in the ECG trace challenging, as the ECG signal is lost for a brief period after every image acquisition.
It should be noted that even with a perfectly clean ECG signal, sequence mistiming could occur. The current image sequence tries to align the preparatory pulses and image acquisition with the diastolic phase by waiting a fixed delay time after detection of a QRS wave. This delay time is fixed at the beginning of the sequence at 80% of the average R-R interval observed before the start of the run. As a result, changes in subject heart rate during a run amounting to more than ±5–10 beats/min end up placing image acquisition too close to systole, during which there is substantial cardiac motion and pulsatile flow.
Motion of this type during image acquisition leads to image blurring and phase-related artifacts that clearly distinguish mistimed images from those acquired during diastole. Hence, such instances are easily recognizable by visual inspection, and their removal from the image time series is an important aspect of data quality control. Although those responsible for assessing image quality were not blinded to the experimental condition being inspected, all data quality assessments were conducted prior to the performance of any analysis. The nature of the data is such that it is not possible to gauge a priori the effect of excluding any image with good cardiac gating on our metrics of interest, and the exclusion of images with poor cardiac gating will generally result in reducing our measures of variability. Thus the logical consequence of any potential bias introduced from the quality control process is to favor the null result and nonsignificance as opposed to the artificial appearance of flow redistribution. For these reasons, we believe that data exclusion method employed was the cautious approach.
In this study, cardiac gating failure resulted in the loss of a substantial number of collected images, with one subject's data of the original 6 subjects deemed totally unusable, and ∼17.6% of the remaining five subjects' data treated as missing time points. For the worst affected subject of the remaining five, ∼51% of the data was considered lost due to overt cardiac gating failure. Meanwhile, the other four subjects had an average image rejection rate of 9.1% ± 11.9% (mean ± SD). Broken out by FiO2, data loss was 21.3% in hypoxia, 17.4% in normoxia, and 13.9% in hyperoxia. The higher propensity for data loss to occur in hypoxia is likely due to increased variability in RR interval length, as opposed to a sustained higher heart rate. Higher heart rates can be accommodated with shorter trigger delay times (except in extreme circumstances); however, this delay is fixed after the sequence begins.
Our previous study found increased temporal variability associated with 12.5% oxygen breathing (5), where in this case we failed to detect any significant differences in FDLocal as a result of FiO2 during the pre-NO periods. It is likely that cardiac gating difficulty in this study had some impact on our measures of FDLocal, and hence our inability to discern changes in temporal variability is not altogether surprising. Flow pattern changes are however considerably larger in magnitude than changes in temporal variability itself and are hence less susceptible to noise in the measurement process. In the future it may be possible to reduce some of the potential for sequence mistiming by adding the capacity to automatically update the delay time based on a running average heart rate during the sequence.
The data collection also depends on the subjects being at or near FRC for each image acquisition (every ∼5 s) requiring active voluntary respiratory gating over a long period of time (∼20+ min per run). This requires the subjects to focus on the scanner activity for a long period of time, and attempt to return to a consistent lung volume throughout the experiment. In practice, we have found that with a few short training runs, subjects can achieve this reliably and repeatably, and overtly bad breath-holds were rare. A deformable image registration algorithm described in detail (5) was used to reduce any remaining small image misalignments but assumes a linear displacement field with the edges fit by three Bezier curves. The possibility exists that remaining misalignment could cause increased variability in our measurements of flow, but because we chose to focus on large lung regions when considering flow redistribution, such effects are minimal.
Use of a representative slice, normalization, and smoothing.
As was the case with our prior study of the temporal pattern of pulmonary blood flow (5), our imaging was limited to a single slice of the lung (in this case a midclavicular sagittal slice) comprising ∼8% of total lung volume. Thus it might be the case that effects in different lung regions went unnoticed, or that changes in the slice mean flow were not representative of changes in blood flow to the entire lung, both of which have the potential to impact our measures of heterogeneity. For example, in the event of a change in blood flow in only one region of the image plane that alters the image mean, but where true mean flow to the organ is preserved, changes in normalized signal in the other areas of the image that were actually unaltered will ensue. This does have the potential to artificially increase measures of flow fluctuation if the imaged slice is no longer sufficiently representative of the whole organ. However, our choice of the sagittal scan plane ensures that we sampled both gravitationally dependent and nondependent regions, over which larger differences in blood flow, more reflective of total lung heterogeneity, can be seen as opposed to isogravitational planes (2). It should also be noted that the degree of smoothing does influence the magnitude of some measures (FD and RD, for instance), but testing with multiple kernel sizes did not affect any of our statistical outcomes in a meaningful way.
A single level of NO exposure.
We employed a single, high-dose level of inhaled NO (40 ppm). This dosage is within the range of those routinely used in clinical practice to examine the pulmonary vasoreactivity of patients with suspected pulmonary arterial hypertension (15, 26, 42), and has been employed by others (14). The logistics of this particular experiment precluded any investigation of a dose-response curve.
Studying a homogeneous population.
We deliberately studied a homogeneous population of young, healthy males, as we did not have an a priori sense of the magnitude of response to iNO nor the number of subjects needed for statistical power. The observation that pulmonary arterial hypertension affects a greater preponderance of women than men may suggest sex-specific differences in the pulmonary circulation (9). Similarly, it is well known that the systemic circulation becomes less vasoactive with age (11). Whether this applies to the pulmonary circulation is unknown. Therefore, any generalization of these results to other populations should be approached with care.
Use of the fast sampling protocol.
The experimental protocol used in this study was based on our methodology from Ref. 5, altered to improve the temporal resolution of measurement from 10 to ∼5 s. In the previous study, image acquisition alternated between flow-sensitive (blood bright) images and flow-insensitive (blood-dark) images. However, further work by our group not yet published has shown that far fewer flow-insensitive images are required in the sequence, as the signal from static protons varies very little over time. In fact, tests showed that provided the blood dark image used came from the same test condition (i.e., was collected using the same FiO2 level), the results were insensitive to the choice of the image. The dependence of blood-dark image signal on test FiO2 most likely reflects underlying differences in MR relaxation rates as a function of oxygenation, where additional deoxyhemoglobin in hypoxia may increase T2 image weighting (44), whereas dissolved O2 accumulation in hyperoxia has a known T1-weighting effect (41).
Given this information, we reconfigured our sequence for this study to acquire a single blood-dark image in 10, yielding 6 such images per experimental condition (pre-NO or +iNO blocks). Only one such image per condition is actually needed; however, the 1 in 10 scheme provided additional replicates in case one or more failed to pass the quality control step. We verified that this change in protocol did not significantly affect the results of our prior study by reanalyzing those data, excluding from that data set the alternating blood dark images. There appeared to be no demonstrable trade-off in accuracy of measurement for the gained temporal resolution.
Contribution of method error to fluctuation dispersion (FD).
The analysis employed did not attempt to correct for the additional heterogeneity added by the measurement process itself. While such attempts were made in our previous study (5), the contribution of measurement noise to overall heterogeneity is sufficiently small as to be negligible. Further, unlike with temporal RD, measurement noise does not make a proportionally larger contribution to the heterogeneity of low flow voxels when assessing FDLocal. The reason for this apparent difference is that FDlocal normalizes variance (of changes during a block) across space to the mean flow of the whole slice (and not individual voxels).
However, we do believe that the linear trend in FDGlobal over time seen in every condition does reflect the accumulation of some small-scale physiological noise. FDGlobal is a measure of spatiotemporal change constructed as the standard deviation of the difference between a particular image and a baseline state (otherwise interpretable as the width of a difference distribution). For perfectly correlated images, the difference distribution is of zero width, and the width increases as images become less correlated and/or flow becomes more heterogeneous. As time progresses, small scale fluctuations in image signal intensity leads to accumulation of differences from baseline flow, and hence a wider difference distribution, analogous to a temporal random-walk.
To test this theory, FDGlobal was calculated on a random-walk simulation and a similar drift was present. We also supposed that if the drift was a result of accumulating small-scale signal differences, increased spatial smoothing would be expected to attenuate the upward trend. Reprocessing the experimental data with a range of spatial smoothing kernel sizes revealed this to be the case, with larger smoothing kernels having a greater attenuating effect until reaching ∼3 cm voxels.
As the upward trend in FDGlobal was the same across all conditions tested and present in each of the three blocks (baseline, challenge, and recovery), we applied a detrending procedure to prevent the accumulation of small-scale physiological noise alone from causing statistically significant differences in spatiotemporal heterogeneity between blocks. Note that even without detrending, the experimental design is such that differences between blocks in the hyperoxic trial (in which the upward trend is also present and an iNO effect is not expected) serve as an effective statistical control for differences observed in the other trials. With the trend included in FDGlobal, the change over baseline in hypoxia with iNO tested significantly different from the same change in hyperoxia with P < 0.01.
This study demonstrates that in our subject group of young healthy males, inhaled nitric oxide at 40 ppm significantly alters the spatial distribution of blood flow, not only under hypoxic conditions, but in normoxia as well. This suggests that vasomotor tone is indeed present during normoxic conditions, consistent with the hypothesis that HPV may play an important role in regional V̇a/Q̇ matching. iNO administration appeared to shift blood flow towards dependent lung, where presumably Po2 would be lowest due to gravitational effects on V̇a/Q̇ ratio, an effect that was abolished when breathing hyperoxic gas (30% O2). Although it has been previously reported that similar doses of iNO have no effect on PA pressure when subjects were breathing room air (14) it remains plausible that subtle changes in the distribution of pulmonary blood flow may have occurred without a significant change in PA pressure. Additionally, while we may have expected given our previous findings that administering a vasodilator during active HPV would have the effect of decreasing temporal variability this did not appear to be the case. The lung may rely on a third factor not yet accounted for to regulate temporal flow variability, such as changing levels of CO2. These results show that measures of spatiotemporal variability and regional blood flow using pulmonary ASL provides a sensitive means to study dynamic regulation of blood flow in the lungs.
This work was supported by the National Institutes of Health through National Heart, Lung, and Blood Institute Grants R01-HL-1104118 and F30-HL-110755 and by the National Space Biomedical Research Institute through NASA NCC 9-58.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: A.K.A., R.J.T., R.B.B., and G.K.P. conception and design of research; A.K.A., R.C.S., N.H.K., R.J.T., S.R.H., and G.K.P. performed experiments; A.K.A. analyzed data; A.K.A., R.C.S., S.R.H., R.B.B., and G.K.P. interpreted results of experiments; A.K.A. and G.K.P. prepared figures; A.K.A. and G.K.P. drafted manuscript; A.K.A., R.C.S., N.H.K., R.J.T., S.R.H., R.B.B., and G.K.P. edited and revised manuscript; A.K.A., R.C.S., N.H.K., R.J.T., S.R.H., R.B.B., and G.K.P. approved final version of manuscript.
We thank S. Lombardi and C. Engler for technical assistance.
- Copyright © 2015 the American Physiological Society