This fMRI study investigated the CVR among native-born HA residents using a visually cued maximum inspiration paradigm with brief breath holding. CVR was significantly reduced among the HA group; several brain regions also showed longer delay in hemodynamic response; IRV as an important aspect of pulmonary function was significantly correlated with the amount of BOLD signal change at multiple brain regions.
During the task, influx of oxygen accompanied by maximum inspiration induces prompt rise of BOLD signal and the following expiration induces drop of the BOLD signal (Figure ). In this process, cerebrovascualr properties can impact the BOLD signal; if CVR is greater, the change of the BOLD signal induced by the task shall be larger [14
]. As shown in Figure ~Figure , the HA group generally had smaller and slower BOLD response. This indicates generally decreased CVR among the HA group, which is consistent with the hypometabolism hypothesis in literature [3
]. It was proposed that during HA adaptation, hypometabolism was developed as a mechanism to cope with the hypoxic stress. The hypometabolism observed on human brains has been compared with the hypometabolism in other species that are hypoxia tolerant such as the diving seal and the aquatic turtle. Such hypometabolism was also observed in our previous two fMRI studies involving gustatory stimulation [9
] and cognitive performance [10
Because a voluntary respiration task was adapted, clusters showing significant group differences are mainly located at brain regions typically related to various aspects of respiratory regulation [27
]. The clusters showing significantly decreased CVR were located at the bilateral primary motor cortex, the right somatosensory association cortex, the right thalamus and the right caudate, the bilateral precuneus, the right cingulate gyrus and the right posterior cingulate cortex, as well as the left fusiform gyrus and the right lingual cortex (Figure ). There was also a cluster showing increased CVR at the anterior cingulate cortex (Figure ). Generally, these brain regions are either involved in respiration modulation, or are important for inhibitory control. The motor cortex controls respiratory muscles in human [25
]. The thalamus has projections to the medulla in the brainstem which controls respiration cycle [29
], and in particular, under hypoxia conditions the thalamus had inhibitory effects on the respiratory neurons of the medulla and reduces the frequency of respiration [30
]. The caudate was previously reported to be associated with voluntary control of breathing [31
]. The somatosensory regions were previously reported to be involved in inspiratory occlusion [32
]. The cingulate gyrus is well known for inhibitory control [33
]; in particular, the posterior cingulate cortex, known as a critical region of default mode [34
], is also known to be important for motor inhibition [36
]. The lingual and fusiform cortexes were activated because of the visual cues in our task; activation at the visual-related brain regions was also reported in a previous study that used visual cues for voluntary breath holding [15
]. The significant group difference at the visual-related cortex indicates that the change in CVR was not restrained to respiratory-control related brain regions, but could have been a general effect across multiple brain regions as long as a consistent regional activation was elicited. The following could be a possible interpretation for the reduced CVR at the brain regions typically involved in respiratory control: under acute exposure to hypoxic stress, the cerebral cortex would modulate relevant mechanisms to enhance ventilation; however, with prolonged chronic HA exposure during which even enhanced ventilation could not increase oxygen supply to cortex, a hypometabolism mechanism could be developed to prevent hyperventilation, so as to reserve energy and to make optimal utilization of the limited oxygen. Such hypometabolism can be pervasive throughout the brain; in the current study, it is expressed in cortical regions typically involved in respiration control and motor inhibition and even vision because these regions were activated during the task thus the BOLD signals at these regions were most reliable (from a cross-subject perspective), thus group differences were demonstrated at these regions. On the other hand, there was increased activation at the anterior cingualte cortex, which is an area generally considered to be important for attention and cognitive control [37
]; such increase can be interpreted as increased mental efforts among HA subjects to maintain attention while following task instructions; such increased attention was also previously observed during a working memory task among this population [11
The HA group showed longer delay in hemodynamic response to the task, with clusters showing significant differences located at the superior and medial frontal gyrus, the superior temporal gyrus, the postcentral gyrus, the insula, and the precuneus. The reasons for a slower hemodynamic response are complicated. It could be related to the decreased CVR; or it could be related to the reduced neuron activity among these regions. A particularly interesting result is the delay of BOLD response at the bilateral insular cortices (Figure ). It has been shown that the insular cortex plays an important role in respiratory modulation [26
]. Our previous study found reduced amount of gray matter volumes at bilateral insula [8
]. Such repeated findings of impairment at the insula points to its reduced functionality among the HA group. One plausible interpretation would be that such reduced functionality contributes to the reduced ventilation associated with long term HA adaptation, in contrast with the hyperventilation typically experienced by new comers. In future studies it would be interesting to investigate why prolonged HA exposure during early development leads to reduced volume and decreased functionality particularly at the insula.
Significant correlations were found between specific aspects of pulmonary function (IRV & ERV) and BOLD signal variation at specific brain regions. These correlations indicate the contribution of specific aspects of pulmonary function to CVR. In pulmonary function testing, the amount of air a person breathes in and out during quiet normal breathing is called the Tide Volume (VT). The additional amount a person could inhale is called the Inspiratory Reserve Volume (IRV). The additional amount a person could exhale is called the Expiratory Reserve Volume (ERV). It seems that IRV had a wide influence on BOLD signal across multiple brain regions, as shown in Figure , not only in cortices involved in respiratory modulation (e.g., the insula, thalamus, and the precentral cortex), but also other cortices such as the fusiform cortex which was activated merely due to the visual cues in the task. This indicates that IRV probably has a general impact on CVR, possibly because individual differences in IRV impacts the individual differences in the amount of intake oxygen, which further influences CVR. In particular, the correlation at bilateral insula might help to explain the longer delay of hemodynamic response among the HA group. The HA group indeed showed a larger IRV than the control group, while other aspects of pulmonary function maintained at a similar level. It could be possible that the it took a longer time for the oxygenation level of the HA group to reach peak, considering that the HA group had similar levels of respiration rate and haemoglobin concentration with the only observed difference at IRV; and this delay was eventually reflected in the delay of BOLD signal. Besides, the correlations between IRV and the BOLD-SD at the thalamus and the precentral gyrus could also be related to the involvement of these brain regions in respiratory modulation. The correlation between IRV and the BOLD-SD at the fusiform cortex, which was significant among the SL group but not the HA group, indicates that probably IRV among the HA group had a relatively more specific impact on respiration modulation, because respiratory modulation was very important for HA adaptation; whereas IRV could have a less specific (thus more general) impact on CVR among SL subjects. The correlation between ERV and the BOLD-SD at the cerebellar tonsil is harder to explain; since there was no significant group difference in ERV, and the function of cerebellar tonsil is not yet clear in literature; besides, in the scatter plot in Figure , distribution of data points in this sub-figure has a less definite pattern compared to other sub-figures, so we would take this result with caution considering our relatively small subject number. In summary, our results suggest the possible contributions of pulmonary functions in CVR; future investigation should probe into the exact physiological mechanisms of how pulmonary functions contribute to CVR.
We hope to point out that the current study purposefully controlled two factors that could have confounded group differences. Firstly, subjects of both groups were from the Han ethnic group so as to control the genetic factor between groups; in many previous studies in the HA literature, indigenous HA local residents were recruited with the control group being indigenous SL residents [3
], such experiment design makes it difficult to dissociate the contribution of genetic and developmental effect; but in our study the genetic factor was controlled as much as possible by selecting subjects from the same ethnic group. Secondly, the HA subjects had been living at SL for at least two years at the time of experiment. It was reported that peripheral physiological parameters, especially hemoglobin concentration, adapt very quickly to hypoxia/normoxia changes in the scale of weeks [51
]; and in our previous study with a larger sample size in which HA subjects had resided at SL for at least one year (in the current study the HA subjects had resided at SL for at least two years), the HA group indeed did not show significant difference on hemoglobin concentration (Table S2) [8
]. Given the controlled genetic factor, and the physiological adaptation to normoxia associated with the long term SL adaptation of our subjects, it is more likely that the observed differences demonstrated the impact of prolonged HA exposure in early childhood on brain development.
There are limitations in the research technique employed in the current study. BOLD fMRI signal comes from multiple physiological factors [12
], it does not provide direct measurement on cerebral blood flow, nor on metabolism, it cannot capture the possible differences in vascular structures either. In order to further explore the impact of HA exposure on the cerebral circulation system, future studies should attempt application of other MRI techniques, such as diffusion tensor imaging (DTI) [53
] or time-of-flight MR angiography (MRA) [54
], which can possibly reveal the fine scale differences in vascular structures; or arterial spin labelling (ASL) [56
], which provides a more direct measurement of the cerebral blood flow, etc.