Effective regulation of the blood supply to the brain is indicative of the overall health of the cerebral vascular system. Sickle cell disease (SCD), a genetic disorder that causes severe anemia reducing the oxygen carrying capacity of the blood is also associated with disrupted cerebrovascular function (Becklake et al., 1955; Abdu et al., 2008), which remains poorly characterized. In the presence of anemia, brain oxygen delivery is restored by increases in cerebral blood flow (CBF) (Milner, 1974; Herold et al., 1986; Bush et al., 2016). Under the assumption that the maintenance of cerebral metabolic oxygen consumption (CMRO₂) is prioritized, deficits in brain oxygen delivery result in increases in the oxygen extraction fraction (OEF) at the expense of a reduction in intracellular partial pressure of oxygen (PO₂). However, it is unknown to what extent the cerebrovascular dysfunction contributes to ischemic injuries often manifesting as silent cerebral infarcts and stroke (Switzer et al., 2006).

In patients with SCD, vasodilatory reserve in large vessels is encroached upon by the increase in resting CBF; the threshold for oxygen release from hemoglobin is raised in left-shifted oxyhemoglobin dissociation curve (Abdu et al., 2008); and increased shear stress of anemia and red cell adhesion compromise endothelial cells via nitric oxide-mediated vasodilator activity (Hoiland et al., 2020) and thereby the vasomotor regulation of the microvasculature (Hillery and Panepinto, 2004). These may result in failure of localized hemodynamic compensation for the reduced arterial oxygen content (CaO₂) of SCD and may account for the observed cognitive deficiencies and widespread ischemic injury in cerebral cortex (Kirk et al., 2009; Pimentel-Coelho et al., 2009) and watershed areas (Hillery and Panepinto, 2004; Fields et al., 2018). We therefore turned our attention to assessing cerebral vascular physiology by observing the patterns and distribution of flow responses to the application of a global vasoactive stimulus (Figure 1A).

A functional assessment of the cerebral vasculature’s regulation of perfusion has been previously tested as the perfusion response to a vasoactive challenge such as acetazolamide (Vaclavu et al., 2017; Vaclavu et al., 2019; Afzali-Hashemi et al., 2021) and hypercapnia (Nur et al., 2009; Prohovnik et al., 2009). The response can be measured in terms of surrogates such as large vessel blood flow velocity measured by trans-cranial Doppler (Nur et al., 2009) or change in Blood Oxygen-Level Dependent (BOLD) signals of magnetic resonance imaging (MRI) (Leung et al., 2016). Early measures of the CBF response to CO₂ were in terms of a simple regression slope expressed as a ratio of the change in CBF surrogate resulting from a change in the end-tidal partial pressure of CO₂ (PETCO₂), termed cerebrovascular reactivity (CVR) (see Fisher and Mikulis (2021) for historic review). In health, the reductions in resistance in all vascular beds is balanced resulting in a sigmoidal increase in flow as a function of PETCO₂ (Bhogal et al., 2015; Bhogal et al., 2016), reasonably well approximated by a linear fit. In the presence of a range of cerebrovascular pathology distributed to multiple vascular beds, cerebral vessels no longer react in a synchronized fashion, making it difficult to extrapolate from CVR to the varying stimulus-response behavior of vascular beds.

In this study we used the ramp portion of the CO₂ stimulus shown in Figure 2A to produce changes in vascular tone from vasoconstriction to vasodilation so that the full pattern of the flow response may be observed. This is a global stimulus and the high resistance of extracranial portion of cerebral vessels (Faraci et al., 1987) limit the total blood flow capacity of the brain. As the brain’s vascular beds are functionally perfused in parallel, they must compete for the restricted blood supply. In the presence of localized differences in vascular resistance, a rising global vasoactive stimulus results in the disruption of the parallel progressive reduction in resistance (or increase in flow), throughout the brain (Sobczyk et al., 2014; Fisher et al., 2018).

When viewed as a response to a progressive increase in stimulus, as in a linear increase in PETCO₂, the BOLD response patterns can be classified into four major recognized patterns (Fisher et al., 2017). The vascular beds with the most robust vasodilatory reserve retain a normal appearing sigmoidal pattern of flow. Those beds with little or no vasodilatory reserve have all their flow redistributed to the beds with robust dilatory responses. Those beds retaining a weakened vasodilatory reserve may reduce their resistance at initial vasoactive stimulation but are unable to further reduce resistance at higher CO₂ levels. Thus, they may increase their flow upon initial increases in CO₂, but at greater CO₂ levels reduce their flow in favor of the beds with stronger vasodilatory responses. This results in a biphasic, concave down flow pattern of flow. Some beds display a concave up flow pattern. They behave as if their partial pressure of carbon dioxide (PCO₂) vs. resistance curve has been shifted to the right: weak reductions in resistance at low PETCO₂ resulting in steal and rallying reductions in resistance at higher CO₂ increase flow.

This system has been modeled as an electrical analog shown in Figure 1B. Using this model, all the flow patterns of response can be translated into a single sigmoidal pattern of change in cerebrovascular resistance (Duffin et al., 2018). In this way the various patterns of flow response can be described with a common set of metrics. Figure 2B shows the resistance sigmoid, and the metrics of the fitted equation: R = S t a r t + A m p l i t u d e / ( 1 + exp ( − ( P E T C O 2 − M i d p o i n t ) / R a n g e ) ) as well as the derived metrics sensitivity and reserve. These five parameters describe the sigmoidal resistance response to CO₂ and are shown in Figure 2B as amplitude, midpoint, range, resting reserve and resting sensitivity (Duffin et al., 2018; Mcketton et al., 2019).

Here we apply the resistance model approach to assess the functioning of the cerebral vasculature in patients with SCD. The vasoactive stimulus consists of a gradually ramping increase in PETCO₂, and BOLD MRI signals are taken as surrogates of the CBF response. Maps of regional CVR as well as maps of the resistance analysis parameters were determined for SCD patients. These parameters were also measured in a healthy population to provide an atlas of the normal range of response in terms of the mean and standard deviation in a voxel for each resistance parameter; this enables the generation of z-maps (Sobczyk et al., 2015), a scoring of the patient data in terms of standard deviations from the mean of the normal cohort.

Cerebral blood flow is regulated over a range of perfusion pressures by adjusting cerebrovascular resistance with the constriction and dilation of arterioles (Klein et al., 2021). Patients or individuals with sickle cell disease compensate for their anemia by increasing CBF (Milner, 1974; Bush et al., 2016), achieved not only by vasodilation but also by an increase in the diameter of major intracranial and extracranial vessels (Guilliams et al., 2020). The main finding of this study is that the functional regulation of the cerebral vasculature of patients with SCD is impaired compared to healthy controls. Detailed findings are summarised in Tables 3, 4 with significant differences for GM and WM in all regions.

Cerebrovascular reactivity for both step and ramp CO₂ challenges is decreased in WM and GM for all ROIs, and the resistance sigmoid parameter resting vasodilatory reserve is decreased in both GM and WM in all vascular territories. These differences show that the ability to increase CBF in response to a challenge is decreased in SCD patients compared to healthy controls. Other parameters of the resistance sigmoids in the SCD patient group are also different from those of healthy controls. These include a reduced resistance sigmoid amplitude in GM, indicating that the full extent of vasodilation and vasoconstriction is restricted in SCD patients. The resistance resting sensitivity is also reduced so that the vasodilatory response to changes in PCO₂ from resting state is less. Moreover, the resistance sigmoid midpoint, the PCO₂ about which vascular changes occur, is elevated in SCD patients. This midpoint finding has implications for cerebrovascular regulation, which are discussed below.

This is the first study to measure the resistance parameters in patients with SCD. Previous studies, which examined the effects of ageing, where some aspects of vascular resistance in parts of the brain were found to be initially maintained with age but then increased in later years (Mcketton et al., 2019), also showed significant reductions in resistance sigmoid parameters in mild cognitive impairment. The same study demonstrated that the resistance measures provided physiological explanations for the various observed types of BOLD-CO₂ response patterns (Duffin et al., 2017). Here we present the first measures of resistance sigmoid parameters in response to a progressive vasoactive stimulus as a manifestation of the pathology in patients with SCD accumulated over childhood and early adulthood. As such we do not have data as to the state of resistance responses of cerebral vessels in SCD at early ages or the rate and pattern of the development of the pathology we observed. However, in this data, one may identify candidate biomarkers of disease progression, prognosis and therapeutic targets.

The virtual resistance model analysis of the CBF response to a ramp change of PETCO₂ was examined in Duffin et al. (2017), and has been described in detail in Duffin et al. (2018). In brief, the observed local (voxel) BOLD signal responses to a global CO₂ stimulus are the result of both changes in local perfusion pressure caused by network competition for a CBF supply limited by the high resistance of the major supply vessels (Faraci and Heistad, 1990), and changes in vessel diameter and therefore vessel flow resistance with CO₂ (Duffin et al., 2021). These factors cannot be separated. However, using a virtual reference resistance in one branch of the model (Figure 1) to compete with an examined voxel’s BOLD response to a ramp of CO₂ can provide a simulated network interaction change in perfusion pressure. The actual reference resistance competition for each voxel is likely different and at this point, unknowable. Consequently, the model only calculates a virtual resistance variation with CO₂ in the examined voxel that would produce the observed BOLD response if it was in competition with the reference, not the actual resistance variation with CO₂. In other words, if the voxel blood flow resulted from competing with the reference resistance, the model calculates the corresponding resistance variation with CO₂.

By making the reference resistance response to CO₂ the same for all observed voxel BOLD signal response to CO₂, the resistance pattern of variation with CO₂ can be compared between voxels because it is entirely determined by the measured BOLD response to CO₂. The model simply converts the BOLD signal variation with the global CO₂ stimulus into a more physiologically meaningful sigmoidal resistance form, where the parameters of the resistance sigmoid provide a description of the voxel BOLD response to a global CO₂ stimulus. The resistance sigmoid is still a result of both vasodilation in response to changes in PCO₂ and vascular network pressure changes, as reflected in changes in the BOLD signal.

Figure 4 shows the regional sigmoids for the SCD and HC groups in both GM and WM. The change in resistance between that indicated by the dots on each sigmoid and the minimum resistance illustrate the resting resistance reserve calculated from the regional averages. The figure illustrates the findings of Tables 3, 4 that resting resistance reserve is reduced in SCD patients compared to healthy controls, and that the resistance sigmoid midpoint PETCO₂ values for the SCD patients are higher. These two observations are linked to differences in the resting PETCO₂, which is a little higher in SCD patients (Table 2). As Figure 5 shows, resting resistance reserves and resistance sigmoid midpoints are proportionate to resting PETCO₂, while resting resistance sensitivities are not as strongly linked. We suggest the following physiological explanation.

In a model we proposed for the smooth muscle regulation of cerebral blood flow (Duffin et al., 2021), vessel tone was related to intracellular [H⁺], which was regulated to a value commensurate with normal intracellular function. Increases in arterial PCO₂ elevate the smooth muscle intracellular [H⁺], and, although intracellular acid-base regulation attempts to return [H⁺] to normal, restoration is not complete, vascular tone relaxes, and cerebral blood flow increases. When ramp increases in arterial PCO₂ are employed as a vasodilatory stimulus, smooth muscle intracellular [H⁺] increases and resistance decreases progressively to increase CBF. This decrease in resistance consumes vasodilatory reserve as measured by the resting resistance reserve. Indeed, as Figure 5 shows there is a proportionate decrease in reserve with increased resting PETCO₂. In order to restore the reserve, the sigmoid resistance relation must be right shifted to a higher PCO₂, as measured by an increase in the midpoint PCO₂. That this adaptation occurs is shown in Figure 5 where there is a proportionate increase in resistance sigmoid midpoints with increased resting PETCO₂.

Although in all the resistance sigmoid fitting the r ² quality exceeded 0.7, the quality measure was lower in the WM and GM of the patients with SCD compared to the HC group. The reasons for this difference in quality are unknown. The percent of interpolated voxels was not significantly different between the groups, and so we do not think this aspect of the analysis was responsible. Whether the difference was due to differences between the groups in the MR acquisition (e.g., motion artifact) or some physiological aspect (cerebrovascular pathology) cannot be determined. A low quality measure in a voxel reflects a low signal to noise ratio and so quality maps are used to identify regions where the resistance measures are less reliable for making conclusions.

The hemoglobinopathies in the SCD patient group varied (29 HbSS, 5 HbSC and 1 HbSD). Since it is uncertain how the BOLD response is affected for specific hemoglobinopathies we continue to interpret BOLD responses in the same way as established with HbA, but this possible confounder should be noted. Furthermore, treatments such as hydroxyurea, analgesics, and the development of other related and unrelated comorbidities and their treatment are added diversifying factors. These many differences within our SCD patient population could impose a degree of variability for which we could not control in such a small cohort, and thus certainly contributed to the obscuration of some of the characteristics of the disease. We also suggest that, because we studied adults rather than children, the distinguishing features of SCD had more time to be established, making them easier to detect in comparison to the healthy cohort. However, our study is biased toward the selection of healthier SCD patients, so that the differences in their resistance parameters from those of the healthy group are an indication of the effects of SCD without the confounding effects of other factor such as the presence of silent infarcts. The characteristics showing statistical significance should be appreciated in this regard.

Cerebrovascular reserve is reduced in patients with SCD compared to healthy controls as evidenced by decreased CVR and decreased resistance reserve at resting PETCO₂, so that these individuals unable to respond as well to a stress as healthy individuals. Moreover, the extent to which cerebrovascular resistance can be adjusted is also compromised, with a reduced resistance sigmoid amplitude. However, the mechanism of adaptation to a generally higher resting PETCO₂ appears similar to that of healthy controls as shown by the increase in resistance sigmoid midpoint.