Introduction
The vascular blood–brain barrier (BBB) is a vital interface that limits the unregulated transfer of circulating substances into the brain and facilitates the regulated transport of substances such as ions, nutrients, and signaling molecules that are essential for CNS homeostasis [
1]. The BBB is primarily comprised of highly specialized brain endothelial cells (BECs) that confer both barrier and interface functions. The barrier functions are attributed, in part, to tight and adherens junction proteins (TJP and AJP), which effectively seal the space between neighboring BECs to prevent paracellular diffusion [
2]. Suppression of transcellular leakage via pinocytic vesicles, expression of efflux transporters, and enzymatic degradation of blood-derived substances also contribute to the barrier properties of BECs [
3,
4]. The interface functions of the BBB include transporter-mediated passage of nutritive and regulatory substances, secretory functions, and the ability of BECs to modulate their activities in response to signals from the blood or brain compartments. Each of these functions can facilitate communication between the brain and the periphery, thus regulating CNS activities [
5].
There is mounting interest in understanding the physiology of the BBB and its pathological changes in disease contexts. The development of in vitro models of the human BBB that derive BEC-like cells (iBECs) from human induced pluripotent stem cells (hiPSCs) offers a robust platform for studying many aspects of BBB function, including interactions between BECs and other cell types of the neurovascular unit (pericytes, astrocytes, neurons) [
6], genetic contributions to BBB dysfunction, and mechanisms of BBB dysfunction in neurological diseases [
7‐
11]. iBECs express TJPs and AJPs, and functional nutrient and efflux transport systems [
12‐
15]. One major benefit of iBECs is that they develop high trans-endothelial electrical resistance (TEER) and low permeability to inert tracers, approximating in vivo barrier properties. This allows for the study of BBB transport systems with minimal confounds of leakage. In most studies, iBECs have been characterized and tested two days after subculturing the differentiated cells (post-subculture), though it has been shown that they retain strong barrier properties for at least two weeks [
14,
15]. The preservation of high TEER in iBECs suggests that they can be used for longer in vitro studies. However, further characterization of the expressional and functional properties of iBECs at extended post-subculture timepoints is needed.
In this study, we characterized the phenotypic changes that occur over time in iBECs derived from the GM25256 hiPSC line by comparing protein expression and functional outcome measures at day 2 and day 9 post-subculture. These time points were selected to reflect a time conventionally used in the literature, day 2 post-subculture, and a later time to which GM25256-derived iBECs reliably sustained high TEER, day 9 post-subculture. We first noted that the amount of cellular proliferation was significantly lower in iBECs subcultured for nine days vs. two days. Subsequently, we characterized how key aspects of the BBB phenotype are altered in iBECs’ more quiescent state. We focused on: (1) metabolic markers of endothelial quiescence, because an altered relation between glucose metabolism and transendothelial glucose transport may be important to consider when using iBECs to study dysfunctional GLUT1-mediated glucose transport at the BBB in Alzheimer’s disease (AD) [
16‐
18], (2) markers of BBB integrity, such as TEER and expression of TJP and AJP, because BEC proliferation has been associated with BBB leakiness [
19,
20], and (3) responses to inflammatory and AD-associated insults, which are insults that can happen concurrently with increased BEC proliferation [
21,
22]. Our findings indicate that length of subculture influences the baseline phenotypes of iBECs, as well as their responses to conditioned medium (CM) from hiPSC-neurons harboring the familial AD-causing Swedish mutation. These findings provide new insight into the relations between BEC proliferative state and BBB functions at baseline and in response to disease-related insults.
Discussion
A unique feature of the iBEC model is its ability to maintain high TEER over a sustained period of time in culture. The maintenance of barrier properties suggests that iBECs are amenable to longer in vitro studies, which could aid the understanding of human disease mechanisms. However, little is currently known about the stability of the iBEC phenotype with prolonged culture. As we demonstrate here, GM25256 iBECs become more quiescent over time accompanied by adoption of a metabolic phenotype that is similar to primary ECs cultured to contact-inhibition [
29]. A limitation of the current study is that we do not know whether iBECs from other commonly used iPSC lines adopt a similar quiescent phenotype, but our work is a first step in showing that GM25256-derived iBECs are suitable for studies on quiescence-related changes to BBB functions.
We first posited that factors inherent to the culture method such as nutrient availability and/or the time elapsed from bFGF and RA exposure could potentially regulate the proliferative status of iBECs [
37]. However, we found that continual medium changes (CMC) could not prevent the transition to a more quiescent state that occurs from day 2 to day 9 post-subculture, nor could the re-application of bFGF/RA fully restore the proliferative status of iBECs after prolonged culture. Only a modest increase in proliferative cells was observed when medium was changed on day 8 post-subculture. These findings indicate that the growth arrest that occurs in iBECs is stable and occurs independently of nutrient depletion or the absence of bFGF and RA.
We and others have consistently observed the preservation of high TEER in iBECs over extended periods of time in subculture [
14,
15]. Here, we demonstrate that TEER is not dependent on length of subculture, as strong barrier functions were upheld in both proliferative and more quiescent iBECs states and there was no mean difference in TEER between early and late post-subculture time points. However, the expression of some TJP and AJP was dependent on length of subculture. Firstly, we found a significant loss of CLDN5 expression over time. Previous studies have also linked increased CLDN5 expression with a more angiogenic EC state [
38]. Interestingly, CLDN5 overexpression promoted proliferation in hCMEC/D3 cells (immortalized human BEC line), while silencing it blocked cell cycle progression at the G0/G1 phase [
39]. Although decreased CLDN5 levels are commonly associated with barrier disruption [
40], upregulation of CLDN5 has also been seen in the context of diminished TEER and increased paracellular permeability [
11,
41]. Intense CLDN5 staining was observed at the junctional zones of more quiescent iBEC monolayers, supporting that, despite reduced expression, CLDN5 expression is sufficient and appropriately localized to support the maintenance of high TEER. In agreement with existing literature, we found ZO-1 expression is significantly increased in more quiescent iBECs. Reduced expression of ZO-1 generally correlates with increased cell proliferation, such as in highly proliferative BECs from human brain tumors [
42,
43]. ZO-1 has been reported to accumulate in the nucleus of proliferating epithelial cells [
44], whereas in high density confluent cells, ZO-1 is primarily localized to cell–cell junctions and inhibits cell proliferation by indirectly controlling expression of cell cycle regulators [
44,
45]. Our immunofluorescence analyses yielded clear evidence of enhanced ZO-1 junctional localization in more quiescent iBECs. We found OCLN levels did not change significantly as proliferation declined, indicating regulation of OCLN expression is likely not modulated by iBEC proliferative state.
For the AJPs evaluated, PECAM-1 expression was not significantly altered over time spent in subculture. However, we found a significant reduction in VE-cadherin expression. VE-cadherin is important in contact inhibition of proliferation [
46]. Although its expression was decreased over time, VE-cadherin junctional localization was maintained at day 9 post-subculture, raising the possibility of its involvement in iBEC adoption of quiescence. Taken together, our data underscore that changes in the levels of any one TJP do not necessarily translate to changes in barrier tightness. This concept has been previously supported in studies demonstrating modulations of BBB permeability with no changes in protein levels [
47‐
49]. How junctional remodeling in highly proliferative BECs occurs without loss of barrier integrity remains to be fully elucidated.
Prior works have characterized iBECs as primarily utilizing glycolysis vs. oxidative phosphorylation for ATP synthesis [
9], which aligns with the metabolic phenotype of primary ECs [
30,
50]. Here, we demonstrate that iBECs undergo a metabolic transition from day 2 to day 9 post-subculture that resembles that of quiescent primary ECs, which alter their metabolism from the high energy requirements of proliferation to the lower energy requirements of quiescent barrier maintenance by downregulating glycolysis [
29,
31]. The distinctive expression changes induced in primary ECs upon contact-inhibition, namely, downregulation of PFKFB3 and upregulation of GLUT1, also occur in iBECs during the transition from a highly proliferative state on day 2 post-subculture to a more quiescent state on day 9 post-subculture. Importantly, we found that nutrient replenishment with CMC did not alter the expression of PFKFB3. Nutrient replenishment did significantly suppress the increase in GLUT1 expression, indicating that GLUT1 expression is regulated by nutrient availability in addition to the proliferative status of iBECs. This finding is consistent with increased BBB GLUT1 expression in response to hypoglycemia [
51].
We further identified reduced HK2 expression and reduced MCT1 expression as two additional indicators of glycolytic repression in iBECs. HK2 is necessary to support EC proliferation as it catalyzes the first rate-limiting reaction in glycolysis, and its depletion inhibits glycolysis and impairs angiogenesis [
32,
52]. Like PFKFB3, HK2 downregulation over time was not affected by CMC. MCT1 facilitates the bidirectional transport of lactate. As MCT1 expression is induced by lactate [
53], decreased lactate production via glycolysis likely contributes the reduced MCT1 expression in more quiescent iBECs on day 9. Furthermore, lactate has been shown to function as a pro-angiogenic effector to promote proliferation in BECs [
54]. Therefore, downregulation of MCT1 may also be a mechanism to inhibit iBEC proliferation by decreasing lactate influx. CMC further suppressed the expression of MCT1, which may be attributed to removal of the lactate produced by the iBECs during glycolysis. Overall, our findings show that more quiescent iBECs adopt a metabolic phenotype reminiscent of quiescent primary ECs. Our results highlight the utility of the model as a platform to study the regulation of BBB functions under defined proliferative and metabolic states.
We confirmed a previous report of a functional and saturable iBEC glucose transport system [
54] and that a GLUT1-selective inhibitor, BAY-876, significantly inhibits glucose transport in GM25256 iBECs. Endothelial upregulation of GLUT1 that occurs concurrently with quiescence is considered to be a mechanism of prioritizing glucose delivery to the surrounding tissues by more mature cells [
29]. Therefore, we predicted that more quiescent iBECs, which were shown to have higher GLUT1 protein levels, would facilitate higher rates of glucose transport. However, the rate of glucose transport, as measured by
14C-DG Pe, was significantly lower in more quiescent iBECs vs. proliferative iBECs, which could not be explained by differences in leakage as quantified by
99mTc-DTPA Pe. Further, the glucose transport rate of iBECs on day 9 post-subculture was not significantly altered by CMC, showing that nutrient depletion does not contribute to decreased glucose transport in more quiescent iBECs.
We also found that luminal-to-abluminal
14C-DG Pe was significantly higher than abluminal-to-luminal
14C-DG Pe, which suggests that the net flux of glucose across iBEC monolayers occurs in the blood-to-brain direction. The localization of GLUT1 on the luminal vs. abluminal membrane is one mechanism by which BECs may regulate the rate of glucose transport. In animal models, the luminal: abluminal ratio of GLUT1 at the BBB has been reported as 1:4 [
55,
56], whereas in humans, the ratio was measured as 1:1 [
57,
58]. Importantly, the transcytosis of glucose in either direction should be limited by the membrane with the fewest functional glucose transporters. We observed a decrease in the transport of
14C-DG in both directions on day 9, which corresponded with an increased expression of GLUT1 and a change in its MW. This discrepancy could be explained by enhanced asymmetry or localization of GLUT1 in a subcellular compartment other than the plasma membrane. It has been shown that GLUT1 translocation to the plasma membrane is one mode of glucose transport upregulation, which can occur without changes in GLUT1 protein expression [
59]. Further, it has been shown that up to 40% of GLUT1 is sequestered in the cytoplasm of BECs [
60].
14C-DG Pe is analogous to fluorodeoxyglucose (FDG), which is an
18F-labeled PET ligand that is routinely used to evaluate glucose uptake into tissues, including the brain [
61]. Decreased brain uptake of FDG can be indicative of impaired glucose utilization by neurons, but also GLUT1 dysfunction at the BBB [
62,
63]. An advantage of using
14C-DG as a glucose surrogate is that only unmetabolized
14C-DG can fully traverse the iBEC monolayer since it becomes trapped within the cell following phosphorylation to 2-deoxyglucose-6-phosphate by hexokinase, and this form is unable to be further metabolized by glycolytic enzymes [
23,
34]. Although our assays did not have the sensitivity to detect intracellular 2-deoxyglucose-6-phosphate accumulation, we posit that the reduced transport
14C-DG in more quiescent iBECs would not be due to increased metabolism, since glycolytic enzymes, including HK2, are reduced.
In addition to observing differences in the total amount of GLUT1, we also observed the appearance of lower MW isoforms of GLUT1 in more quiescent iBECs, suggesting differences in post-translational modifications (PTM). Glycosylation is one PTM which alters the MW of GLUT1 [
64], and can also regulate its function by facilitating membrane clustering on lipid rafts [
65]. However, some of the apparent shifts in GLUT1 MW were mitigated by medium changes, which had no effect on functional glucose transport in our studies. It is also possible that other PTMs such as phosphorylation could regulate GLUT1 activity [
66]. Presently, very little is known about the mechanisms of GLUT1 regulation in proliferating vs. quiescent BECs. iBECs could potentially be used to delineate these mechanisms as a future direction.
Prior studies have shown that iBECs are responsive to pro-inflammatory cytokine stimulation within two days post-subculture [
67]. We explored the functional consequences of altered proliferation by comparing iBEC vulnerability to barrier disrupting insults on day 2 vs. day 9. TNF-α treatment produced comparable disruption in proliferative and more quiescent iBEC states. Therefore, the effects of TNF-α on TEER appear to be independent of the proliferative status of iBECs. Additionally, we modeled an Alzheimer’s brain milieu using conditioned medium (CM) from neuronal cultures differentiated from hiPSCs harboring the Swedish mutation, a familial AD-causing mutation resulting in elevated A
\(\beta\) production [
35]. We found that in proliferative iBECs, CM from hiPSC-neurons harboring the Swedish mutation (APP
Swe/+) caused a significant reduction in TEER, whereas CM from isogenic control hiPSC-neurons (APP
WT) did not change TEER. In contrast, CM from neurons derived from either hiPSC line did not have a significant effect on TEER in more quiescent iBECs. These data suggest that BBB disruption in response to this AD-associated stimulus is dependent on iBEC proliferative state.
Our results align with a previous study showing a selective vulnerability of proliferative aortic ECs to the injurious effects of A
\(\beta\) [
36] and suggest that actively dividing BECs, such as those in injured or growth states, are more vulnerable to disruption by Aβ-associated metabolites. In vitro A
\(\beta\) treatment studies generally require concentrations in to nM to µM range to induce BBB disruption [
68], whereas A
\(\beta\)-40 and -42 levels in the APP
Swe/+ CM were much lower, around 1000 pg/mL for A
\(\beta\)-40 and 100 pg/mL for A
\(\beta\)-42. It is possible that natural production by the hiPSC-derived neurons confers increased A
\(\beta\) toxicity, or that metabolites other than A
\(\beta\) produced as a result of the Swedish mutation are responsible for the barrier disrupting effects on iBECs. Interestingly, the presence of another familial AD-causing mutation in presenilin-1 or -2 in iBECs is not only associated with reduced barrier function, but also with impaired glycolysis [
9]. As a future direction, iBEC metabolism could be evaluated after CM treatment to elucidate potential metabolic mechanisms driving the specific vulnerability of iBECs on day 2 post-subculture, providing insight into the relation between BEC proliferation and BBB disruption in the context of AD. Our findings highlight that CM from iPSC-neurons can be used to study interactions of neuronal secretions with iBECs in context of AD. An additional future direction of this work could involve the co-culturing of APP
Swe/− iPSC neurons with iBECs [
69] to more accurately model the close proximity of neurons with BECs in vivo.
Both the barrier and the interface functions of the BBB are influenced by BEC proliferation. Maturation of the brain vasculature into the stable adult BBB depends on the establishment and maintenance of a predominantly quiescent BEC monolayer. On the other hand, aberrant brain angiogenesis caused by inappropriate BEC proliferation is observed in pathologies such as Alzheimer’s disease (AD), glioblastoma multiforme, and cerebrovascular malformations [
70‐
72]. A recent single-cell transcriptome analysis revealed that a subpopulation of angiogenic ECs is induced in the AD brain [
73]. A central role for angiogenic vessels in the progression of AD is suggested by their presence in brain regions affected by AD pathology [
70] and raises the possibility of inhibiting angiogenesis therapeutically [
74]. A better understanding of the differential regulatory mechanisms of BBB functions in proliferating vs. quiescent BECs would inform novel therapeutic strategies to target vascular abnormalities and BEC activation in AD and other neurological diseases.