Literature search was conducted in November 2021 on the Ovid MEDLINE, EMBASE, and Web of Science databases for entries in English from inception to search date. Briefly, the search strategy (Supplementary Table 1) used the keywords: (“motor neuron disease” or “motor neurone disease” or “amyotrophic lateral sclerosis” or “ALS” or “primary lateral sclerosis” or “PLS” or “progressive muscle atrophy” or “PMA” or “Lou Gehrig Disease”) AND (“blood brain barrier” or “BBB” or “blood spinal cord barrier” or “BSCB” or “blood cerebral spinal fluid barrier” or “blood cerebrospinal fluid barrier” or “BCSFB” or “blood central nervous system barrier” or “BCNSB”).

Four independent reviewers (Abrahao, Mirian, Moszczynski, and Soleimani) performed a primary title and abstract screening and extracted full data from eligible articles following the PRISMA extension for scoping reviews (PRISMA-ScR) guidelines (Tricco et al., 2018) and a pre-specified protocol registered with PROSPERO (CRD42017065405) to mitigate selection bias.

Included studies were original, peer-reviewed, full-text published or accepted articles assessing the following research questions: (1) in the context of ALS pathogenesis, what is known from the literature about the structure and function of the BBB and BSCB, as well as therapeutic interventions targeting these barriers, in animal and human ALS studies? (2) in the context of ALS treatment development, how does the BBB and BSCB impact therapeutic access to the CNS in animal and human ALS studies? Exclusion criteria included non-peer-reviewed articles; publications in the format of poster abstracts, editorial letters, conference papers; and articles without extractable data on ALS or its variants, primary lateral sclerosis (PLS) or progressive muscular atrophy (PMA). Studies cited in peer-reviewed review papers were also hand-searched for broader inclusion. Additional studies that fell outside the scope of the search strategy but were considered relevant were also included for the discussion of the clinical context of the BCNSB pathological findings and therapeutic delivery.

Data charting from each study was performed in duplicate by independent reviewers. Data items included the studied structure (BSCB and BBB), the research questions, interventions, experimental groups or population, sample size, animal disease stage (pre vs. symptomatic), ALS model description, BCNSB functional or structural measures, and descriptive findings. The synthesis and interpretation of the extracted data were presented in a descriptive manner without statistical inferences or meta-analyses given the largely qualitative nature of the studies in this review. Discrepancies in data extraction and synthesis were resolved by consensus decision of all reviewers. Supplementary Table 2 reports the PRISMA-ScR checklist.

In transgenic superoxide dismutase-1 (SOD1) rodent models of ALS, studies have demonstrated abnormal neurovascular unit and BCNSB ultrastructure (Garbuzova-Davis et al., 2007a,b) with upregulated pharmacoresistance mechanisms (Jablonski et al., 2012). Neurovascular unit changes include dysfunction of glia, endothelial cells, and pericytes, downregulation of tight junction expression, upregulation of drug efflux proteins, and increased paracellular permeability. TAR-DNA binding protein of 43 kDa (TDP-43), TANK-binding kinase 1 (TBK1), and Fused In Sarcoma (FUS) rodent models of ALS have also shown impairment of the BCNSB manifesting as ZO-1 adherens protein dysfunction, reduced vascular density, decreased expression of AQP4 and increased albumin levels in the CSF, among other changes (Jablonski et al., 2012; Sasaki et al., 2015; Ouali Alami et al., 2020).

Early morphological changes in the BSCB preceded motor neuron degeneration in SOD1 rodent models, including reduced capillary density and premature dissociation of the astrocyte end-foot and endothelial cells (Figure 1; Zhong et al., 2008; Miyazaki et al., 2011). Blood flow was reduced by 30–45% in the lower spinal cord prior to symptom onset in SOD1 models compared to non-diseased animals (Zhong et al., 2008). During the presymptomatic stage, activation of the platelet-derived growth factor C (PDGF-CC) pathway may contribute to BSCB leakage (Lewandowski et al., 2016). Deposits of hemosiderin, a hemoglobin degradation marker and indicator of microhemorrhage, were observed in the anterior horn of SOD1 mice (Zhong et al., 2008) and SOD1 rats (Nicaise et al., 2009a). Through BSCB dysfunction, it was proposed that iron from hemoglobin metabolism, endogenous immune or inflammatory mediators (Garbuzova-Davis et al., 2019) and reactive oxidative stress can contribute to early motor neuron damage or exacerbate ongoing motor neuron damage (Winkler et al., 2014). Furthermore, electron microscopy demonstrated focal accumulation of extracellular fluid between vessel walls and adjacent parenchyma, along with swollen astrocyte foot processes in SOD1 rodents (Garbuzova-Davis et al., 2007a).

Ultrastructural abnormalities accumulated over time in the BSCB of symptomatic SOD1 animal models as neuronal loss progressed (Garbuzova-Davis et al., 2007a). Endothelial cells showed cytoplasmic disorganization, vacuolated mitochondria, and impaired basal membrane with reduced expression of tight junction proteins claudin-5, occludin, ZO-1 and proteoglycan agrin (Nicaise et al., 2009a; Garbuzova-Davis et al., 2012; Meister et al., 2015; Ouali Alami et al., 2020; Tang et al., 2021).

Structural changes were associated with progressive functional impairment of the BSCB in ALS animals. IgG, an immunoglobulin typically unable to cross the BSCB, had higher deposition in the extra- and intra-neuronal spaces within the spinal cord of pre-symptomatic and symptomatic SOD1 rodents (Zhong et al., 2008; Nicaise et al., 2009a; Garbuzova-Davis et al., 2012; Boswell et al., 2013; Rabinovich-Nikitin et al., 2016). The mechanism of IgG deposition remains unknown as it is unclear whether immunoglobulins have extravasated from the bloodstream through a disrupted BSCB, by transcytosis across endothelial cells (Poduslo et al., 1994), or have originated from a non-vascular mechanism of retrograde axonal transport by the motor neurons (Fabian and Petroff, 1987; Fratantoni et al., 1996). BSCB permeability increased with disease progression in SOD1 rodents, as suggested by the extravasation of intravenous Evans Blue dye into the spinal cord of symptomatic animals, in contrast to the lack of Evans Blue leakage in presymptomatic animals (Garbuzova-Davis et al., 2007a; Zhong et al., 2008; Nicaise et al., 2009b; Miyazaki et al., 2011).

Compared to the number of animal studies focusing on the spinal cord, fewer reports have demonstrated neurovascular unit changes in the brain and brainstem. BBB disruption in the midbrain of symptomatic SOD1 ALS rats was detected by the extravasation of gadolinium from the intravascular compartment into the midbrain on MRI. Gadolinium enhancement was correlated with activation of microglia and immune cell infiltration as detected by immunohistochemistry (Andjus et al., 2009; Bataveljic et al., 2011). Downregulation of potassium channel Kir4.1 and upregulation of water channel aquaporin-4 (AQP4) in astrocytic end-feet were associated with BBB dysfunction within the motor cortex and brainstem of mutant SOD1 rats (Bataveljic et al., 2012). This astrocyte-induced impairment of water and potassium homeostasis was indicated as the driving factor for increased BBB permeability and leakage of intravascular toxic mediators into the motor neuron microenvironment. Additional toxic effects occur with the accumulation of extracellular ions, such as potassium, which overcomes physiological buffer mechanisms in the ALS disease state (Bataveljic et al., 2012, 2014).

AQP4 upregulation has also been demonstrated in mouse models of ALS carrying the SOD1^G93A, SOD1^G85R, and LoxSOD1^G37R mutations (Watanabe-Matsumoto et al., 2018). In these studies, AQP4 mislocalization along astrocyte end-feet was unique to ALS pathology and was not seen in other nerve damage models, such as sciatic nerve axotomy. While these studies noted upregulation of AQP4, other studies have demonstrated downregulation (Ouali Alami et al., 2020). This discrepancy may be due to variability in models or timepoints, as the response of regulation of the protein may go through different phases of dysregulation after its equilibrium has shifted, downregulating and then upregulating to compensate or vice versa. Future research may focus on this phenomenon. Furthermore, SOD1 ALS mice with knocked out AQP4 gene did not develop BBB dysfunction to the same extent as their AQP4 expressing counterparts. Despite improved BBB integrity, SOD1 ALS mice lacking AQP4 demonstrated accelerated disease progression and shortened survival (Watanabe-Matsumoto et al., 2018). This seemingly opposite effect whereby a lack of AQP4 improved BBB integrity but worsened disease phenotype may be due to independent roles of AQP4 in BBB-mediated CNS homeostasis and ALS pathogenesis.

The BCNSB findings observed in animal studies have also been corroborated in postmortem tissue from ALS patients (Garbuzova-Davis and Sanberg, 2014). In comparison to non-ALS human controls, there was disorganization of the microvascular architecture in the anterior horn, along with reduced capillary diameter (Sasaki, 2015) and density (Yamadera et al., 2015). Pericytes and endothelial cells within the spinal cord and brainstem exhibited degeneration, along with dissociation of astrocyte end-feet (Garbuzova-Davis et al., 2012; Winkler et al., 2013; Sasaki, 2015) and downregulation of tight junction proteins (Henkel et al., 2009). Disruptions in the endothelial lining indicating BCNSB breakage were also identified, with deposition of IgG, fibrin, thrombin, hemoglobin, and erythrocytes in the anterior horn tissue samples of patients with ALS (Miyazaki et al., 2011; Winkler et al., 2013; Sasaki, 2015). In the interstitial component of the BSCB, there was evidence for reduced and fragmented collagen bundles in the anterior horn of patients with ALS (Ono et al., 1998; Garbuzova-Davis and Sanberg, 2014). There was significant accumulation of perivascular collagen IV in the spinal cord and medulla (Garbuzova-Davis et al., 2012), a factor that can limit effective drug influx to the CNS (Garbuzova-Davis et al., 2016). The morphological and functional changes in the BCNSB in the late stages of ALS are schematized in Figures 2, 3.

TDP-43 deposition in endothelial cells has been observed in both ALS and frontotemporal lobar dementia with TDP-43 pathology (FTLD-TDP) in postmortem tissue (Ferrer et al., 2021). This has led to speculation over whether the pathological protein is being taken up by endothelial cells from neurons, or whether the observed deposits are from endogenous TDP-43 in the endothelial cells themselves. Both cases indicate that a vasculopathy is present in TDP-43 proteinopathies.

Whether the post-mortem BCNSB changes described in the vicinity of lower motor neurons (LMNs) in the spinal cord and brainstem also apply to the BBB surrounding the upper motor neurons (UMNs) in primary motor cortex is undetermined. Historically, studies in ALS have primarily investigated the LMNs and their connectivity with muscles at the neuromuscular junction, with less focus on the motor cortex. Given the robust data supporting a key pathological role of UMN dysfunction in ALS (Eisen, 2021), future preclinical, clinical and post-mortem studies on the interplay of astrocytes, UMNs, and motor cortex capillaries are warranted.

This scoping review did not identify validated imaging, serum, or CSF biomarkers to measure BCNSB integrity and function directly and reliably in patients living with ALS. Therefore, our understanding of BCNSB pathology throughout early to late stages of ALS in humans is limited. However, a small number of studies attempted to evaluate the BCNSB function in patients with ALS using pharmacokinetic assumptions and indirect markers.

Among these indirect measures, increased cerebrospinal fluid (CSF) total protein or albumin levels and CSF/serum concentration ratio of any given therapeutic may correlate with BCNSB permeability. Elevated CSF protein has been reported in some case series of ALS (Chelstowska and Kuzma-Kozakiewicz, 2014). However, protein may also leak into the CSF from the degenerating motor roots, limiting the utility of this measure as a BCNSB functional biomarker. For instance, breakdown of the blood-peripheral nerve barrier has been demonstrated by gadolinium enhancement on MRI of lumbosacral roots and leptomeninges of patients with ALS (Luigetti et al., 2010; Young et al., 2010). Circulating endothelial cells in peripheral blood was not a reliable indicator of BCNSB endothelial damage in patients with ALS (Garbuzova-Davis et al., 2010).

Measuring BCNSB permeability using CSF/serum therapeutic concentrations has limitations. It relies on the assumption of therapeutic extravasation from the vascular compartment to the CNS parenchyma and then to the subarachnoid space without direct CSF excretion at the choroid plexus (i.e., via the BCSFB) or complete therapeutic metabolism within neural tissue. Pharmacokinetic studies investigating CSF/serum concentration ratios of multiple ALS therapeutics have typically demonstrated low CSF/serum ratios suggesting limited therapeutic access across the BCNSB (Sussmuth et al., 2010; Wu et al., 2020; Prell et al., 2021; Waters et al., 2021).

Current structural and functional neuroimaging techniques lack the spatial resolution and accuracy required to directly assess the structure and function of BCNSB in patients with ALS. Radiologically, the BCNSB appears intact in ALS patients on standard imaging techniques, in contrast to multiple sclerosis and other CNS inflammatory conditions whereby BCNSB disruption is typically demonstrated by gadolinium extravasation on MRI. While early clinical ALS studies using novel 7T MRI scanners have largely explored the CNS anatomy and connectivity (Verstraete et al., 2010b; Cosottini et al., 2016; Barry et al., 2021), BCNSB leakage has not been extensively investigated with this technique. Of note, indirect measures of BBB integrity such as microbleeds in ALS were not increased as measured by 7T MRI (Verstraete et al., 2010a).

Despite increased permeability in preclinical and postmortem models, therapeutics with high molecular weight still have limited access to the CNS (Garbuzova-Davis et al., 2016). While lipid soluble and low molecular weight molecules can diffuse across the endothelial cell membranes, most of these compounds do not effectively enter the CNS and are transported back to the bloodstream by BCNSB-driven efflux or pharmacoresistance systems.

These pharmacoresistance proteins include specialized ATP binding cassette (ABC) efflux transporters that actively pump a number of endogenous and exogenous substrates out of the BCNSB (Mohamed et al., 2017). Relevant to ALS, the upregulation of key pharmacoresistance efflux proteins, P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP), has been shown in multiple animal models (Milane et al., 2010; Jablonski et al., 2012; Chan et al., 2017), a finding also supported by postmortem human tissue studies (Jablonski et al., 2012; Qosa et al., 2016; Van Vliet et al., 2020).

Increased expression and activity of P-gp began after symptom onset in the BSCB and BBB of mutated SOD1 models (Boston-Howes et al., 2008; Milane et al., 2010; Chan et al., 2017). TDP-43^A315T mutant mice also demonstrated P-gp and BCRP overexpression (Jablonski et al., 2012). In vitro, the P-gp upregulation in the endothelial cells was mediated by nuclear factor κB (NF-κB) activation and was induced when these cells were co-cultured with ALS-derived astrocytes (Qosa et al., 2016; Mohamed et al., 2019). In postmortem tissue, P-gp and BCRP upregulation was observed in the spinal cord and motor cortex of both sporadic and familial ALS cases (Van Vliet et al., 2020). It has been suggested that P-gp evolved for the purpose of responding to harmful substances (Broeks et al., 1995). Therefore, in ALS it may be responding to the toxic sequelae of aberrant protein in the CNS or other circulating toxins. Upregulation may be the result of a physiological attempt to dispense of harmful substrates. A consequence, however, may be decreased drug efficacy.

Of clinical relevance, riluzole, the first approved ALS drug, is a substrate of these efflux proteins (Milane et al., 2009b, 2010). The P-gp overexpression over the course of the disease may explain the poor bioavailability of riluzole within CNS parenchyma thereby limiting its therapeutic efficacy (Milane et al., 2010; Jablonski et al., 2014). In contrast, edavarone, a free radical scavenger approved for ALS treatment, has negligible P-gp binding and is potentially less influenced by BCNSB efflux activity (Dash et al., 2018; Hyung et al., 2018).

Reversing BCNSB pharmacoresistance in ALS is a long-sought approach to enhance therapeutic effect via inhibition of efflux transporters (Milane et al., 2009a; Jablonski et al., 2014). There was a significant increase in the parenchymal concentration of riluzole in rodents co-treated with efflux transporter inhibitors, such as minocycline (Milane et al., 2009a), verapamil liposomes (Yang et al., 2018) and elacridar (Jablonski et al., 2014). For instance, riluzole plus elacridar, a third-generation inhibitor BCRP and P-gp, improved muscle function, disease progression and survival in ALS mice (Jablonski et al., 2014).

While inhibition of efflux transporters is a promising strategy for clinical trials there are some challenges with current P-gp inhibitors. This includes the high dose needed to block P-gp which may lead to adverse reactions and systemic toxicity (Milane et al., 2007; Kalvass et al., 2013), as well as the need for selectivity and specificity of P-gp only at the BBB to prevent adverse effects on other organs (Amin, 2013).

Repair of the BCNSB pathology in ALS via stem cell transplantation has provided some evidence of symptomatic improvement in SOD1^G93A mice. The intravenous transplantation of human bone marrow-derived CD34+ cells (hBM34+) and endothelial progenitor cells (hBM-EPCs) enhanced the replacement of damaged endothelial cells in the CNS capillaries (Garbuzova-Davis et al., 2017, 2018, 2019, 2021; Eve et al., 2018). One mechanism of hBM-EPCs on the endothelium includes excretion of extracellular vesicles that transfer biomolecules to facilitate the repair of damaged microvascular endothelium in ALS (Garbuzova-Davis et al., 2020). Additionally, mice receiving hBM-EPCs have increased tight junction protein levels, capillary pericyte coverage, and basement membrane laminin expression, all of which maintain capillary endothelium integrity (Garbuzova-Davis et al., 2021). Ultimately, these mechanisms may prevent the entry of immune or inflammatory mediators which can contribute to motor neuron dysfunction (Garbuzova-Davis et al., 2019). At the symptomatic disease stage, hBM-EPC-treated mice have shown an improvement of behavioral outcomes and motor neuron survival, making it a promising therapeutic strategy for future translational clinical trials (Garbuzova-Davis et al., 2019).

Other therapeutics, such as an antagonist of the chemokine receptor CXCR4 (Rabinovich-Nikitin et al., 2016) and the myelosuppressive dealkylating agent busulfan (Peake et al., 2017), demonstrated improved BCNSB function in ALS models. Edaravone has been previously shown to mediate BBB repair in models of ischemic stroke, in addition to its primary mechanism of action as a free radical scavenger (Miyamoto et al., 2014; Tóth et al., 2014; Watanabe et al., 2015). Yet, BCNSB repair largely remains only a theoretical consideration in edavarone's mechanism of action in ALS.