Immunopathology and parasite sequestration cause severe cerebral trypanosomiasis in animals

Sara Silva Pereira Instituto de Medicina Molecular Joao Lobo Antunes https://orcid.org/0000-0002-6590-6626 Mariana De Niz Instituto de Medicina Molecular Joao Lobo Antunes https://orcid.org/0000-0001-6987-6789 Karine Serre Instituto de Medicina Molecular Joao Lobo Antunes https://orcid.org/0000-0001-9152-4739 Marie Ouarné Instituto de Medicina Molecular Joao Lobo Antunes https://orcid.org/0000-0003-4724-4363 Claudio A Franco Instituto de Medicina Molecular Joao Lobo Antunes https://orcid.org/0000-0002-2861-3883 Luisa M Figueiredo (  gueiredolm@gmail.com ) Instituto de Medicina Molecular Joao Lobo Antunes https://orcid.org/0000-0002-5752-6586


Introduction
Endothelial sequestration enables pathogen cytoadhesion to the blood vasculature and is key for survival of certain parasites and bacteria in their mammalian hosts. Trypanosoma congolense, the most prevalent and pathogenic African trypanosome species in African livestock (van den Bossche et al., 2011), is a fully intravascular parasite that employs sequestration by cytoadhering to the endothelial cells of several mammals (Banks, 1978;Hemphill et al., 1994).
Understanding the molecular basis of animal African trypanosomiasis (AAT) disease severity is of foremost importance because natural infections display wide variability in pathogenesis and clinical outcome. Disease can be acute with ≤70% mortality per herd, or chronic; and signalment can range from mild fever, anemia, and weight loss, to cachexia, acute in ammatory syndrome, disseminated intravascular coagulation syndrome, and neurological impairment, that ultimately culminate in generalized organ failure and death. Animals can experience necrotic and hemorrhagic lesions in major organs, particularly the brain, liver, and spleen (Silva Pereira et al., 2019), but can also remain chronically infected with less severe signalment, or be asymptomatic (Berthier et al., 2015;Mamoudou et al., 2016).
The reasons behind such large phenotypic variability are unknown, but we hypothesize that differential parasite sequestration plays an important role. This has been observed in other parasitic infections, such as Plasmodium falciparum, the causative agent of malaria, where sequestration is directly linked to disease severity, by causing cerebral malaria when happening in the brain (Ghazanfari et al., 2018), malaria-associated acute respiratory distress syndrome when in the lungs (Van den Steen et al., 2013), or placental malaria when in the placenta (Rogerson et al., 2007).
Parasite sequestration usually results in an in ammatory response (Storm and Craig, 2014). We know that T. congolense cytoadhesion to host cell membranes triggers antibody-complement cascades and increases vascular permeability, suggestive of endothelium damage (Banks, 1980). The parasite itself has also been reported to release soluble molecules, like trans-sialidades, that activate the endothelium, enhancing in ammation (Ammar et al., 2013). It is therefore plausible that the physical damage caused by parasite cytoadhesion and the resulting host's immune response affect disease progression.
Here, we report the rst mouse model of cerebral trypanosomiasis in animals and investigate its disease mechanism in living animals. We characterized parasite distribution in the mouse vasculature, vessel diameter preference, and duration of parasite-endothelial cell interaction. Furthermore, we assessed the impact of infection on the brain endothelium and on the onset of cerebral trypanosomiasis. Our data showed that cerebral trypanosomiasis is caused by a combination of increased parasite sequestration in the brain and T cell activation, via upregulation of intercellular adhesion molecule 1 (ICAM1) expression in both endothelial and circulating myeloid cells. These ndings highlight the importance of parasite sequestration and provide a cellular mechanism for the development of cerebral trypanosomiasis in animals.

A virulent strain of T. congolense causes cerebral trypanosomiasis
To investigate the basis of disease severity caused by T. congolense, we used two parasite strains of different virulence. Infection of C57BL/6J mice with strain 1/148 resulted in acute disease with mean mouse survival of 9.0±0.4 days (N = 4) ( Figure 1A). The majority of mice did not survive beyond the rst peak of parasitaemia ( Figure 1B). Within one to three days to the time of death, mice showed growing signs of neurological impairment, including loss of proprioception, hemiparesis (i.e. weakness in one side of the body), and strength and grip loss in the limbs. In contrast, infections with strain IL3000 resulted in 3-5 de ned peaks of parasitaemia ( Figure 1B). These mice died within 77.5±4.0 days ( Figure 1A) with multi-systemic pathology.
We compared organ pathology at the rst peak of parasitaemia and observed that infection with strain 1/148 resulted in marked damage to the brain, kidney, and thymus; moderate damage in the liver; and mild damage to the heart, lungs, and spleen ( Figure 1C). Infection with strain IL3000 resulted also in moderate damage to the liver, but only mild damage to the spleen, thymus and testes, and minimal to the heart ( Figure 1C). Importantly, whilst there were no pathological changes in the brain of mice infected with IL3000, 1/148 infections resulted in large brain lesions characterized by multifocal grey matter vacuolation, ischemia, neuronal loss and hemorrhages ( Figure 1D), which were the most likely cause of death. Parasite labelling by immunohistochemistry showed large parasite accumulation in the brain vasculature ( Figure 1E), to the point of vascular occlusion in smaller capillaries. Parasite accumulation in the brain vasculature during natural infections has been described in the literature, including in cattle, dogs, horses, and wild animals, some of which have been reported to develop neurological impairment (Losos and Gwamaka, 1973;Losos and Ikede, 1972;Losos et al., 1971;Savage et al., 2021).
The proportion of Trypanosoma congolense sequestration varies with strain and tissue, independently of parasite load Given the severity of lesions in the host and the large accumulation of parasites in the brains of animals infected with the T. congolense strain 1/148, we hypothesized that early death was caused by parasite sequestration in the brain vasculature. Cytoadhesion and sequestration are often used interchangeably.
Here, we refer to cytoadhesion as the physical act of parasite binding to the endothelium, and to sequestration as the mechanism. Therefore, we started by con rming that T. congolense parasites cytoadhered to the endothelium by intravital microscopy. Since we did not have access to uorescent 1/148 parasites, we used FITC-Dextran and Hoechst dye to visualize the intravascular environment and the nuclei of circulating cells, respectively (Figure 2A), as well as α-A637-CD31 to mark the vascular endothelium ( Figure 2B). This combination of markers allowed us to quantify parasites in vessels. Then, based on parasite mobility during live imaging, we compared total parasite load and the percentage of cytoadhered parasites in major organs in the rst 6 days of 1/148 infection and at three points in IL3000 infection ( rst peak, second peak, and postsecond peak). For representative image acquisition, we perfused mice to remove free-owing parasites (i.e. not cytoadhered) ( Figure 2B). In 1/148 infections, total parasite load increased linearly with time in the brain (R 2 = 0.83, Pearson's correlation) and adipose tissue (R 2 = 0.68, Pearson's correlation) ( Figure  2C, left). In remaining organs, parasite load oscillated during infection. For instance, in the liver, parasite load was maximal already on day 2; in the spleen, maximum parasite load was reached on day 5 p.i; and in the lungs on day 4 p.i.. The brain was the organ with the highest parasite load at day 6 p.i., and variability across replicates was low ( Figure 2C and 2D, left, Video S1). This was also the maximum parasite load registered throughout the course of infection amongst all organs. In IL3000 infections, parasite load was highest in the lungs, at both peaks of parasitaemia (1.25 to 8.64 x 10 5 parasites / cm 2 vessel), followed by heart (1.14 to 2.15 x 10 5 parasites / cm 2 vessel) and brain (0.61 to 2.16 x 10 5 parasites / cm 2 vessel). Parasite distribution from the rst to second peaks of parasitaemia did not change drastically, but there was a 2-fold decrease in the spleen ( Figure 2C, right). After the second peak of parasitaemia, in a window of undetectable peripheral parasitaemia by hemocytometry, there was a small number of parasites detectable in the brain, heart, lungs, pancreas, and spleen. An example of a parasite cytoadhered to the endothelium of a pancreatic arteriole can be seen in Video S2.
Unexpectedly, parasite sequestration did not increase linearly with parasite load in neither 1/148 nor IL3000 infections ( Figure 2E). For all organs, parasites began sequestering at day 1 p.i., but throughout infection the percentage of cytoadhered parasites oscillated in an organ-dependent manner. In 1/148 infections, in the brain and pancreas, parasite sequestration was maximal at day 6 p.i. and the proportion of sequestered parasites (over the total number of parasites recorded) was higher than 80%. In remaining organs, the percentage of parasite sequestration was highest earlier in infection ( Figure 2E and 2F, left). In IL3000 infections, parasite sequestration at the rst peak of parasitaemia was highest in the kidney and spleen (all detected parasites were cytoadhered), followed by the heart (92% ± 3, respectively) ( Figure 2E and 2F, right). During the second peak of parasitaemia, sequestration was lowest in the lungs (11% ± 3, compared to 75% ± 3 in the rst peak), and highest in the kidney (85% ± 4). After the second peak, and despite the above-mentioned low parasite load, all parasites in the spleen and lungs and 92% ± 6 in the heart were cytoadhered, suggesting that these sequestered parasites escaped immune or splenic clearance. Representative images of T. congolense sequestration in the different organs at the rst peak of parasitaemia (day 6 p.i. for 1/148 and days 7-10 p.i. for IL3000 infections) are shown in Figure S1.
Together, these data showed that parasite sequestration is tissue-, strain-and time-dependent, but not a consequence of high parasitaemia. Importantly, given that the brain was the organ with the highest parasite load and the highest sequestration at day 6 after 1/148 infection, it supported the hypothesis that parasite sequestration contributed to cerebral trypanosomiasis and early death.
Individual parasites sequester in the brain vasculature for up to 8 hours We sought to investigate how tropism is established by parasites of strain 1/148 preferentially sequestering in the brain. Therefore, we used ex-vivo microscopy to understand whether they accumulated in particular brain regions or brain vessels. Mice were infected and, at day 6 post-infection, brains were surgically removed, dissected into ten anatomical regions, and imaged immediately. We found that, in both 1/148 and IL3000 infections, parasites accumulated in the posterior parts of the brain (i.e. cerebellum, midbrain, pons, medulla) (Figures 3A and 3B,left). However, while IL3000 sequestration reproduced this pattern, 1/148 sequestration was evenly distributed across vessels of all brain sections (Figures 3A and 3B,right). Next, we questioned whether vessel caliber was a determinant of parasite sequestration, so we compared the number of parasites in capillaries (φ < 10 µm), and arterioles/venules of different diameters (10 ≤ φ > 20 µm, 20 ≤ φ > 40 µm, φ ≥ 40 µm). We found that T. congolense parasites distributed evenly ( Figure 3C, left) and sequestered in similar proportions throughout vessels of all diameters ( Figure 3C, right).
We observed that T. congolense parasites can adhere to the brain endothelium with every part of their cell ( Figure 3D), including the cell body (Video S3), the posterior end (Video S4), the distal agellum (Video S5), or a mixture (i.e. when during the duration of the video parasites attached with more than one part of their cell). IL3000 parasites adhered less with the posterior end and more with the distal agellum than 1/148 parasites (10% ± 6 vs. 24% ± 5 and 45% ± 5 vs. 32% ± 7, respectively; p < 0.001, unpaired t-test) ( Figure 3E). We then followed 100 parasites for up to 10 hours and discovered that 1/148 parasites remained cytoadhered to the same endothelial cell for longer than IL3000 and to a maximum of 8 hours (median of 2 hours ± 7 minutes and 5 hours ± 11 minutes for IL3000 and 1/148, respectively) ( Figure 3F).
In summary, we observed that parasites of both strains preferentially colonized the vasculature of the posterior parts of the brain, but that 1/148 cytoadhered equally well across brain regions, corroborating our earlier observation that sequestration was independent of parasite load. These experiments also showed very little physical constraints for parasite distribution and sequestration, as parasites distributed and cytoadhered to vessels of all calibers and interacted with the endothelium using any part of their bodies. Importantly, we showed that 1/148 sequestration in the brain is a longer-lived interaction, thus corroborating the tropism of this strain to the brain vasculature.
These observations, together with the fact that T. congolense 1/148 infections result in shorter mouse survival, signi cant neuropathology, and increased parasite sequestration in the brain, indicate that T. congolense 1/148 infection in C57BL6/J mice are a good model of cerebral trypanosomiasis. For the remaining of this study, we investigated the mechanism behind cerebral trypanosomiasis in mice.

T. congolense 1/148 infection induces a pro-in ammatory pro le in brain endothelial cells
We started by characterizing how the brain endothelium responded to 1/148 infection. To achieve this, we infected RiboTag.PDGFb-iCRE mice, which, upon induction of Cre recombinase activity, express a hemagglutinin tag at the ribosomes of the endothelial cells. We harvested the brain, captured the polysomes of the endothelial cells by immunoprecipitation, and sequenced the mRNA that was in translation (Sanz et al., 2009) (Figure 4A). We compared the expression pro les of brain endothelial cells of mice non-infected and infected with T. congolense 1/148 and found that infection resulted in downregulation of 588 genes and upregulation of 612 genes ( Figure 4B and Table S1).
To study the predicted biological function of upregulated genes in a more systematic and unbiased way, we performed gene set enrichment analysis of the 612 genes (GSEA) ( Table S2). Results were consistent with the signature we observed from the most upregulated genes, as 63% of the enriched gene sets related to in ammatory responses (e.g. response to interferon gamma, positive regulation of defense response, response to tumor necrosis factor, cytokine-mediated signaling pathway, regulation of innate immune response, response to virus, response to interleukin-1, antigen processing and presentation, regulation of in ammatory response, leukocyte migration, etc.), and 7% to angiogenesis (e.g. angiogenesis, regeneration) ( Figure 4D).
Given that the transcriptomic analysis revealed signs of a strong pro-in ammatory response, we asked if the serum cytokine levels were consistent with such response. For that, at days 3 and 6 p.i., we assessed the serum levels of six cytokines present either in the 20 most upregulated genes or the enriched gene sets (i.e. IL-1α, IL-1β, IFNγ, CXCL10, CXCL9, TNFα). We con rmed a signi cant increase in the levels of IFNγ at day 3 p.i. (log 2 FC = 2.82) (2-way ANOVA, p-value = 0.01), an increase in CXCL10 at day 6 p.i. (log 2 FC = 2.53) (2-way ANOVA, p-value = 0.01), and an increase in CXCL9 at both days 3 and 6 p.i. compared to the non-infected control (log 2 FC = 3.45 and 4.30, respectively) (2-way ANOVA, p-value = 0.03 and 0.002, respectively) ( Figure 4E). These results support the presence of a systemic pro-in ammatory response in the rst week of infection with 1/148 strain.
Given that an activated endothelium upregulates the expression of integrins which control the immune response, we also assessed the expression changes of intercellular adhesion molecule (ICAM) 1, ICAM2, and vascular cell adhesion molecule (VCAM) 1, three integrins involved in leukocyte transendothelial migration. We compared expression of ICAM1, ICAM2, and VCAM1 proteins in the vasculature of the brain at day 6 p.i., by ex-vivo microscopy ( Figure 4F). Consistent with the endothelium ribosome pro ling data (which revealed ICAM1 to be upregulated upon infection (FC = 1.62), but not of ICAM2 or VCAM1), we observed increased expression of ICAM1 (unpaired t-test with Welch's correction, p-value < 0.01), but not of ICAM2 (unpaired t-test with Welch's correction, p-value < 0.0001), nor VCAM1 (unpaired t-test with Welch's correction, p-value < 0.01) ( Figure 4G). To determine whether the increase in ICAM1 was brainspeci c, we checked ICAM1, ICAM2, and VCAM1 expression in remaining organs, and observed it also increased in the lungs, heart, and liver (unpaired t-test with Welch's correction, p-value < 0.01), whereas ICAM2 expression increased in the heart and kidneys (unpaired t-test with Welch's correction, p-value < 0.0001), and VCAM1 in the liver, spleen, and kidneys (unpaired t-test with Welch's correction, p-value < 0.01) ( Figure S2A). Similar to what we did to assess the distribution of parasite load and sequestration, we asked whether the increase in ICAM1 expression was localized to a particular region in the brain. We observed that expression of ICAM1 increased in all anatomic regions of the brain, with the exception of the hypothalamus ( Figure S2B and S2C).
In summary, analysis of the brain vasculature revealed that endothelial cells respond to T. congolense infection by displaying a pro-in ammatory response. IFNγ and/or associated cytokines (i.e. CXCL9, CXCL10), as well as genes involved in leukocyte transendothelial migration (i.e. ICAM1) may play a role in the development of T. congolense 1/148-associated neuropathology.
Blocking of ICAM1 reduces disease severity and parasite sequestration in the brain The increase in ICAM1 expression in the brain endothelium, but not of ICAM2 or VCAM1 suggests that ICAM1 may play a role in disease severity, perhaps by favoring parasite sequestration and/or in ammation. We found previously that T. brucei interacts with several endothelial receptors (De Niz et al., 2021). To test if T. congolense sequestration depends on similar host molecules, we blocked ICAM1 in vivo, by administering α-ICAM1 antibody 24 hours before infection, repeating daily during the course of infection. We noticed that α-ICAM1-treated mice showed fewer signs of cerebral disease, so we used SHIRPA (i.e. SmithKline Beecham, Harwell, Imperial College, Royal London Hospital, phenotype assessment), a mouse phenotypic assessment protocol (Rogers et al., 1997), to test whether α-ICAM1treated mice were neurologically tter than isotype-treated controls.
We selected 5 parameters of the semi-quantitative SHIRPA protocol: posture, velocity of escape upon touching, positional passivity, type of locomotion, and grip strength. Mice infected with the strain 1/148 showed defects in all these parameters on day 6 post-infection ( Figure 5A). These alterations are consistent with the changes described in dogs infected with T. congolense (Harrus et al., 1995), showing that SHIRPA protocol is a useful quantitative method to score neurological alterations associated to cerebral trypanosomiasis in this mouse model. The SHIRPA protocol revealed that α-ICAM1-treated mice performed better than the control group in all parameters relating to neurological impairment, and particularly those relating to movement and strength ( Figure 5A). Importantly, we also found that α-ICAM1-treated mice survived longer than isotype-treated controls, as 50% of the mice were still alive at day 20 p.i, when the experiment was terminated ( Figure 5B). Brain necropsy also showed reduced neuropathology ( Figure 5C).
To understand the cellular basis of this ICAM1-dependent phenotype, we quanti ed parasite load and sequestration percentage in the brain at day 6 p.i., by intravital microscopy ( Figure 5D). We found that total parasite load was unaffected by treatment ( Figure 5E, left), but the percentage of cytoadhered parasites was reduced by 44% (2-way ANOVA, p-value = 0.0025) ( Figure 5E, right). These results could suggest that ICAM1 directly promotes parasite sequestration. To test if ICAM1 is a potential receptor for parasite adhesion, we immobilized recombinant ICAM1, CD36 and BSA on a plastic surface, in static conditions, and measured 1/148 parasite adherence by microscopy after a 30-minute incubation ( Figure   5F). Parasites did not adhere more to recombinant ICAM1 protein than to either BSA or recombinant CD36 protein ( Figure 5G), suggesting that parasite sequestration is not mediated by ICAM1 binding.
These data show that blocking ICAM1 reduces parasite sequestration in the brain, improves disease severity and promotes survival, but this effect appears to be independent of direct parasite binding to the ICAM1 receptor.

T. congolense 1/148 infection results in an increase in the number of circulating ICAM1 + monocytes
Given that ICAM1 is important in leukocyte recruitment and in ltration (Lawson and Wolf, 2009), next we investigated whether the phenotypic effect of ICAM1 blocking could derive from an impaired immunological response. First, we measured the impact of ICAM1 blocking in the number of circulating nucleated cells speci cally in the brain vasculature. By intravital imaging, we detected a 67%±5% reduction in the number of circulating nucleated cells (assumed to be leukocytes due to their size and complexity) (one-way ANOVA, p-value = 0.0270), which was not observed when blocking either ICAM2 or VCAM1 ( Figure 6A and 6B). These results indicate that the ICAM1 blocking may improve disease outcome by reducing the magnitude of the immune response.
Next, we characterized the immune response in the systemic circulation at day 6 p.i. using ow cytometry. We observed no signi cant changes in the number of myeloid cells (i.e. dendritic cells, monocytes and macrophages) at day 6 post-infection ( Figure 6C), but the number of B cells was 1.63-fold reduced (2-way ANOVA, p-value = 0.0277) ( Figure 6D). Interestingly, ICAM1 + monocytes in the blood increased 13-fold upon infection (2-way ANOVA, p-value < 0.0001) ( Figure 6E).
Then, we characterized the leukocyte populations within the brain and within the brain vasculature. Thus, to distinguish and quantify the number of intra-and extravascular leukocytes, we administered α-CD45-APC antibody intravenously, 3 minutes prior to mouse euthanasia. In such a short incubation time, the antibody does not traverse the vasculature and thus only stains intravascular leukocytes (Anderson et al., 2014;Morawski et al., 2017). Figure S3A con rms that the α-CD45-APC antibody was e cient at staining intravascular leukocytes. As expected, we found an increase in the number of both extravascular (unpaired t-test, p-value = 0.0482, t = 2.813, df = 4) and intravascular (unpaired t-test, p-value = 0.0160, t = 4.0007, df = 4) CD45 + cells upon infection ( Figure S3B). Additionally, the contribution of intravascular leukocytes to the total immune cells in the brain is higher in infected animals (14% ± 4 vs. 25% ± 1).
Leukocytes are a heterogenous population of cells. To characterize the changes within the myeloid (e.g. monocytes, macrophages and dendritic cells) and lymphoid (B and T cells) subpopulations, we stained brain leukocytes for various subset markers, and quanti ed them using ow cytometry ( Figure S4). We did not stain for neutrophils because preliminary experiments revealed that neutrophil numbers do not change signi cantly upon infection ( Figure S3C). As expected, the numbers of extravascular DCs and monocytes was low in both infected and non-infected conditions. However, the number of CD11b + F4/80 + macrophages increased 2.4-fold upon infection, suggesting in ltration of circulating in ammatory monocytes into the brain parenchyma (2-way ANOVA, p-value < 0.0001) ( Figure 6F). In contrast, the number of intravascular CD11b + CD11c + dendritic cells (DC) and CD11b + Ly6C + monocytes increased by 8-fold (2-way ANOVA, p-value = 0.0113, < 0.0001, respectively) ( Figure 6G). Importantly, 93 ± 16% of the intravascular monocytes expressed ICAM1 during infection ( Figure 6H) and at higher mean uorescent intensity (MFI) (2013 ± 206 in non-infected vs. 2962 ± 206 in infected mice, 2-way ANOVA, pvalue = 0.0006) ( Figure S3D), compared to only 32 ± 14% in the non-infected control, which is re ected in a signi cant increase (91-fold) in the number of ICAM1-expressing monocytes circulating in the brain vasculature (2-way ANOVA, p-value < 0.0001) ( Figure 6H).
The numbers of lymphoid cells present in the brain showed a tendency to increase, but this effect was re ected in both extravascular and intravascular populations ( Figure 6I and 6J). The numbers of extravascular B cells increased 5-fold, CD4+ T helper cells increased 9-fold, and CD8+ cytotoxic T cells increased 8-fold. Whilst the number of intravascular B cells remains similar upon infection, the number of CD4+ T helper cells and CD8+ cytotoxic T cells increased 4.5 and 4.7-fold, respectively. In terms of ICAM1 expression, barely any T cells were detected to express ICAM1 regardless of infection status, whereas the number of ICAM1 + extravascular B cells increased 13-fold (2-way ANOVA, p-value < 0.0353) ( Figure 6K).
These data show that, upon infection, the development of cerebral trypanosomiasis is associated with the recruitment of myeloid cells to the brain vasculature, most of which express ICAM1. Cerebral animal African trypanosomiasis is prevented in the absence of T cells To test if the increase in circulating monocytes cells is the cause of cerebral trypanosomiasis, we depleted circulating monocytes and in ammatory macrophages by intravenous administration of clodronate liposomes 24 hours prior and 3 days post-infection. We observed that mice treated with clodronate liposomes showed high peripheral parasitaemia earlier than mice receiving either PBS-lled liposomes or PBS only ( Figure S5A). Survival was also affected. Mice receiving clodronate liposomes, died between days 6 and 7, compared to days 7-8 when receiving control liposomes, and days 7-10 if only PBS was administered ( Figure S5B). Moreover, we performed mouse behavioral assessment on day 6 post-infection and observed that monocyte/macrophage depletion increased severity of neurological impairment ( Figure S5C). These results show that neuropathology cannot be speci cally attributed to monocytes/in ammatory macrophages, as these cells seem to have a principal protective effect against infection.
In other parasitic diseases, like cerebral malaria, brain immunopathology is caused by small numbers of in ltrating CD8 + T cells (Belnoue et al., 2002). Moreover, upregulation of ICAM1, IFNγ, and related cytokines is known to increase T cell adhesion and transendothelial cell migration (Ashok Sonar et al., 2017;May and Ager, 1992;Sancho et al., 1999). Thus, we hypothesized that T cells could cause neuropathology. We used mice of the same genetic background, but that do not produce mature T and B cells (RAG2 KO mice) to directly investigate the role of T cells in parasite sequestration and pathology in the brain. Although they also lack mature B cells, mice with cerebral trypanosomiasis do not survive past the rst peak of parasitemia, so the antibody-mediated response is unlikely to play an important role in ICAM1-mediated neuropathology. Besides, we did not observe a signi cant variation in B cell numbers in the brain upon infection.
We performed intravital microscopy in the brain of RAG2 KO mice infected with T. congolense 1/148 at day 6 p.i. ( Figure 7A) and quanti ed the number of circulating nucleated cells. We observed that it was reduced relative to WT mice and similar to the levels observed in α-ICAM1-treated mice ( Figure 7B and Figure 6B). When we followed disease progression, we observed that RAG2 KO mice performed better than WT in every parameter of the SHIRPA test, with only one mouse (out of ve) showing signs of neurological impairment ( Figure 7C). As expected due to the absence of B cells, mice did not clear parasitaemia, and remained heavily parasitized, with minor oscillations ( Figure 7D). Remarkably, unlike WT controls, mice survived up to day 40 p.i. ( Figure 7E). Brain necropsy corroborated these observations as the level of brain lesions was lower overall, despite one mouse still showing moderate neuropathology ( Figure 7F). These data suggest that neuropathology associated with cerebral trypanosomiasis is mainly due to T cells.
Given our previous results indicating a role of ICAM1 in cerebral trypanosomiasis, we investigated whether T cells could cause neuropathology via ICAM1 signaling. Therefore, we administered α-ICAM1 or its isotype control into RAG2 KO mice and infected them with T. congolense 1/148 parasites. At day 6 p.i., α-IgG2-treated RAG2 KO mice had similar parasite load in the vasculature of the brain to WT ( Figure 7G), but presented lower levels of parasite sequestration (2-way ANOVA, p-value = 0.0241) ( Figure 7H). This phenotype is similar to what we observed in α-ICAM1-treated WT mice ( Figure 5E). Furthermore, we observed that α-ICAM1-treatment caused no effect on RAG2 KO mice in terms of both parasite load and percentage of parasite sequestration ( Figure 7G and 7H), which shows that ICAM1 depletion plays no additional role when T cells are absent.

Discussion
Animal African trypanosomiasis comprises a spectrum of diseases. In this work, we established a mouse model of cerebral trypanosomiasis. We show that C57BL6/j mice infected with T. congolense 1/148 strain show signi cant and quanti able neurological clinical signs, which are associated to parasite prolonged sequestration in the brain vasculature and an overwhelming ICAM1-mediated proin ammatory immune response.

Immunopathology in cerebral animal African trypanosomiasis
In the acute disease caused by strain 1/148, the brain is the site of preferential parasite sequestration. The presence of T. congolense has previously been shown to result in endothelium activation via the release of soluble factors in vitro (Ammar et al., 2013). Given our observations of T. congolense 1/148 parasites accumulating in the brain vasculature and having a prolonged interaction with the endothelium, we propose that parasite sequestration further enhances endothelial activation in vivo. Ultimately, tropism to the brain vasculature is a virulence factor and is a common feature of intravascular parasites. Trypanosome sequestration on the brain endothelium seems to cause the vascular occlusion that results in ischemia and tissue hypoxia, accounting for some of the pathological lesions we observed in infected mice. In both cerebral malaria and acute babesiosis, attenuation of strain virulence is accompanied by loss of cerebral capillary sequestration (Dondorp et al., 2004;Medana and Turner, 2006;Sondgeroth et al., 2013), much like what we observed when comparing T. congolense strains 1/148 and IL3000.
Ribosome pro ling shows that brain endothelial cells respond to T. congolense infection strongly, by upregulating angiogenic and proin ammatory pathways. These responses include upregulation of genes involved in the increase of vascular permeability and vasodilation, as well as the recruitment and activation of innate and adaptive immune cells. Currently, we do not know whether the endothelium is activated by the direct effect of prolonged parasite sequestration, or due to the immune response.
T cell suppression by cyclosporine A administration has been shown not alter disease course nor associated in ammation in chronic models of trypanosomiasis (Noyes et al., 2009). In contrast, we would expect that the same treatment in a 1/148 infection would result in prolonged survival, mimicking the effect we observed in RAG2 KO mice. Moreover, although ICAM1 is upregulated in endothelial cells during IL3000 co-culture (Ammar et al., 2013), it does not result in life-threatening neuroin ammation ( Figure 1C). This suggests that, whilst the exacerbated immune response may be the cause of neuropathology, parasite tropism to the brain, and the consequent increase in parasite sequestration, may be essential for the development of cerebral trypanosomiasis. This mechanism may not be triggered in IL3000 infections because of the reduced parasite sequestration in the brain. We propose a model for the mechanism of disease of cerebral trypanosomiasis (Figure 8). Upon adhesion to endothelial cells, the parasite releases pro-in ammatory soluble molecules, such as transsialidades, and causes activation of the endothelium via the NF-κB pathway (Ammar et al., 2013). This would result in the observed increase in ICAM1 expression in the plasma membrane of endothelial cells.
Ultimately, parasite sequestration could trigger a pro-in ammatory response, possibly mediated by IFNγdependent and the NF-κB pathway. Pro-in ammatory cytokines lead to the increase in circulating myeloid cells in the brain, particularly of ICAM1 + monocytes. Some of the monocytes cross the BBB, accumulate in the brain and differentiate into macrophages. These cytokines, chemokines and myeloid cells also promote T cell recruitment to the brain vasculature (Clark et al., 2007). T cells interact with ICAM1 at the surface of endothelial cells, and possibly myeloid cells, which facilitates activated T cell adhesion and extravasation to the brain parenchyma (Dietrich, 2002). We propose this is the cause of the immunopathology that culminates in cerebral trypanosomiasis.
The similarities of the T. congolense-associated neuropathology with that of cerebral malaria are clearly remarkable, but perhaps not unexpected if we reconsider that the initial events that trigger the immune response, namely sequestration and endothelial damage, are alike. ICAM1 is one the endothelial cell adhesion receptors for P. falciparum (Berendt et al., 1989), but has a multifaceted role in cerebral malaria pathogenesis (Storm and Craig, 2014). Whilst we did not nd evidence for ICAM1 to be an endothelial cell receptor for T. congolense parasites, this hypothesis cannot be excluded until more sensitive and complex adhesion assays are performed. Both IFNγ and CXCL10 play important roles in the in ammation that drives cerebral malaria-associated neuropathology, namely in terms of T cell adhesion to endothelial cells (Sorensen et al., 2018), and recruitment to the brain parenchyma (Campanella et al., 2008). Monocytes promote brain in ammation, secretion of IFNγ, and activation of CD8 + T effector cells (Schumak et al., 2015).

Advantages of parasite sequestration
Our data shows that T. congolense parasites of both 1/148 and IL3000 strains sequester in the vasculature of major organs, albeit to different degrees. Previous studies of T. congolense in experimental and natural infections are consistent with our observations (Fiennes, 1952;Losos and Gwamaka, 1973;Losos et al., 1971;Maxie and Losos, 1977). It emerges that sequestration is a trait of T. congolense as a species and an essential mechanism of interaction with the mammalian host. However, sequestration is not a characteristic of all African trypanosomes. T. brucei parasites, for instance, do not sequester (Silva Pereira et al., 2019). Although they interact with the endothelium and may in fact adhere to it, T. brucei parasites do it as means to invade the extravascular spaces of tissues (De Niz et al., 2021). So, why is it more advantageous for T. congolense to sequester, but for T. brucei to cross the endothelium? Other parasites employing sequestration, like P. falciparum and B. bovis, do so to evade the immune response and prevent splenic clearance of infected red blood cells (Allred and Al-Khedery, 2004; Dinko and Pradel, 2016). However, being extracellular, trypanosome clearance is not exclusively spleen-dependent. Some IL3000 parasites remain cytoadhered after the antibody-mediated clearance that follows the peaks of parasitemia, suggesting sequestration may help to keep enough parasites to start a new wave of parasitaemia. Nonetheless, clearance of T. brucei from the blood is also not entirely e cient (De Niz et al., 2021), which indicates that sequestration is not a requirement for parasite persistence, even in the absence of tissue invasion.
Morphological and physiological characteristics of T. congolense parasites facilitate sequestration: they can cytoadhere with any part of the cell and have a rounder cell shape, stiffer cell body, and shorter agellum that results in slower motility (Bargul et al., 2016). Cytoadhesion requires considerable force to counteract the speed of the blood blow, so these features may help reduce energy expenditure.
Simultaneously, it is possible that sequestration enhances parasite-endothelium interactions by facilitating hijacking of cellular functions and/or host's nutrients. For instance, sequestration might facilitate iron uptake, which is both essential for trypanosome survival and a source of host pathology due to anemia (Stijlemans et al., 2015). It would be interesting to test whether cytoadhered T. congolense parasites are functionally different from circulating parasites and for example, express higher levels of transferrin or haptoglobin-hemoglobin receptors, necessary for iron and hem internalization, respectively. Our work lays the ground for more detailed examinations of sequestration, including the characterization of movement type, binding forces and receptor-ligand interactions.
In conclusion, we show that cerebral trypanosomiasis is caused by the combination of parasite sequestration in the brain and immune cell recruitment, and consequently can be prevented. As we reveal that T. congolense sequestration is a virulence factor, we bring a novel hypothesis forward: trypanosome sequestration patterns can determine strain virulence, and thus the risk of disease severity. The translation potential of our ndings is large as they expose possible drug treatment strategies targeting mediators of sequestration and/or immunomodulators, as well as to the design of targeted strategies for vector control and surveillance frequency, when coupled with epidemiological mapping.

Declaration of Interests
The authors declare no competing interests.

Animal Experiments
This study was conducted in accordance with EU regulations and ethical approval was obtained from the Animal Ethics Committee of Instituto de Medicina Molecular (AWB_2016_07_LF_Tropism). Infections were performed at the rodent facility of Instituto de Medicina Molecular, in 6-10 weeks old, wild-type, male C57BL/6J mice (Charles River, France), RiboTag.PDGFb-iCRE mice, or RAG2 KO mice of the same age, bred in-house. Mice were infected by intraperitoneal injection (i. p.) of 2 x 10 3 [T.

Histology and Immuno-histochemistry
Formalin-xed organs were embedded in para n and 3 µm sections were stained with hematoxylin and eosin (H&E). For immunohistochemistry, 3 µm sections were immunostained with a non-puri ed rabbit serum α-T. brucei histone 2A (H2A) (generated against a recombinant protein) (kind gift of Christian Janzen), diluted 1:5000. Antigen heat-retrieval was performed in a microwave oven (800 W) for 15 minutes with pH 9 Sodium Citrate buffer (Leica Biosystems, MO, USA). Incubation with ENVISION kit (Peroxidase/DAB detection system, Dako Corp, Santa Barbara, CA) was followed by Mayer's hemalumen counterstaining. Tissue sections were examined by a pathologist, blinded to experimental groups, in a Leica DM2500 microscope coupled to a Leica MC170 HD microscope camera.

Surgical Procedures and Intravital Microscopy
For intravital microscopy, surgeries were separately performed by groups of organs, as previously described for the brain (De Niz et al., 2019a), the lungs and heart, the liver, pancreas, spleen and kidneys (De Niz et al., 2020), and the adipose tissues (De Niz et al., 2019b). In summary, mice were anaesthetized prior to surgery with a mixture of ketamine (120 mg/kg) and xylazine (16 mg/kg), by intraperitoneal injection. Re ex responses were induced, and surgery was initiated once these responses were nonexistent. For all experiments used for quanti cation of cytoadhered and owing parasites, mice were injected with 3 markers: Hoechst 33342 to label nucleic acids, 70 kDa FITC-Dextran to label the intravascular space and provide contrast within the vessel walls, and a uorescently conjugated antibody against the pan-vascular marker CD31 (PECAM1). Prior to surgery, mice were intraocularly injected with Hoechst 33342 (stock diluted in dH2O at 100 mg/ml, injection of 40 µg/kg mouse), 70 kDa FITC-Dextran (stock diluted in 1x PBS at a concentration of 100 mg/ml, injection of 500 mg/kg mouse), and A637-CD31 (BioLegend, used at 20 µg/mouse). For imaging, a temporary glass window (Merk rectangular cover glass, 100 mm x 60 mm) or a circular cover glass (12 mm) was implanted in each organ. The windows were secured surgically using stitches, or immobilized using surgical glue. For imaging the heart and lungs, the windows were immobilized using a vacuum to prevent collapse of the thoracic cavity.
All intravital microscopy performed with the aim of identifying owing and circulating parasites was performed in spinning disc microscopes. These included a Zeiss Cell Observer SD (Carl Zeiss Microimaging, equipped with a Yokogawa CSU-X1 confocal scanner, and an Evolve 512 EMCCD camera and a Hamamatsu ORCA-Flash 4.0 VS camera) or a 3i Marianas SDC (spinning disc confocal) microscope (Intelligent Imaging Innovations, equipped with a Yokogawa CSU-X1 confocal scanner and a Photometrics Evolve 512 EMCCD camera). Laser units 405, 488, and 647 were used to image Hoechst, FITC-Dextran and AF67-CD31 respectively. Visualization was done using either an oil-immersion plan apochromat 63x objective with 1.4 NA and 0.17 WD, or a 40x LD C-Apochromat corrected, water immersion objective with 1.1 NA and 0.62 WD. Images were obtained for a total of 20 seconds, at an acquisition rate of 20 frames per second. For all acquisitions, the software used was either ZEN blue edition v.2.6., or 3i Slidebook reader v.6.0.22.
For the quanti cation of the duration of sequestration in the brain, longer imaging sessions were required. Therefore, T. congolense-infected mice were imaged at day 6 post-infection for a total of 12 hours, with images acquired every 30 minutes. In this case, a non-invasive cranial window was implanted on the skull, and anesthesia was injected every hour intraperitoneally, in addition to inhalable anesthesia (iso urane) being supplied as required to ensure analgesic and anesthetic effect for the full imaging time.
Cytoadhered parasites were de ned as those that did not change positions relative to the immediately preceding time, in half-hourly periods.

Ex-vivo Microscopy
While intravital imaging in living mice mostly allows visualization of parasite-host interactions in the cerebral cortex, the meninges, and the olfactory bulb, we were interested in visualizing parasite dynamics throughout the entire brain. For this purpose, we followed the same procedures as previously described for vascular and nucleic acid labeling, and we then surgically exposed ten different regions of the mouse brain, namely the olfactory bulb, the cerebral cortex, the septum, the hypothalamus, the thalamus, the hippocampus, the midbrain, the pons, the medulla, and the cerebellum. Imaging was performed within a humid chamber with 37°C stable temperature and oxygen ow to delay loss of oxygenation and therefore alter parasite dynamics. Using these conditions parasites retain their motility for an average of 1 hour, during which we quanti ed cytoadhered and owing parasites in the vascular area of each region.

Image Analysis
Quanti cation of cytoadhered and owing parasites In order to quantify cytoadhered and flowing parasites, we used as reference the 70 kDa Cytoadhered parasites were defined as those not displaying throughout the time-lapse acquisition. These parasites often displayed movement of the body and flagellum, but despite these local movement, the bodies remained attached at the same spot of the vessel wall. To enable quantifications by vascular area, we determined the total area of the field of view, and extracted the vascular area using the FITC-Dextran and AF647-CD31. We calculated the percentage of vascular area (Av%) using the formula = , where A T is the total area of the field of view, A T is 100% and A v is the total area marked by FITC-Dextran and CD31. Then, we normalized the total quantity of cytoadhered and flowing parasites, to the calculated vascular area. These measurements were performed daily in T. congolense Tc1/148-infected C57BL/6J mice, throughout 6 days of infection, at peaks of infection in T. congolense IL3000-infected C57BL/6J mice, and at day 6 of infection in several antibody-treated C57BL/6J mice as well as RAG2 KO mice. In addition to vascular area, we quantified vascular diameter of vessels in each field of view, to calculate number of adhered and cytoadhered parasites in 4 categories of vessel diameters (0-10 µm, 10-20 µm, 20-40 µm, and > 40 µm). Vascular area and diameter measurements were calculated using Fiji software.

Quanti cation of parasite region enabling sequestration
To evaluate and quantify parasite regions enabling sequestration, we used spinning disc confocal microscopy as previously described. Visualization was done using an oil-immersion plan apochromat 100X objective with 1.4 NA and 0.17 WD. Throughout 5 minutes of time-lapse imaging at a rate of 20 frames per second, we determined whether cytoadhered parasites remained attached using the mid-body, the parasite posterior, the agellar tip, or variable. The latter was de ned as several regions of the parasite involved in sequestration, with the attachment point shifting throughout the time lapse imaged.
Quanti cation of kinetoplast and nuclear numbers, and nuclear area To quantify nuclear and kinetoplast numbers, and nuclear area in cytoadhered and owing parasites, visualization was done using an oil-immersion plan apochromat 100x objective with 1.4 NA and 0.17 WD.
In order to visualize the full parasite body, we generated confocal images of 12-16 stacks, with step sizes of 0.2 mm. To be able to capture the entire parasite body, we surgically stopped blood ow. This resulted in owing parasites continuing to displace, while cytoadhered parasites remained in the same position throughout the entire time-lapse acquired. Using the contrast generated by the FITC-Dextran, we de ned the parasite regions, and classi ed parasites into 3 categories: those having 1 kinteoplast and 1 nucleus (1K1N), 2 kinetoplasts and 1 nucleus (2K1N), and 2 kinetoplasts and 2 nuclei (2K2N). Of important note, the distance between the two nuclei in both T. congolense strains used in this work, is much smaller than the one observed in T. brucei parasites. For this reason, we also measured nuclear area. Nuclear segmentation was facilitated by the use of Hoechst 33342, and measurements were performed using Fiji software.

Leukocyte quanti cation by intravital microscopy
Intravital microscopy allowed us to observe an increase of nucleated, large cells in the vasculature of the brain in infected mice. This visualization is achieved by the contrast provided by the 70 kDa FITC-Dextran, and the nuclear labeling allowed by Hoechst 33342. Although these combined dyes do not allow the de nition of types and sub-types of leukocytes during infection, they allow segmentation and quanti cation of new cell populations compared to control uninfected mice. Having obtained information from intravital microscopy, we characterized immune cell populations in infected animals by ow cytometry, using speci c antibodies as described below.
Non-expressed genes were removed from the analysis and the remaining were ranked by differential expression Log2 fold change. Gene sets were de ned based on gene ontology for biological function.

Cytokine Pro ling
Blood was collected from mice either non-infected or infected with T. congolense 1/148 at days 3 and 6 post-infection by cardiac puncture. Blood was allowed to clot for 30 minutes at room temperature and then centrifuged at 1000 x g for 10 minutes at 4ºC. Serum was collected from each sample and added to an equal volume of PBS (pH 7.4). Samples were immediately frozen at -80ºC, and shipped in dry ice to Eve Technologies (Canada), where a Mouse Cytokine Array / Chemokine Array 31-Plex was performed, in duplicate.

Receptor Expression by Mean Fluorescent Intensity
To investigate the relative expression of endothelial receptors in different organs, we intravenously injected 20 µg per mouse, of uorescently conjugated antibodies against ICAM1, ICAM2 (BioLegend, conjugated to AF647), and VCAM1 (Invitrogen, conjugated to FITC). These were administered intravenously, by retroorbital injection, into either non-infected mice or mice infected with T. congolense 1/148 at day 6 of infection. We measured mean uorescent intensities of at least 100 vessels per organ in 3 separate mice using an LSM 880 Zeiss microscope, and a 40x oil objective (1.3 NA). were spotted in a radial pattern onto the center of the wells, and incubated in a cell culture incubator for 3 hours at 37ºC, 5% CO2. Plates were washed three times with PBS (pH 7.4) and 100 µl of BSA were added.
Plates were incubated again for 30 minutes at 37ºC, 5% CO 2 . Simultaneously, parasites were isolated from mouse by anion exchange chromatography (Lanham and Godfrey, 1970) and stained with 5mM Vybrant™ CFDA SE Cell Tracer dye (#V12883, Invitrogen) diluted 1000 times in trypanosome dilution buffer (TDB) (5 mM KCl, 80 mM NaCl, 1 mM MgSO4, 20 mM Na2HPO4, 2 mM NaH2PO4, 20 mM glucose, pH 7.4), and incubated for 25 minutes at 34ºC, 5% CO 2 . At the end of the incubation period, parasites were washed and resuspended in TDB, added to the previously-washed pre-coated plates, and incubated for 1 hour at 34ºC, 5% CO 2 . Plates were washed twice with PBS (pH 7.4) to remove unbound parasites. We added 100 µl of TDB to the washed plates, now containing only adhered parasites, and proceeded with live imaging on a Zeiss Cell Observer (Carl Zeiss Microimaging) with a 40X waterimmersion objective. We acquired 15 elds of view per replicate well (3 replicates), per condition (BSA alone, rICAM1, rCD36) using bright-eld light and green laser. We used Fiji software to count adhered parasite per eld of view and total area imaged.

SHIRPA Test
We performed the primary screen of the previously described SHIRPA protocol (Rogers et al., 1997),  Infected and non-infected mice were anaesthetized with iso urane and received 3µg of α-APC-CD45 intravenously, by retroorbital injection. Antibodies were allowed to circulate for a maximum of three minutes, mice were euthanized by cervical dislocation, 50-150 µl of blood were collected by cardiac puncture, added to 2 ml ACK lysis buffer (155mM Ammonium Chloride, 10 mM Potassium Bicarbonate, 0.1 mM EDTA), left to incubate at room temperature for 15 minutes for red blood cell lysis, and centrifuged for 5 minutes at 550 x g, 4ºC. The supernatant was discarded the procedure repeated. At the end, the cell pellet was gently resuspended in RPMI 1640 medium (#11875093, Gibco) supplemented with 10% FBS (#10270106, Gibco). Brains were dissected, added to 5ml supplemented RPMI 1640 medium, cut in small pieces, and incubated with 100 µg/ml DNAse and 1.5 mg/ml collagenase D for 30 minutes at 37ºC with periodic agitation. After the incubation, an additional 15ml of medium were added, the organs were forced to pass through a 70 µm cell strainer, and centrifuged at 550 x g for 5 minutes at 4ºC. Resulting pellet was resuspended in 3 ml of 40% percoll and passed to a 15 ml tube, after which 2 ml of 70% percoll were slowly added to the bottom. Samples were centrifuged for 30 minutes at 2400 rpm with no acceleration or brake. The interphase was carefully collected with a pipette, added to 5ml of supplemented RPMI 1640 medium, and centrifuged at 550 x g for 5 minutes at 4ºC. The supernatant was discarded by inversion and the pellet resuspended in 2ml of supplemented RPMI 1640 medium. Isolated blood and brain immune cells were counted on a hemocytometer, and between 2.5 x 10 5 and 2  Parasite presence induces a pro-in ammatory pro le in brain endothelial cells. A. Schematics of the methodology used to compare the transcriptomes of brain endothelial cells. RiboTag.PDGFb.iCRE mice, under Cre recombinase induction, were infected with 2x103 T. congolense 1/148 parasites intraperitoneally. At day 6 post-infection, mice were euthanized, perfused and brains were dissected and homogenized. The polysomes of endothelial cells were immunoprecipitated and the RNA extracted and converted into cDNA, which was sequenced on a NextSeq 500 platform as 75bp single-end reads (N = 4-5). Enriched gene sets upon infection (FDR< 0.05) detected by gene ontology-based gene set enrichment analysis (Subramanian et al., 2005), produced by WEBgestalt (Wang et al., 2017), and ordered by the normalized enriched score. E. Serum concentration of IL-1α, IL-1β, IFNγ, CXCL10, CXCL9, TNFα in noninfected mice and at days 3 and 6 post-infection (N = 3-4). F. Representative images of ICAM1, ICAM2, and VCAM1 expression in the brain endothelium of non-infected and infected mice, measured by uorescence. G. Mean uorescent intensity of ICAM1, ICAM2, and VCAM in the brain endothelium in noninfected mice and in mice infected with T. congolense 1/148, at day 6 post-infection. Black lines represent mean. Stars indicate statistically signi cant results; unpaired t-test, * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001. Characterization of immune cell populations in ltrating and circulating in the brain and systemic vasculature at day 6 post-infection with T. congolense 1/148. A. Representative images obtained by intravital microscopy, of leukocytes in the brain vasculature in non-infected mice, and mice infected with T. congolense 1/148, but treated with antibodies against IgG2 (isotype control), ICAM1, ICAM2, or VCAM1.
Nuclei are stained with Hoechst dye and shown in green. B. Percentage of leukocytes found in the brain