Human CCR5^high effector memory cells perform CNS parenchymal immune surveillance via GZMK-mediated transendothelial diapedesis

Abstract

Although the CNS is immune privileged, continuous search for pathogens and tumours by immune cells within the CNS is indispensable. Thus, distinct immune-cell populations also cross the blood–brain barrier independently of inflammation/under homeostatic conditions. It was previously shown that effector memory T cells populate healthy CNS parenchyma in humans and, independently, that CCR5-expressing lymphocytes as well as CCR5 ligands are enriched in the CNS of patients with multiple sclerosis. Apart from the recently described CD8⁺ CNS tissue-resident memory T cells, we identified a population of CD4⁺CCR5^high effector memory cells as brain parenchyma-surveilling cells. These cells used their high levels of VLA-4 to arrest on scattered VCAM1, their open-conformation LFA-1 to crawl preferentially against the flow in search for sites permissive for extravasation, and their stored granzyme K (GZMK) to induce local ICAM1 aggregation and perform trans-, rather than paracellular diapedesis through unstimulated primary brain microvascular endothelial cells. This study included peripheral blood mononuclear cell samples from 175 healthy donors, 29 patients infected with HIV, with neurological symptoms in terms of cognitive impairment, 73 patients with relapsing-remitting multiple sclerosis in remission, either 1–4 weeks before (n = 29), or 18–60 months after the initiation of natalizumab therapy (n = 44), as well as white matter brain tissue of three patients suffering from epilepsy. We here provide ex vivo evidence that CCR5^highGZMK⁺CD4⁺ effector memory T cells are involved in CNS immune surveillance during homeostasis, but could also play a role in CNS pathology. Among CD4⁺ T cells, this subset was found to dominate the CNS of patients without neurological inflammation ex vivo. The reduction in peripheral blood of HIV-positive patients with neurological symptoms correlated to their CD4 count as a measure of disease progression. Their peripheral enrichment in multiple sclerosis patients and specific peripheral entrapment through the CNS infiltration inhibiting drug natalizumab additionally suggests a contribution to CNS autoimmune pathology. Our transcriptome analysis revealed a migratory phenotype sharing many features with tissue-resident memory and Th17.1 cells, most notably the transcription factor eomesodermin. Knowledge on this cell subset should enable future studies to find ways to strengthen the host defence against CNS-resident pathogens and brain tumours or to prevent CNS autoimmunity.

CNS immune surveillance, granzyme K, CCR5, multiple sclerosis, HIV infection

Introduction

Efficient defence against pathogens and tumours (Ratnam et al., 2019) within the CNS is of paramount importance, as the brain is an organ with limited capacities for regeneration. However, to limit potential bystander damage via unregulated immune reactions, the CNS has evolved into an immune privileged site surrounded by the blood–brain barrier and the blood–CSF barrier (Ransohoff and Engelhardt, 2012; Russo and McGavern, 2015). Under homeostatic conditions, the blood–brain barrier consists of a tightly connected monolayer of endothelial cells, strictly regulating the entry of immune cells to the CNS (Greenwood et al., 2011; Engelhardt et al., 2017). Disturbance in the integrity of the blood–brain barrier leads to enhanced entrance of lymphocytes, associated with neuro-inflammatory disorders such as multiple sclerosis (Alvarez et al., 2011; Reich et al., 2018). So far, most studies have analysed immune cell infiltration into the CNS during pathological and inflammatory conditions. Studies investigating CNS immune surveillance in healthy individuals are sparse with regard to patrolling the parenchyma (Loeffler et al., 2011; Smolders et al., 2018) or focused on the CSF compartment (Svenningsson et al., 1995; Kleine and Benes, 2006). However, knowledge of immune cell subtypes and mechanisms involved in patrolling healthy CNS tissue is crucial for designing more specific therapeutics against immunological disorders allowing immune surveillance (i.e. to control infections and cancer), while inhibiting pathology-associated immune cell invasion into the brain (Engelhardt and Ransohoff, 2005; Vestweber, 2015). A prominent example of disturbed immune surveillance is the development of CNS opportunistic infections during HIV infection (Tan et al., 2012). C-C chemokine receptor 5 (CCR5) is a major co-receptor for HIV-1 entry to immune cells; accordingly, CCR5-expressing T cells are specifically depleted upon HIV infection (Grivel and Margolis, 1999). Thus, individuals lacking functional immune cell surface expression of CCR5 (CCR5-delta32 allele mutation) on lymphocytes are protected from infection by the CCR5-tropic HIV variant (Liu et al., 1996). A second example of potentially impaired CNS immune surveillance is found in West Nile fever, which is asymptomatic in the majority of infected patients. However, a minor fraction of individuals infected with the West Nile virus develop severe, potentially fatal encephalitis or meningitis (Hayes and Gubler, 2006). Interestingly, these patients have a four to eight times higher occurrence of the CCR5-delta32 allele with a resulting CCR5 deficiency (Glass et al., 2006). Accordingly, CCR5-mediated CNS immune surveillance was causally related to clearance of CNS viral burden and survival after West Nile virus infection in mice (Glass et al., 2005). Overall, these data indicate that CCR5-expressing T cells could be involved in anti-viral defence within the CNS.

For chronic inflammatory diseases of the CNS such as multiple sclerosis, it is conceivable that CNS infiltration by immune cells during phases of remission could be causally related to subsequent relapses (Geginat et al., 2017). Therefore, chemokines and their receptors, putatively encoding the target organ specificity of patrolling immune cells, have been investigated within this context. CCR5-, CCR2- and CXCR3-expressing T cells, as well as their ligands, could be detected in active multiple sclerosis brain lesions, but were also increased in the CSF in comparison to peripheral blood (Balashov et al., 1999; Sorensen et al., 1999; Sato et al., 2012), indicating potentially similar mechanisms of chemotactic attraction to CNS and CSF.

Memory CD4⁺ T cells are the dominant population in the CSF (Svenningsson et al., 1993) and during acute CNS inflammation mainly CCR7⁺ central memory T cells (T_CM) are found in CSF (Kivisäkk et al., 2004). We could previously show that under non-inflammatory conditions, as well as under treatment with the monoclonal VLA-4 antibody natalizumab, mainly T cells of the CD4⁺ effector memory (T_EM) phenotype are found in the CSF (Schneider-Hohendorf et al., 2014). The role of T_CM or T_EM cells in immune surveillance has been discussed, but is still a matter of research, especially in the human system (Sallusto et al., 1999; Lanzavecchia and Sallusto, 2000; Kleine and Benes, 2006; Gebhardt et al., 2013). In contrast to CSF, immune cell composition of the healthy CNS parenchyma consists of tissue-resident memory (T_RM) CD8⁺, but also CD4⁺ T cells (Smolders et al., 2018), both lacking CCR7 expression (Smolders et al., 2013). Together, this indicates that a T_EM subtype with a distinct C-C chemokine receptor (CCR) expression pattern and effector functions differing from its T_CM counterpart, is involved in human parenchymal immune surveillance (Ifergan et al., 2011).

In this study, we identify a CCR5-high-expressing T_EM subset with exceptionally high granzyme K (GZMK) expression present in healthy peripheral blood. These CCR5^high cells also reside in healthy brain tissue and show transcriptomic characteristics of Th17.1 cells, as well as similarities to T_RM cells. They are able to adhere to and to migrate through primary human endothelium under homeostatic conditions. On one hand, this immune-cell subpopulation is specifically depleted in HIV-positive patients with neurological symptoms. On the other hand, this cell population is more prevalent in peripheral blood of multiple sclerosis patients without current clinical activity. Moreover, blockade of CNS infiltration via treatment with natalizumab/anti-VLA-4 leads to specific sequestration of this population in the periphery. Therefore, these data strongly suggests a role of this cell population in CNS immune surveillance in healthy individuals with implications for infectious diseases with CNS tropism, as well as CNS autoimmunity.

Material and methods

Healthy donors and patients

Peripheral blood mononuclear cells (PBMCs), CNS tissue and CSF cells from 175 healthy donors and 109 patients were collected alongside voluntary blood donations, routine examinations or clinical interventions because of the underlying diagnosis [Patients (CSF) with non-inflammatory neurological disorders (NIND, n = 4)]. White matter brain tissue of patients with epilepsy was collected during brain surgery (n = 3). Tissue samples stemmed from non-epilepsy-afflicted regions of the resected material. PBMCs were collected from patients infected with HIV, presenting with neurological symptoms in terms of cognitive impairment (n = 29). A second set of PBMCs was collected from relapsing-remitting multiple sclerosis patients in remission, either 1–4 weeks before (n = 29), or 18–60 months after the initiation of natalizumab therapy (n = 44).

Standard protocol approvals, registrations, and patient consents

This non-interventional study was approved by the local ethics committee (University of Muenster: Ethik-Kommission der Ärztekammer Westfalen-Lippe und der Medizinischen Fakultät der Westfälischen Wilhelms-Universität, registration number: 2010–245-f-S) and informed written consent was obtained from all participants. This study was performed according to the Declaration of Helsinki.

Cells and cell culture

PBMCs, CD4⁺ and CD8⁺ cells were isolated from fresh human blood samples of healthy donors by density gradient centrifugation using phosphate-buffered saline (PBS) (Sigma-Aldrich), RosetteSep™ Human CD4⁺ T Cell Enrichment Cocktail (Stemcell Technologies Inc.) or RosetteSep™ Human CD8⁺ T Cell Enrichment Cocktail (Stemcell). Lymphoprep™ cells were cultivated in RPMI-1640 medium supplemented with 10% heat-inactivated foetal calf serum and 1% penicillin/streptavidin.

Brain tissue dissociation into single cell suspension was performed using the gentleMACS™ Octo Dissociator and the Adult Brain Dissociation Kit according to the manual (Miltenyi Biotech).

The used human brain microvascular endothelial cells (HBMECs) derived from autopsy tissue were purchased from PELOBiotech and have been characterized in detail previously (Schulte-Mecklenbeck et al., 2017). An additional phenotypical characterization of HBMECs is found in the Supplementary material (Supplementary Fig. 1A and B). HBMECs were grown in endothelial cell medium supplemented with foetal bovine serum, endothelial cell growth supplement and penicillin/streptomycin solution (ScienCell Research Laboratories). If indicated, HBMECs were treated with tumor necrosis factor-alpha (TNF-α) (stimulated condition) (10 U/ml; R&D Systems), for 18 h.

Isolation and fluorescence activated sorting of cells

For T-cell arrest under shear flow, crawling and diapedesis assays, PBMC, CD4⁺ or CD8⁺ cells were purified further using fluorescence activated cell (FAC) sorting with a FACSAria III Cell Sorter (BD Bioscience).

Flow cytometry

For labelling, PBMCs or T cell subpopulations were stained with fluorochrome-conjugated antibodies as described previously (Breuer et al., 2014). Example gating strategies are shown in Supplementary Fig. 1C and D.

Measurement of high affinity (HA)-LFA-1 was performed as described previously by using ICAM1–Fc/anti-Fc multimeric complexes (Dimitrov et al., 2018).

Video-microscopy shear flow assay

Shear flow assays were performed in µ-Slide I channel slides (IBIDI) containing a previously grown HBMEC monolayer (48–72 h) and captured with time-lapse live cell microscopy (×10 magnification; 2 frames/s) according to Steiner et al. (2010) and Wimmer et al. (2019). In brief, 10⁶ cells/ml were applied per channel with a shear rate of 0.25 dyn/cm² for 5 min to allow accumulation of lymphocytes and physical contact with the endothelial monolayer. Subsequently, shear force was increased to 1.5 dyn/cm² for 2 min to simulate physiological conditions. Arrested cells were quantified by analysing 15 fields of view alongside the channel. If indicated, cells were incubated with blocking antibodies against CD49d (natalizumab, Biogen), CD106 (clone: BBIG-V1) (R&D Systems) or CD11a (clone: HI111) (BD Biosciences). Videos were recorded using a BZ-9000 BioRevo microscope (Keyence) and BZ II viewer software (Keyence). The numbers of arrested cells per field of view was analysed using ImageJ software (NIH).

Video-microscopy crawling and diapedesis assay

Crawling and diapedesis was performed in µ-Slide 8 well plates (IBIDI) containing a previously grown HBMEC monolayer (48–72 h) and captured with time-lapse live cell microscopy (×10 magnification). 5 × 10⁴ cells FAC sorted T-cell subpopulations were added per slide. If indicated, cells were incubated with blocking antibodies against CD49d (natalizumab, Biogen), CD11a (clone: HI111) (BD Biosciences), CD54 (clone: HCD54) (BioLegend), CD102 (clone: CBR-IC2/2) (BioLegend), CCL3 (clone: 93321), CCL4 (clone: 24006) (both R&D Systems), CCL5/RANTES (clone: 2D5) (BD Biosciences) or chemical antagonists batimastat (Merck), maraviroc, (Sigma-Aldrich) or JNJ0996 (MedChemExpress). Videos were recorded using a BZ-9000 BioRevo microscope (Keyence) and BZ II viewer software (Keyence) (Supplementary Video 1). Distance, motility and numbers of crawling and migrating cells per field of view was analysed using ImageJ software.

Quantification of para- and transcellular diapedesis was conducted according to the method published by Mamdouh et al. (2009). Briefly, transmigrating CCR5^low, CCR5^high CD4⁺ T_EM and HBMECs were fixed with 4% paraformaldehyde (PFA) after diapedesis assays. To visualize cell borders and to distinguish T cells, fixed cells were incubated with antibodies targeting VE-cadherin (AF488-labelled; Thermo Fischer Scientific) and CD4 (Serotec) and stained with fluorochrome-conjugated antibody donkey IgG anti-rabbit IgG (H+L)-Cy3 (dianova). Image stacks obtained by fluorescence microscopy were analysed for quantification of diapedesis. T cells migrating in more than 3 μm distance from the junctions were counted as a transcellular diapedesis event, all other diapedesis events were classified as paracellular migration.

Immunofluorescence staining

Staining was performed as previously described (Breuer et al., 2017). A full list of antibodies is provided in Supplementary Table 1.

GZMK and actin staining for confocal microscopy

HBMECs were grown on glass coverslips for 48–72 h until they constituted a confluent cell layer. FAC sorted CCR5^high T_EM cells (5 × 10⁴) were added per coverslip. After 20 min of co-cultivation cells were fixed by adding 4% PFA for 10 min at room temperature. Cells were permeabilized by applying PBS + 100 mM glycine + 0.1% Triton™ X for 10 min at room temperature and blocked by PBS + 0.1% Tween20 + 1.5% BSA for 60 min at room temperature. Staining with antibody targeting GZMK (Santa Cruz) and subsequently with secondary goat anti-mouse Alexa Fluor® 488 conjugated antibody (Life Technologies) was performed for 60 min at room temperature in the dark each. Next, cells were incubated with Alexa647 Phalloidin (life technologies) for 10 min at room temperature in the dark and stained with DAPI (Sigma) in PBS for 10 min room temperature in the dark. Cells were mounted on object slides for microscopy with a LSM700 System on an inverted Axio Observer Z1 microscope (Zeiss) with Planapochromat lenses (Zeiss) and T-PMT transmissions-photomultiplier.

Microarray and RNAseq data analysis

After fluorescence readout of raw data (chip type: HuGene-2_0-st-v1) and background substraction using Affymetrix’ program NetAffx 1.1. Gene expression of microarray data were analysed using Illumina® GenomeStudio 2011 V2011.1 (Ilumina Inc., USA) software. Microarray data were clustered in R using the function ‘hclust’ according to Ward, the dendrogram was sorted using the packages ‘dendsort’ and ‘factoextra’ and cluster number was determined using gap statistics.

RNAseq data were clustered in R according to Ward and only genes are displayed passing a false discovery rate (FDR)-adjusted P-value of <10⁻⁶. See Supplementary material for further details of the RNAseq analysis used.

Statistical analysis

Data were analysed using the Prism Version 7.04 software (GraphPad). We did not assume normality of data and, therefore, two-tailed Mann-Whitney tests were used in the comparison of two non-parametric distributions and matched two-tailed Wilcoxon signed-rank test was used for different treatments or different subpopulations within the same patient/donor and for comparisons between peripheral blood and CSF. Data depicted in figures display mean ± standard deviation (SD). Measurements were taken from distinct samples. Differences were considered statistically significant with the following P-values: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; asterisk indicates FDR-passed in the case of RNAseq analysis (Fig. 5).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. Microarray data are publicly available in the ArrayExpress database (EMBL-EBI) with the accession number E-MTAB-7413. RNAseq data are publicly available in the SRA database (NCBI) with the bioproject accession number PRJNA503537. A detailed list of used material including RRID is provided in the Supplementary material (Supplementary Table 1).

Results

CCR5^high T_EM dominate human CNS parenchyma under non-inflammatory conditions

CCR5-expressing lymphocytes have been previously detected in multiple sclerosis CNS tissue (Balashov et al., 1999; Sorensen et al., 1999; Sato et al., 2012). Independently, a CCR7-negative T_EM phenotype has been reported to populate the CNS parenchyma of healthy individuals (Smolders et al., 2013). To evaluate whether CCR5 could also be involved in CNS immune surveillance without (recent) inflammation, we assessed freshly isolated brain tissue resections from patients with NIND (surgical resections) by flow cytometry. In line with previous reports on CD8⁺ T_RM in CNS of healthy individuals and multiple sclerosis patients (Machado-Santos et al., 2018; Smolders et al., 2018), tissue-derived T cells were shifted towards the CD8⁺ lineage in comparison to peripheral blood (Fig. 1A and B). While CD4⁺ and CD8⁺ CNS populations were dominated by T_EM, only few T_CM were present, and T_Naïve as well as T_EMRA were undetectable in CNS parenchyma (Fig. 1C and Supplementary Fig. 2A). In contrast to the periphery, the majority of CNS CD4⁺ and CD8⁺ T cells expressed CCR5 with a notable presentation of a CCR5^high expressing phenotype (Fig. 1D and Supplementary Fig. 2B; an exemplar gating for CCR5 subpopulations of healthy PBMCs is depicted in Fig. 1E). Additionally, markers associated with CD8⁺ T_RM cells, which are found on cells dwelling in non-lymphoid tissue with presumed prior inflammation (Schenkel and Masopust, 2014), were assessed. CD4⁺ and CD8⁺ CCR5^high T_EM lacked pronounced CD69 and CD103 expression, but highly expressed PD-1/CD279⁺ (Fig. 1F and Supplementary Fig. 2C).

Figure 1

CCR5^high T_EM dominate human CNS parenchyma under non-inflammatory conditions. (A) Quantification of the percentage of CD3⁺ T cells of healthy donor PBMCs (n = 13), brain tissue from patients with NIND (surgical resections) (n = 3), and CSF of patients with NIND (n = 4) expressing CD4 and CD8. (B) CD4/CD8 ratio of freshly isolated HD PBMC, brain tissue from patients with NIND (surgical resections) (n = 3) and CSF of patients with NIND (n = 4). (C) Quantification of the percentage of CD45RA⁺ T_Naïve, CD45RA⁻ CD62L⁺ T_CM, CD45RA⁻ CD62L⁻ T_EM CD45RA⁻ CD62L⁺ T_EMRA within CD4⁺ T cell of healthy donor PBMCs (n = 13) and brain tissue from patients with NIND (surgical resections) (n = 3), and CSF of patients with NIND (n = 4). (D) Quantification of the percentage of CCR5^neg, CCR5^low, CCR5^high expressing cells of healthy donor PBMCs (n = 13), brain tissue from patients with NIND (surgical resections) (n = 3), and CSF of patients with NIND (n = 4) within CD4⁺ T_EM. (E) Representative density plot of CCR5^neg, CCR5^low, CCR5^high staining of peripheral CD4⁺ T cells. (F) Quantification of the percentage of CCR5^high CD4⁺ T_EM of brain tissue from patients with NIND (surgical resections) (n = 3) expressing CD69, CD279 (PD-1) and CD103.

Although their CD4/CD8 ratio differed from CNS parenchyma and peripheral blood, CSF lymphocytes from NIND patients showed a similar enrichment of CCR5^high T_EM (Fig. 1A and E)_. Overall, this indicates that despite apparent differing propensities for tissue retention [as evident by CD4/CD8 ratios and the known propensity of CD8⁺ T cells for tissue retention (Gebhardt et al., 2011)], a consistent phenotype of CD4⁺ CCR5^high T_EM is attracted to healthy CNS parenchyma and CSF, potentially by a similar mechanism and chemoattractant.

Within CD4⁺ T_EM GZMA and GZMK expression is a unique feature of CCR5^high cells

The major known difference between T_CM and T_EM is the lack of lymphoid homing markers such as CD62L and CCR7 in T_EM. To screen for additional functional differences in depth, peripheral blood from three healthy donors was divided into T_Naïve, T_CM, T_EM, and transitional memory cells (CD45RA^lowCD45RO^low) by flow cytometry sorting. Extracted RNA was analysed using an Affymetrix HuGene-2_0-st-v1 chip comprising 352 859 probes. Cell subsets could be grouped into two major clusters, which can be viewed as naïve and memory clusters (Fig. 2A, detailed overview of genes and clustering is given in Supplementary Table 2). Analysis of the differential expression score showed that T_EM cells express significantly more GZMA and GZMK (Fig. 2B), than T_CM. This could be verified by flow cytometry analysis (Fig. 2C). GZMA and GZMK are serine proteases, tryptases, located in granules and share high structural and functional homologies with each other, but are functionally separate from molecules such as GZMB, which is mostly known for its function in cytotoxic CD8⁺ T cells and natural killer (NK) cells (Chowdhury and Lieberman, 2009).

Figure 2

Within CD4⁺ T_EM GZMA and GZMK expression is a unique feature of CCR5^high cells. (A) Heat map shows z-normalized expression of genes with at least one significant differential expression between CD45RA⁺ T_Naïve, CD45RA^low transitional CD45RA⁻ CD62L⁺ T_CM, CD45RA⁻ CD62L⁻ T_EM CD4⁺ T cells clustered according to Ward. (B) Heat map shows z-normalized expression of genes sorted by the most significant differential expression between T_CM and T_EM CD4⁺ T cells. (C) Quantification of the percentage of T_CM and T_EM CD4⁺ T cells expressing GZMA and GZMK. (D) Spearman correlation of GZMA/K and CCR5 expression by T_EM cells, r = Spearman’s rank correlation coefficient. (E) Representative density plot of GZMA, GZMK, and CCR5 staining of T_Naïve, T_CM, and T_EM CD4⁺ T cells. (F) Quantification of the percentage of CCR5^neg, CCR5^low, CCR5^high T_EM CD4⁺ T cells expressing GZMA and GZMK.

Unexpectedly, a predominant CCR5 expression by T_EM cells was not found on the RNA level. However, flow cytometric comparison of intra- and extracellular CCR5 expression showed that T_EM and T_CM cells contain similar amounts of cytosolic CCR5 but show differences in extracellular expression (Supplementary Fig. 3A). Extracellular CCR5 expression positively correlated with GZMA and GZMK expression (Fig. 2D) and only CCR5⁺ memory cells, predominantly T_EM cells, expressed GZMA and GZMK (Fig. 2E). Importantly, GZMA and GZMK expression within T_EM cells was not simply associated with CCR5, but with CCR5^high expressing cells (Fig. 2F), similar to the phenotype found in the CNS.

CD4⁺ CCR5^high T_EM show higher migratory capacity over unstimulated endothelium

To evaluate whether the identified CD4⁺ CCR5^high T_EM phenotype characterized by GZMA and GZMK expression could cross the uninflamed blood–brain barrier, we assessed their capacity for firm arrest, crawling, and crossing using primary unstimulated HBMECs in vitro. Spontaneous arrest of CD4⁺ CCR5 T_EM on unstimulated HBMECs was a rare event, as on average 56 of 2 × 10⁶ applied CD4⁺ CCR5^high T_EM (0.0028%) arrested under shear flow in the window of observation (Fig. 3A). Accordingly, VCAM1 expression was low without stimulation and scattered to singular endothelial cells (Supplementary Fig. 1A). CD4⁺ CCR5^high T_EM showed a 3-fold higher capacity for arrest under shear flow conditions, as compared to CCR5^low T_EM cells (Fig. 3A). This could be attributed to higher VLA-4 expression levels of CCR5^high T_EM (Supplementary Fig. 4A and B)_. T-cell arrest depended on VLA-4/VCAM1 interaction as their blockade completely abrogated CCR5^high T_EM arrest, whereas blockade of LFA-1 did not impair arrest (Fig. 3B).

Figure 3

CD4⁺ CCR5^high T_EM show higher migratory capacity over unstimulated endothelium. (A) Number of CCR5^low and CCR5^high CD4⁺ T_EM cells/15 fields of view that arrested on HBMEC monolayers (n = 6, fields of view = 24 mm²). (B) Number of CCR5^high CD4⁺ T_EM/field of view that arrested to HBMEC monolayers after incubation with anti-VCAM1 (on HBMECs, n = 6), anti-VLA-4 (on CCR5^high T_EM, n = 6), anti-LFA-1 (on CCR5^high CD4⁺ T_EM, n = 6) blocking antibodies (10 µg/ml), field of view = 24 mm². (C) Number of crawling CCR5^low and CCR5^high CD4⁺ T_EM /field of view against flow direction on HBMEC monolayers during arrest assays (n = 6), field of view = 1.6 mm². (D) Number of crawling and filopodia forming CCR5^low and CCR5^high CD4⁺ T_EM/field of view on HBMEC monolayers (n = 6), field of view = 1.6 mm². (E–G) Cell motility [(E) degree in per cent of locomotion; (F) speed excluding non-moving periods; (G) moving distance/30 min] of CCR5^high CD4⁺ T_EM on HBMECs. Graphs show individual cells of three donors and three independent experiments. (H) Number of crawling and filopodia forming CCR5^high CD4⁺ T_EM/field of view on HBMEC monolayers after incubation with anti-LFA-1 (on CCR5^high CD4⁺ T_EM, n = 6), anti-VLA-4 (on CCR5^high CD4⁺ T_EM, n = 6), anti-ICAM2 (on HBMECs, n = 6) or anti-ICAM2 (on HBMECs, n = 6) blocking antibodies (10 µg/ml), field of view = 1.6 mm². (I) Quantification of the degree of CCR5^low and CCR5^high CD4⁺ T_EM expressing open conformation HA-LFA-1 detected with fluorescent ICAM1–Fc/anti-Fc multimeric complexes. Normalized with corresponding negative control (n = 6, three independent experiments). (J) Number of migrated CCR5^low, CCR5^high CD4⁺ T_EM/field of view through HBMEC monolayers (n = 6) tracked via live cell microscopy, field of view = 1.6 mm². (K) Number of migrated CCR5^high CD4⁺ T_EM/field of view through a HBMEC monolayer tracked via live cell microscopy after incubation with anti-LFA-1 (on CCR5^high CD4⁺ T_EM, n = 6), anti-VLA-4 (on CCR5^high CD4⁺ T_EM, n = 6), anti-ICAM1 (on HBMECs n = 6) or anti-ICAM2 (on HBMECs n = 6) blocking antibodies (10 µg/ml), field of view = 1.6 mm².

Subsequent to arrest, T cells crawl along the luminal surface of the endothelium to detect potential spots for diapedesis. Confirming previous reports CCR5^high T_EM crawling was observed preferentially against the direction of shear flow (Fig. 3C; an exemplar crawling cell against the flow direction is shown in Supplementary Fig. 4C) (Coisne et al., 2013). The number of crawling, filopodia and uropod forming CCR5^high T_EM cells was higher as compared to CCR5^low cells (Fig. 3D; an exemplary crawling cell is shown in Supplementary Video 1). While locomotion (amount spent moving in the window of observation) of crawling cells was similar for both subsets (Fig. 3E), CCR5^high T_EM showed higher motility as assessed by crawling speed and the resulting covered distance (Fig. 3F and G), fitting to their increased LFA-1 expression (Supplementary Fig. 4D and E). LFA-1/ICAM2 interaction was crucial for cell motility as blocking either prevented crawling, whereas blockade of ICAM1 or VLA-4 did not influence crawling events (Fig. 3H). It was shown previously that ICAM1 expression on CNS capillaries is low under steady-state conditions (Laschinger et al., 2002), which was also seen in the used HBMECs in vitro (Supplementary Fig. 1B) in contrast to ICAM2, which was expressed constitutively (Supplementary Fig. 1B) (de Fougerolles et al., 1991). Cell shape and F-actin remodelling, as well as the subsequent forming of filopodia and uropods, are highly associated with open conformation HA-LFA-1 and interactions with its ligands (Porter et al., 2002; Abadier et al., 2015). Therefore, the amount of ex vivo available HA-LFA-1 was assessed (Dimitrov et al., 2018). HA-LFA-1 for both CCR5^low and CCR5^high T_EM cells was expectedly low, but twice as high for CCR5^high T_EM cells compared to CCR5^low T_EM (Fig. 3I).

Similar to arrest, diapedesis through unstimulated endothelium was a rare event (diapedesis-performing cells are shown in Supplementary Video 1). However, CCR5^high T_EM cells showed a unique ability to cross the unstimulated endothelial monolayer (Fig. 3J). Blockade of LFA-1, but not VLA-4 on CCR5^high T_EM cells, and ICAM2, but not ICAM1, on HBMECs, abrogated diapedesis (Fig. 3K).

GZMK but not GZMA expression is a prerequisite of CCR5^high lymphocytes’ transcellular diapedesis

Under steady-state conditions, blood–brain barrier endothelial cells express none to very low amounts of the three ligands of CCR5 (CCL3, MIP-1α; CCL4, MIP-1β; CCL5, RANTES) (Ebnet et al., 1996; Chui and Dorovini-Zis, 2010), which was also seen in HBMECs in vitro (Supplementary Fig. 5A and B). Blocking neither CCL3, CCL4, CCL5, nor CCR5 impaired CCR5^high T_EM diapedesis (Supplementary Fig. 5C). However, chemotaxis of CCR5⁺ cells towards CCL5 was expectedly CCR5-dependent: CCR5-blockade abrogated chemotaxis towards CCL5 in a Boyden-chamber migration assay without HBMECs (Supplementary Fig. 5D). CCR5 was, therefore, deemed to indicate the direction of migration/target organ, but not to be functionally involved in the step of transendothelial diapedesis over unstimulated endothelium.

As diapedesis could almost exclusively be performed by CCR5^high T_EM and this subset predominantly expresses GZMA and GZMK, the potential involvement of these proteases in mediating the process of diapedesis was assessed. The number of diapedesis events correlated with the percentage of cells positive for both GZMA and GZMK (Fig. 4A). To investigate individual roles of the GZMA and GZMK, diapedesis of other cell subpopulations apart from CD4⁺, previously found in CNS parenchyma (Gross et al., 2016) expressing intersections of GZMA and GZMK (Smolders et al., 2018), was assessed. In contrast to CD4⁺ T_EM, where GZMA and GZMK expression correlates with CCR5 expression, all CD8⁺ T_EM and NK cells expressed GZMA. However, only CD8 CCR5^high T_EM and NK^bright cells expressed GZMK (Fig. 4B and C, left, and Supplementary Fig. 5E) and only these subsets were able to perform diapedesis over unstimulated HBMECs indicating a distinguished role for GZMK (Fig. 4B and C, right).

Figure 4

GZMK but not GZMA expression is a prerequisite of CCR5^high lymphocytes’ transcellular diapedesis. (A) Spearman correlation of GZMA/K expression and number of diapedesis events through a HBMEC monolayer by CCR5^high CD4⁺ T_EM cells. (B) Left: Representative density plot of GZMK and CCR5 staining of CD4⁺ T_EM T cells. Right: Number of migrated CCR5 CCR5^low, CCR5^high CD8⁺ T_EM /field of view through a HBMEC monolayer (n = 3) tracked via live cell microscopy, field of view = 1.6 mm². (C) Left: Representative density plot of GZMK and CD56 staining of NK cells. Right: Number of migrated CD56^dim, CD56^bright NK cells/field of view through a HBMEC monolayer (n = 6) tracked via live cell microscopy, field of view = 1.6 mm². (D) Number of migrated CCR5^high CD4⁺ T_EM/field of view through a HBMEC monolayer tracked via live cell microscopy after incubation with batimastat (50 µM, n = 6) or JNJ0966 (10 µM; n = 6). (E) Quantification of the percentage of HBMECs expressing ICAM1 after incubation with recombinant human GZMK (250 nM, n = 6) and batimastat (50 µM, n = 3) for min, field of view = 1.6 mm². (F) Representative immunofluorescence images of in vitro transmigrating CCR5^high CD4⁺ T_EM cells (white arrow) through unstimulated HBMECs for the discrimination of trans- (left) paracellular (right) diapedesis. Immunofluorescence staining for VE-Cadherin (green), CD4 (red), and nuclear staining (DAPI, blue). Scale bar = 30 µm. (G) Quantification of the percentage of transcellularly migrating CCR5^low, CCR5^high CD4⁺ T_EM through unstimulated HBMEC/field of view, (n = 7, field of view = 5 mm²). (H) Exemplary transmission electron microscopic image of in vitro transcellular migration CCR5^high T_EM cells (depicted in blue) across HBMECs (depicted in red). Endothelial nuclei are indicated by a dashed white line. Scale bar = 1.3 µm. (I) Exemplary confocal microscopic images of in vitro transmigrating CCR5^high T_EM cells (white arrow) through unstimulated HBMEC. Immunofluorescence staining for GZMK (green), actin (red) and nuclear staining (DAPI, blue). Scale bar = 10 µm.

Figure 5

CCR5^high CD4⁺ T_EM are a phenotypically and functionally distinct subpopulation sharing features of Th17.1 and T_RM. (A) Heat map shows z-normalized expression of RNAseq data of CCR5^low versus CCR5^high CD4⁺ T_EM of four healthy donors. (B) Heat map of normalized gene expression in CCR5 CCR5^neg, CCR5^low and CCR5^high. Depicted are genes that are associated with either the Th1, Th2, Th17 or Th17.1 phenotype according to Monaco et al. (2019) (for each phenotype top 100 FDR, log2FC > 0 genes, genes found in more than one phenotype were dropped). Association of the genes with the respective phenotype is colour-coded in the second, qualitative heat map. (C) Heat map of normalized gene expression in CCR5^neg, CCR5^low and CCR5^high for genes that were shown to be up- or downregulated in human CD4⁺ CD69⁺ T_RM cells according to Kumar et al. (2017). (D) Dot plots of genes commonly associated with either the Th1, Th2, Th17, or Th17.1 phenotype according to CCR5 expression (negative, low or high) extracted from the RNAseq normalized counts. Colour-coded boxes show association of genes with phenotype. Error bars represent standard deviation. Comparisons that passed FDR in the RNAseq analysis are marked with an asterisk.

Immunofluorescence staining of GZMK showed that the protein is heavily stored in granules (Supplementary Fig. 5F). As prevention of transcription would, therefore, not suffice, a direct inhibition of GZMK was necessary to prove involvement of GZMK in this process. As there are no agents or antibodies available to block GZMK specifically, the synthetic broad spectral protease inhibitor batimastat, which is mainly used as an antagonist of matrix metallopeptidases (MMP), was evaluated for its capability to block GZMK. It inhibits enzymes by peptide mimicking and competitive binding to the active sites (Botos et al., 1996; Rasmussen and McCann, 1997). Cleaving activity of recombinant human GZMK against β-tubulin, a known specific substrate of GZMK (Bovenschen et al., 2009), was investigated and revealed that batimastat decreased GZMK activity (Supplementary Fig. 5G). Accordingly, application of this inhibitor abrogated the number of CCR5^high T_EM diapedesis events (Fig. 4D). This finding was specific for the diapedesis performed by CCR5^high T_EM under homeostatic conditions, as applying this inhibitor to CCR5^neg T_EM on stimulated HBMECs showed no effect (Supplementary Fig. 5H). Of note, it was shown previously that CCR5⁺ T cells, enriched in CSF of patients with multiple sclerosis, express MMP9 after stimulation (Sato et al., 2012), and MMP9 was shown to be involved in blood–brain barrier diapedesis (Song et al., 2015). However, MMP9 expression was low ex vivo with no differences within CCR5^neg, CCR5^low or CCR5^high T_EM (Supplementary Fig. 5I). In line with this, specific blockade of MMP9 in CCR5^high T_EM did not interfere with diapedesis (Fig. 4D).

As it had been shown previously that T cells degranulate and can secrete GZMA towards endothelial cells upon arrest and T-cell receptor triggering and that this degranulation can induce surface expression of endothelial adhesion molecules (Manes and Pober, 2014), we hypothesized that GZMK interaction could lead to increased surface expression of cell adhesion molecules on endothelial cells, as well. Incubation of HBMECs with GZMK induced ICAM1 surface expression within minutes, which could be reversed by simultaneous application of batimastat (Fig. 4E). We further hypothesized that the induction of ICAM1 could be mediated by endothelial protease activated receptors (PAR). It was shown that GZMK can activate endothelial PAR (Cooper et al., 2011; Sharma et al., 2016), which induces cytoskeletal rearrangement. The used HBMEC expressed PAR-1 to PAR-3, but not PAR-4 and incubation of HBMECs with GZMK increased expression of endothelial PAR-2 (Supplementary Fig. 5J and K).

As the endothelial tight junctions should be firmly closed under steady state conditions, putatively hampering paracellular transmigration, and GZMK can induce endothelial surface ICAM1, we hypothesized that the observed diapedesis of CCR5^high T_EM might favour the transcellular pathway. Therefore, diapedesis events were spatially analysed by staining the endothelial junctions with a fluorescently-labelled antibody for VE-cadherin (Fig. 4F). Compared to CCR5^low, CCR5^high T_EM predominantly performed transcellular diapedesis (Fig. 4G), indicating that GZMK-mediated endothelial ICAM1 induction facilitates transcellular diapedesis (Fig. 4H). Staining of CCR5^high T_EM for GZMK and actin during migration revealed that GZMK was polarized at the uropod and absent in the pseudopod (visualized by actin staining) (Fig. 4I). Consistent with this finding, it was previously shown that GZMB in murine cytotoxic lymphocytes is polarized in the uropod during migration on a Matrigel® matrix (Prakash et al., 2014).

CCR5^high CD4⁺ T_EM are a phenotypically and functionally distinct subpopulation sharing features of Th17.1 and T_RM

CCR5⁺ expressing cells have previously been associated mainly with the Th1 phenotype (Bonecchi et al., 1998; Loetscher et al., 1998). To analyse the molecular repertoire of CCR5^high T_EM, bulk RNAseq analysis of CCR5^high T_EM versus CCR5^low T_EM from four healthy donors was performed (Fig. 5A). Up- and downregulated genes, as defined by a FDR-adjusted P-value of <10⁻⁶, are shown in Supplementary Fig. 6A and B. The complete gene list and raw data are listed in the Supplementary Table 3. Relevant genes for CNS migration (e.g. CXCR6, CCL4, CCR3 and EOMES) were upregulated for CCR5^high T_EM. To analyse the pathways significantly modified within this population, the KEGG database (Hiroyuki et al., 2017) was used (Supplementary Fig. 6C). Consistent with differences found in their migration and diapedesis pattern, the cytokine-cytokine receptor pathway was the most prominent result. Lineage analysis referring to Monaco et al. (2019) and Kumar et al. (2017) revealed clustering within the genes characteristic for the Th17.1 phenotype (TBX21, RORC, EOMES; Fig. 5B). Fitting to their detection in CNS parenchyma and their PD-1 expression, congruence with the T_RM cell phenotype was also observed (Fig. 5C). Genes commonly associated with either the Th1, Th2, Th17, or Th17.1 phenotype according to expression by CCR5^neg, CCR5^low and CCR5^high are depicted in Fig. 5D. We validated cytokine expression of CCR5^low versus CCR5^high T_EM via flow cytometry and could show that both expressed equally high levels of TNF-α and low levels of IL-17, IL-10 and GM-CSF. CCR5^high T_EM expressed twice as much of the Th1 characterizing cytokine IFN-γ (Supplementary Fig. 6D). CD4⁺ T_EM express low amounts of transcription factor FoxP3, compared to T_Naïve and T_CM. Expression was absent in the CCR5^high fraction (Supplementary Fig. 6E). Flow cytometric analysis of CXCR3 and CCR6 showed that CCR5 is highly co-expressed with CXCR3, but CCR6 expression on CCR5^high was lower than on CCR5^low T_EM (Supplementary Fig. 6F). The characterization of the CNS surveilling cells’ cellular setup is depicted visually in Supplementary Fig. 6G.

Alterations in peripheral CCR5^high GZMK⁺ cells of HIV⁺ and multiple sclerosis patients indicate clinical relevance

As CCR5 is a co-receptor for HIV infection of CD4⁺ T cells and CD4⁺ CCR5⁺ cells are specifically depleted by the CCR5-tropic HIV strain, we hypothesized that HIV-positive patients presenting with neurological symptoms should have impaired CNS immune surveillance, demonstrated by low percentages of CCR5^high T_EM cells. Despite their increased GZMA expression (Fig. 6A), flow cytometric analysis of PBMCs of these HIV-positive patients revealed decreased amounts of CCR5^high GZMK⁺ CD4⁺ T_EM in the periphery as compared to healthy individuals (Fig. 6B). Accordingly, CD4 counts positively correlated with the amount of remaining GZMK⁺ cells among CCR5^high CD4⁺ T_EM, indicating that among CCR5⁺ T cells, CCR5^high GZMK⁺ cells are specifically depleted upon HIV infection (Fig. 6C).

Figure 6

Changes in the peripheral prevalence of CCR5^high GZMK⁺ cells of HIV-positive and multiple sclerosis patients emphasize their importance in both pathologies. (A) Quantification of the percentage of CD4⁺ T_EM of healthy donors (n = 8) and HIV-positive patients (n = 29) expressing GZMA. (B) Quantification of the percentage of CD4⁺ T_EM of HD (n = 28), HIV-positive patients (n = 29), treatment naïve multiple sclerosis (n = 28), and natalizumab treated patients (n = 44) being CCR5^high GZMK⁺ double positive. (C) Spearman correlation of % CCR5^high GZMK⁺ T_EM and CD4 count of HIV-positive patients. Spearman rho = 0.4363, P = 0.0292, n = 22.

In contrast to the absence in HIV, presence of CCR5-expressing lymphocytes is associated with multiple sclerosis pathology. Concordant with these findings, we found CCR5^high GZMK⁺ CD4⁺ T_EM to be increased in peripheral blood of patients with multiple sclerosis without clinical activity. Treatment of patients with multiple sclerosis with the VLA-4 blocking antibody natalizumab specifically sequestered CCR5^high GZMK⁺ among peripheral CD4⁺ T_EM (Fig. 6B).

Discussion

CCR5 expression has been previously shown to be associated with the recruitment of T cells to sites of inflammation (Kawai et al., 1999) as well as with multiple sclerosis (Balashov et al., 1999; Sorensen et al., 1999; Sato et al., 2007), but was also found on T cells in the CSF of patients with NIND (Kivisäkk et al., 2002). As CCR5 was shown to be dispensable for diapedesis, its expression seems to determine its target organ via chemotactic stimuli (Shulman et al., 2012), which fits with findings of CCR5 ligands in brain lesions (Balashov et al., 1999). CCR5^high T_EM predominantly express GZMK and the amount of ex vivo expression correlated with the number of diapedesis events in vitro. However, in contrast to CCR5 inhibition, blockade of granzymes successfully abrogated diapedesis. Although inhibition with the antagonist batimastat is not as specific for GZMK as an antibody could be, this finding provides strong evidence for the importance of GZMK in CCR5^high T_EM cells for diapedesis across unstimulated endothelium. In a previous study we could show that GZMK expressing T and NK cells are found in brain lesions of patients with multiple sclerosis (Gross et al., 2016) supporting the view that this molecule might be of general importance for the CNS infiltration of immune cells. Recently, GZMK together with CCR5 were found to be associated with CD4⁺ T_RM (Smolders et al., 2018; Beura et al., 2019). Granzymes are a family of serine proteases located in granules and are predominantly known for their intracellular cytotoxic role in immunity and cancer (Cullen et al., 2010). Nevertheless, during the last decade, the importance of the extracellular function of granzymes has emerged, as it was shown that they could play a role in diapedesis of T_EM (Manes and Pober, 2014). Lymphocytes had been shown previously to predominantly migrate through endothelium via the paracellular pathway, while transcellular diapedesis was observed to a lesser extent (Lutz et al., 2017; Wimmer et al., 2019). Here, we show that CCR5^high GZMK⁺ T_EM preferentially use the transcellular pathway under steady-state conditions. We hypothesize that CCR5^high T_EM can use GZMK to engage the endothelial cell via the activation of PAR, which has already been suggested (Cooper et al., 2011; Sharma et al., 2016), possibly via induction of cytoskeletal rearrangement through Rho activation (Vouret-Craviari et al., 1998). Additionally, our group recently found that in mice PAR-2 engagement is necessary for blood–brain barrier inflammation by inducing ICAM1 and VCAM1 expression, thereby influencing endothelial transmigration of lymphocytes in the context of experimental autoimmune encephalomyelitis (Göbel et al., 2019). Data from our current study corroborate that engagement of endothelial cells by GZMK leads to higher surface expression of ICAM1 within minutes with no possibility of transcription or translation, indicating that ICAM1 is readily stored in these cells and transported quickly to the cell surface.

The high ex vivo presence of CCR5^high CD4⁺ T_EM in the CNS of donors without CNS inflammation reflects and validates the in vitro observations. In line with a recent study, we found more memory CD8⁺ T cells (compared to CD4⁺) in the CNS of healthy donors (Smolders et al., 2013), whereas the CSF is mainly populated by memory CD4⁺ T cells (Svenningsson et al., 1993; Schneider-Hohendorf et al., 2014), indicating differing propensities for CD4⁺ and CD8⁺ tissue retention. Accordingly, CD8⁺ T_RM reside in non-lymphoid tissues confined to the site of infection to maintain immediate local immune surveillance upon reinfection (Schenkel and Masopust, 2014), whereas CD4⁺ T_RM were shown to exert higher motility in tissues with a wider recirculation pattern (Gebhardt et al., 2011). CD8⁺ T_RM have been detected in multiple compartments, including the CNS (Owens et al., 2016; Steinbach et al., 2016; Machado-Santos et al., 2018; Smolders et al., 2018). They are often characterized by their expression of CD69, CD103 and PD-1 (Kumar et al., 2017). However, it was suggested that some T_RM, depending on the tissue, might lose or lack CD69 expression (Walsh et al., 2019). Especially, CD4⁺ T_RM might, therefore, be less confined to the site of initial infection (Collins et al., 2016). Although we could only detect PD-1 in CNS parenchyma-derived lymphocytes, transcriptome analysis strongly suggests congruence with a migratory T_RM phenotype, especially because of high eomesodermin (EOMES) expression. EOMES has been shown to regulate migration to tissues via modulation of T-bet and CXCR3 (Knox et al., 2014). Additionally, single nucleotide polymorphisms of EOMES were recently confirmed to be strongly associated with multiple sclerosis in genome-wide association studies (GWAS) (Sawcer et al., 2011; International Multiple Sclerosis Genetics Consortium, 2019). The transcription factor T-box 21 (TBX21), where the expression has been shown to positively correlate with EOMES expression in multiple sclerosis patients (Parnell et al., 2014), was also strongly upregulated in CCR5^high T_EM. Given the fact that multiple sclerosis has repeatedly been associated with viral infection (Ascherio and Munger, 2007; Belbasis et al., 2015), an increased immune surveillance elicited by e.g. EBV or JCV activity might contribute to multiple sclerosis risk and pathology. EOMES has recently been identified as the most prominent transcription factor for Th17.1 (also called ex-Th17 or non-classical Th1) cells. This cell type is characterized by CXCR3 (Th1) and CCR6 (Th17) expression (while lacking CCR4 and IL-17), but also CCR5, and had been associated with multiple sclerosis pathology early on (Kebir et al., 2009), and again very recently (Paroni et al., 2017; Van Langelaar et al., 2018). Jelcic et al. (2018) recently described the association of a brain-homing T cell population of CXCR3⁺ IFN-γ⁺ IL-17⁻ to multiple sclerosis, which roughly describes the CCR5^high T_EM in our study. Fitting to these recent observations, we found higher frequencies of these cells in peripheral blood of multiple sclerosis patients in periods of clinical stability, corroborating that these presumed CNS-homing cells might contribute to disease initiation and perpetuation.

The role of CD4⁺ CCR5⁺ cells in immune surveillance has been a focus in the research of acquired immune deficiency syndrome (AIDS) caused by infection with HIV (Okoye and Picker, 2013) and it was recently shown that increased gut-homing through T-cell β7-integrin expression leads to a more severe disease course in HIV patients, as the gut is a primary reservoir for HIV (Sivro et al., 2018). The virus uses CCR5 as a co-receptor to infect CD4⁺ T cells (Chan et al., 1999; Hung et al., 1999), which are then rapidly and specifically depleted (Grivel and Margolis, 1999; Brenchley et al., 2004). This would be consistent with our data suggesting this cell population to confer immune surveillance, as HIV-positive patients without treatment often present with neurological symptoms of their disease (Mateen et al., 2012) and that this depletion was shown to correlate with the degree of neurological symptoms (Power et al., 1994). We could show as an additional proof of concept ex vivo that CD4 counts of these patients positively correlated with the presence of these cells. Surprisingly, one possible compensatory mechanism of upregulating GZMA could be observed in CD4⁺ T_EM cells of HIV patients. This has previously been shown to be a misguided maturation process in HIV-positive patients specifically impairing defence against the virus (Haridas et al., 2003). Consistent with these studies, the finding that these patients still presented with neurological symptoms supports our hypothesis that GZMA is not critically important in the CNS immune surveillance process. This fits together with GZMA expression in other non-CNS infiltrating cell types such as CCR5^neg CD8⁺ T cells and CD56^dim NK cells.

Taken together, our study identifies a subpopulation of CD4⁺ T_EM cells, phenotypically identified by their high expression of CCR5 and functionally employing GZMK to migrate through the blood–brain barrier under homeostatic conditions and subsequently be found in the CNS. Knowledge of this population could be used either to prevent autoimmune pathology in the context of multiple sclerosis or to improve CNS immune surveillance in patients with or at risk of neuro-infections or brain tumor development e.g. by designing chimeric antigen receptor (CAR) T cells (Johnson et al., 2015; Brown et al., 2016) with the necessary receptors to mediate their function within the CNS.

Abbreviations

Abbreviations
 
• HBMEC

  human brain microvascular endothelial cell

 
• NIND

  non-inflammatory neurological disorders

 
• PBMC

  peripheral blood mononuclear cell

 
• T_CM

  central memory T cells

 
• T_EM

  effector memory T cells

 
• T_RM

  tissue-resident memory cells.

Acknowledgements

The authors would like to thank Petra Kotte, Barbara Meyring, and Jila Akbari for excellent technical assistance. Many thanks to Vivian Rieping for help with electron microscopy work.

Funding

This study was funded by the Deutsche Forschungsgemeinschaft (DFG) Grant CRC128 Project B01 to N.S. and A.Z., A09 to C.C.G. and H.W., A10/Z2 to H.W., as well as the Kompetenznetz Multiple Sklerose (Competence Network for Multiple Sclerosis) funded by the Federal Ministry of Education and Research (FKZ 01GI1308B 01GI0907) to H.W.

Competing interests

T.S.H. received research and travel support from Novartis. A.S.M. received funding from Novartis. C.J. received travel support from Novartis. CCG received speaker honoraria, travel expenses for attending meetings, and research support from Biogen, Euroimmun, Genzyme, Mylan, Novartis, and Bayer Health Care. Her work is funded by the German Ministry for Education and Research (BMBF; 01GI1603A), the German Research Foundation (DFG; GR3946/3–1 and SFB/Transregio 128 A09), the European Union (Horizon2020, ReSToRe), the Interdisciplinary Center for Clinical Studies (IZKF) Munster (Kl3/010/19), and Novartis. N.S. received honoraria for advisory boards and travel expenses from Novartis and Sanofi-Genzyme. H.W. received honoraria and consultation fees from Bayer Healthcare, Biogen Idec, Fresenius Medical Care, GlaxoSmithKline, GW Pharmaceuticals, Merck Serono, Novartis Pharma, Sanofi Genzyme, and TEVA Pharma. All other authors report no competing interests.

Author notes

Sebastian Herich and Tilman Schneider-Hohendorf authors contributed equally to this work.