Novel insights into neuroinflammation: bacterial lipopolysaccharide, tumor necrosis factor α, and Ureaplasma species differentially modulate atypical chemokine receptor 3 responses in human brain microvascular endothelial cells

Background Atypical chemokine receptor 3 (ACKR3, synonym CXCR7) is increasingly considered relevant in neuroinflammatory conditions, in which its upregulation contributes to compromised endothelial barrier function and may ultimately allow inflammatory brain injury. While an impact of ACKR3 has been recognized in several neurological autoimmune diseases, neuroinflammation may also result from infectious agents, including Ureaplasma species (spp.). Although commonly regarded as commensals of the adult urogenital tract, Ureaplasma spp. may cause invasive infections in immunocompromised adults as well as in neonates and appear to be relevant pathogens in neonatal meningitis. Nonetheless, clinical and in vitro data on Ureaplasma-induced inflammation are scarce. Methods We established a cell culture model of Ureaplasma meningitis, aiming to analyze ACKR3 variances as a possible pathomechanism in Ureaplasma-associated neuroinflammation. Non-immortalized human brain microvascular endothelial cells (HBMEC) were exposed to bacterial lipopolysaccharide (LPS) or tumor necrosis factor-α (TNF-α), and native as well as LPS-primed HBMEC were cultured with Ureaplasma urealyticum serovar 8 (Uu8) and U. parvum serovar 3 (Up3). ACKR3 responses were assessed via qRT-PCR, RNA sequencing, flow cytometry, and immunocytochemistry. Results LPS, TNF-α, and Ureaplasma spp. influenced ACKR3 expression in HBMEC. LPS and TNF-α significantly induced ACKR3 mRNA expression (p < 0.001, vs. control), whereas Ureaplasma spp. enhanced ACKR3 protein expression in HBMEC (p < 0.01, vs. broth control). Co-stimulation with LPS and either Ureaplasma isolate intensified ACKR3 responses (p < 0.05, vs. LPS). Furthermore, stimulation wielded a differential influence on the receptor’s ligands. Conclusions We introduce an in vitro model of Ureaplasma meningitis. We are able to demonstrate a pro-inflammatory capacity of Ureaplasma spp. in native and, even more so, in LPS-primed HBMEC, underlining their clinical relevance particularly in a setting of co-infection. Furthermore, our data may indicate a novel role for ACKR3, with an impact not limited to auto-inflammatory diseases, but extending to infection-related neuroinflammation as well. AKCR3-induced blood-brain barrier breakdown might constitute a potential common pathomechanism. Electronic supplementary material The online version of this article (10.1186/s12974-018-1170-0) contains supplementary material, which is available to authorized users.


Background
The two human Ureaplasma species (spp.) Ureaplasma (U.) urealyticum and U. parvum are among the smallest self-replicating pathogens and, unlike other bacteria, lack a cell wall [1]. They are typical commensals of the adult urogenital tract and are generally considered as being of low virulence [1].
Meningitis and neuroinflammation profoundly contribute to neonatal morbidity and mortality [16]. Long-term sequelae develop in up to 25-50% and comprise cerebral palsy, hydrocephalus, or neurodevelopmental impairment [17,18]. While the typical pathogens causing neonatal meningitis, such as Escherichia (E.) coli and group B Streptococcus [19], are well described, little is known about Ureaplasma-driven inflammation of the neonatal central nervous system (CNS). Although Ureaplasma spp. can be commonly detected in the cerebrospinal fluid of preterm neonates [20,21], clinical data on Ureaplasma-induced meningitis are scarce [14,15], and in vitro data are nonexistent so far. We therefore aimed to establish a cell culture model of Ureaplasma meningitis to assess the inflammatory pathomechanisms involved.
Cytokines and their subgroup chemokines act as key regulators in inflammation and immune activation [22]. Recently, atypical chemokine receptors (ACKRs) have been described, which seem to bear a crucial impact on autoinflammatory CNS disorders such as multiple sclerosis (MS) or its animal model, experimental autoimmune encephalomyelitis (EAE) [23,24]. Unlike other receptors, ACKRs do not seem to induce G protein signaling, but act as scavenger receptors [25]. As one of the four ACKRs described to date, ACKR3 (also known as CXCR7) binds C-X-C motif chemokine ligand (CXCL) 11 and CXCL12 [23,26]. Its activation ultimately permits leukocyte entry into the CNS [27] and furthermore seems to mediate migration of activated microglia [23,24], the latter strongly influencing multiple CNS diseases such as MS or Alzheimer's disease, but also stroke or brain traumata, and even psychiatric diseases [28]. In animal models, ACKR3 was shown to be expressed within various brain regions, with an upregulation observable in case of inflammation, as for instance in EAE [29]. ACKR3 increases were furthermore found upon cytokine stimulation or induction of hypoxia in human astrocytes [30]. To date, only a single study addressed ACKR3 expression and inducibility in human brain endothelial cells [26]. We evaluated ACKR3 responses of human brain microvascular endothelial cells (HBMEC) upon exposure to bacterial lipopolysaccharide (LPS) and tumor necrosis factor-α (TNF-α), and furthermore used ACKR3 levels to depict the interactions of Ureaplasma spp. with native and LPS-primed HBMEC.

Stimulation assays
For mRNA analysis and flow cytometry, HBMEC were transferred to gelatin-coated six-well culture plates (Greiner Bio-One) at a density of 2 × 10 5 cells/well and grown to confluence for 48 h. Cells were washed, and 1 ml fresh growth medium was added per well. For immunocytochemistry, HBMEC were seeded on gelatincoated 24-well culture plates (Thermo Fisher) at a density of 5 × 10 4 cells/well and cultivated for 48 h. Cells were washed, and 0.5 ml fresh growth medium was added per well. Uu8 and Up3 suspensions were fourfold concentrated by centrifugation and resuspended in fresh broth and 10 9 -10 10 CCU in 250 μl broth were added per ml of HBMEC medium. CCU were determined by 10-fold titration, as described previously [31], and corresponding genome equivalents were identified (Institute of Medical Microbiology and Hospital Hygiene, Duesseldorf, Germany). 10 9 -10 10 CCU/ ml equated to 5 × 10 7 -6 × 10 8 copy numbers/ml. Ureaplasma concentrations used in this study corresponded with in vivo amniotic fluid levels in pregnancies with premature rupture of membranes < 37 weeks, which were reported to range from 4 × 10 2 to 5 × 10 7 copy numbers/ml in one study [32], while another described bacterial numbers between 1.9 × 10 7 and 1.1 × 10 12 /ml [33]. Bacterial viability was verified by simultaneous culture on selective agar plates (medco Diagnostika GmbH, Ottobrunn, Germany). For negative controls in selected experiments, samples of Ureaplasma spp. were heat-killed at 60°C for 15 min. LPS from E. coli serotype 055:B5 (Sigma-Aldrich) was added to HBMEC at a concentration of 100 ng/ml. TNF-α (Sigma-Aldrich) was used at a concentration of 10 ng/ml. Doses of LPS, TNF-α, and Ureaplasma spp. were determined by preliminary experiments (Additional file 2) analogous to previous approaches [26,31,[34][35][36], testing 100, 500, 1000, and 2000 ng/ml LPS; 10, 40, and 100 ng/ml TNF-α; and 0.2, 0.3, 0.4, 0.5, and 1 × 10 9 -10 10 CCU/ml Ureaplasma isolates. According to the results of preliminary time kinetic experiments (Additional file 2) with 2-, 4-, 6-, 20-, 24-, 30-, 36-, and 48-h incubation periods, exposure times of 4 and 30 h were chosen for mRNA analysis, while flow cytometry assays were performed after 24 and 48 h. Unstimulated HBMEC served as negative controls. In order to adjust for potential confounding effects of the in-house medium, HBMEC exposed to Ureaplasma isolates were compared to broth control in all experiments.

RNA extraction and reverse transcriptase PCR (RT-PCR)
Total RNA was extracted using NucleoSpin® RNA Kit (Macherey-Nagel, Dueren, Germany) according to the manufacturer's protocol. Total RNA was eluted in 60 μl nuclease-free H 2 O (Sigma-Aldrich), quantified using a Qubit® 2.0 Fluorometer (Thermo Fisher), and stored at − 80°C until reverse transcription. For RT-PCR, 1 μg of total RNA was reverse transcribed using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher) according to the manufacturer's instructions. First-strand cDNA was diluted 1:10 with deionized, nuclease-free H 2 O (Sigma-Aldrich) and stored at − 20°C until analysis.

RNA sequencing
Total RNA was extracted using NucleoSpin® RNA Kit (Macherey-Nagel) according to the manufacturer's protocol, and samples were stored at − 80°C until further processing. Experiments were repeated three times (n = 3). Library preparation was performed at the Core Unit Systems Medicine, University of Wuerzburg, Germany, according to the Illumina TruSeq stranded mRNA Sample Preparation Guide, using the Illumina TruSeq stranded mRNA Kit (both: Illumina, San Diego, CA, USA) with 700 ng of input RNA and 13 PCR cycles. Thirteen to fourteen libraries were pooled and sequenced on a NextSeq 500 (Illumina) with a read length of 75 nucleotides. A total of~34-40 million raw reads per library were produced and processed using FastQC 0.11.5 [38] for assessing read quality, amount of duplicates, and presence of adapter sequences. The Illumina TruSeq adaptors were cleaved using cutadapt (version 1. 14) [39]. Reads were additionally trimmed keeping a quality drop value below a mean of Q20. The processed sequences were mapped to the human genome using the short read aligner STAR (version 2.5.2b) [40] with genome and annotation files retrieved from GENCODE (version 25-March 2016 freeze, GRCh38). For all samples, the proportion of reads mapped to the human reference genome ranged between 76 and 90% in total. Sequences aligning to specific genes were quantified using bedtools subcommand intersect (version 2.15.0) [41]. Differentially expressed genes were identified with the help of DESeq2 (version 1.16.1) [42]. Only genes with a Benjamini-Hochberg corrected p value below 0.05 were classified as significantly differentially expressed. To compare different groups, reads per kilo base per million mapped reads (RPKM) were calculated for individual genes using DGEList and RPKM function from edgeR [43].

Flow cytometry
Cells were harvested and incubated with 4 mg/ml Gamunex 10% (Grifols, Frankfurt, Germany) for Fcreceptor blocking. Cells were separated by centrifugation and stained with APC-H7-conjugated fixable viability dye (eBioScience, San Diego, CA, USA), a BV510conjugated antibody to cluster of differentiation (CD) 31 (platelet endothelial cell adhesion molecule-1, BD Biosciences), and an APC-conjugated antibody to surface ACKR3 (BioLegend, San Diego, CA, USA). After centrifugation, cells were resuspended in PBS (Sigma-Aldrich) containing 1% human serum (Biochrom GmbH) and samples were read on a FACSCanto™ II flow cytometer (BD Biosciences), acquiring a minimum of 10,000 events to be analyzed with FACSDiva v6.1.3 software (BD Biosciences). Events were gated via forward and side scatter. A SSC-height versus FSC-width dot plot was used to exclude doublets, and cells were gated for viability-dye negative events. CD31 functioned as endothelial marker. Experiments were repeated five times (n = 5).

Statistical analysis
PCR, RNA sequencing, and flow cytometry results were analyzed by a one-way ANOVA with Tukey's multiple comparisons test using Prism® 6 software (GraphPad Software, San Diego, CA, USA). To account for normal distribution, data were transformed logarithmically using the equation LN(x + 0.1). Throughout this manuscript, multiplicity adjusted p values were calculated with logtransformed data, whereas untransformed data were employed for diagrams or calculation of x-fold change and standard deviation (SD). Differences with p < 0.05 were considered to be statistically significant. Data are shown as mean ± SD.

Basal expression of ACKR3 mRNA in HBMEC
With cycle threshold values between 22.9 and 24.9 in native HBMEC, qRT-PCR results indicated relevant amounts of ACKR3 mRNA even in unstimulated HBMEC. Basal ACKR3 gene expression was confirmed by RNA sequencing, with ACKR3 RPKM ranging from 1.8 to 2.8 in unstimulated cells.
ACKR3 response of HBMEC upon co-stimulation with LPS and Ureaplasma spp.
RNA sequencing revealed a non-significant trend towards higher ACKR3 mRNA levels following 4 h Ureaplasma exposure of LPS-primed cells (Fig. 2b).

Immunocytochemistry
Native HBMEC as well as those stimulated with LPS or TNF-α exhibited a faint and homogenous cytoplasmic staining for ACKR3 (Fig. 3). Incubation with Ureaplasma spp. enhanced the cytoplasmic signal. Upon Ureaplasmastimulation of LPS-primed HBMEC, we additionally observed intense submembranous granules.

Discussion
A compromised integrity of the blood-brain barrier (BBB) is a common pathological feature in numerous inflammatory disorders of the CNS [44,45]. Recently, ACKR3 has been recognized as a contributor to BBB impairment by permitting passage of inflammatory cells into the CNS in the event of inflammation [27].
The present study is the first to analyze proinflammatory influences of Ureaplasma spp. as well as LPS and TNF-α on HBMEC, which constitute a major component of the BBB [45]. Our data reveal that all given stimuli are able to induce ACKR3 responses in HBMEC. These results (i) suggest a novel role of ACKR3 not only in autoimmune neuroinflammation, but also in CNS infections; (ii) introduce an in vitro model of Ureaplasma meningitis; and (iii) for the first time indicate a proinflammatory capacity of Ureaplasma spp. in HBMEC.
The current data demonstrate basal ACKR3 mRNA expression in unstimulated HBMEC. These results are in accordance with the only previous study addressing ACKR3 expression in HBMEC [26]. In addition, we found a significant increase in ACKR3 mRNA upon stimulation not only with TNF-α, as described in one previous study [26], but also with LPS. While TNF-α is an endogenous proinflammatory cytokine particularly involved in autoimmune disorders [46,47], LPS, as a component of bacterial membranes, is widely used to experimentally mimic gramnegative bacterial infections. Our data on LPS-induced ACKR3 amplification therefore propose an involvement of ACKR3 not only in autoimmune neuroinflammatory disorders, but also in infection-associated CNS diseases. These results are in line with previous studies describing LPSinducible ACKR3 expression in choroid endothelial cells [48], a colorectal carcinoma cell line [49], and pulmonary This study is the first to address the interactions of viable Ureaplasma isolates with cells constituting the BBB. While pro-inflammatory cytokine responses to Ureaplasma colonization or infection were described for chorioamnion [51] and lung [52], no such correlation was made for the CNS to date. By demonstrating an ability of Ureaplasma spp. to induce ACKR3 in HBMEC, our data provide evidence of a pro-inflammatory capacity of Ureaplasma spp. in brain endothelial cells. Several case reports identified Ureaplasma spp. as causative in neonatal meningitis [14,15], and some authors furthermore described a higher risk for intraventricular hemorrhage (IVH) or cerebral palsy (CP) in Ureaplasma-colonized preterm infants [2,53,54]. Taking together these observations and our in vitro findings, the clinical relevance of Ureaplasma spp. in inflammatory CNS disorders may be much higher than contemplated to date. Ureaplasma-driven neuroinflammation may be mediated by ACKR3, potentially in terms of a consecutive BBB breakdown, ultimately allowing the influx of inflammatory cells into the CNS.
Flow cytometry and immunocytochemistry data from HBMEC co-stimulated with LPS and Ureaplasma spp. indicate impairments of BBB integrity particularly in the event of co-infection with Ureaplasma spp. and a second pathogen. While evoking only moderate ACKR3 responses themselves, Ureaplasma spp. caused an overproportional ACKR3 increase in HBMEC if combined with LPS. These results are in line with in vitro studies in human monocytes, describing an aggravation of LPS-induced inflammatory responses by Ureaplasma spp. [31,55,56]. Furthermore, clinical data revealed a more intense intrauterine inflammation in polymicrobial chorioamnionitis with Mycoplasma or Ureaplasma spp. and other bacteria compared to amniotic infection with other bacteria alone [57]. Together, this may indicate an immunomodulatory capacity of Ureaplasma spp., which is likely to be of particular clinical relevance. Polymicrobial colonization and co-infections are a common problem particularly in an intensive care setting, and Ureaplasma spp. may predispose for invasive CNS infections with other pathogens. However, clinical data addressing this topic are pending.
The exact mechanism of ACKR3 induction by Ureaplasma spp. is yet to be determined. However, results from a knock-out experiment in a colorectal carcinoma cell line suggest that LPS-induced increase in ACKR3 expression might be mediated by Toll-like receptor (TLR) 4 [49]. Considering that TLR signaling is crucial in mediating pro-inflammatory effects evoked by Ureaplasma spp. [58,59], TLRs might be involved in the pathway of Ureaplasma-driven ACKR3 induction as well. In earlier studies, we could demonstrate an Ureaplasma-induced TLR2 mRNA increase in neonatal and adult monocytes [31]. Opposed to TLR4, generally being a potent activator of LPS-induced immune responses, TLR2 primarily mediates bacterial lipopeptide recognition [60]. Differences in TLR engagement may therefore be a potential explanation for the diverging effects of LPS and Ureaplasma spp. on mRNA and protein levels observed in this study. These aspects need to be addressed in further studies.
ACKR3 acts as a receptor for chemokines CXCL11 and CXCL12 [23,26] and is known to scavenge its ligands to some extent [25,61]. CXCL12 restricts leukocyte entry into the CNS [62] and is therefore relevant in limitation of inflammation. Animal studies indicate a beneficial role for CXCL12 in post-ischemia white matter injury [63], whereas antagonist studies provide evidence that an internalization of CXCL12 by ACKR3 and the resulting decline of CXCL12 allow consecutive inflammatory parenchymal infiltration [27]. Loss of CXCL12 has been observed in MS [64], but may be relevant in other inflammatory morbidities as well. In accordance with these observations, we could demonstrate reduced CXCL12 mRNA levels in HBMEC upon exposure to LPS and TNF-α. The second receptor ligand CXCL11 and its interactions with ACKR3 are less well characterized, but CXCL11 also seems to relevantly contribute to EAE [65] and is adjudged a role in chemotaxis of mononuclear cells [65,66]. Inverse to CXCL12, we found an increase of CXCL11 in response to LPS and TNF-α stimulation of HBMEC. In line with previous studies [27,65,66], these results may provide evidence for CXCL12 loss and CXCL11 increase being the causal link between ACKR3 elevation and consecutive neuroinflammation. Furthermore, ACKR3 mediates microglial chemotaxis, relevant in several neuroinflammatory morbidities [28], by activation of extracellular signal-regulated kinases (ERK) 1/2 [24]. This mechanism offers yet another possible explanation for the relevance of ACKR3 in inflammatory diseases of the CNS.
Moreover, given the abovementioned protective role of CXCL12 in white matter integrity [63], our results may imply a relevance of ACKR3 in other diseases potentially associated with Ureaplasma spp., such as IVH or CP. AKCR3 might be a missing link between pathogen and morbidity, an aspect that should be addressed in future studies.
The strength of this study relates to the use of two different, viable Ureaplasma serovars and the assessment of pro-inflammatory responses at the level of mRNA synthesis and protein expression as well as visualization of ACKR3 expression via immunocytochemistry.
One potential limitation of this study is the use of a commercially available single donor cell line. Although commonly used and widely accepted, these cell lines bear a certain risk of not being fully representative. Furthermore, this study was conducted with an adult cell line. Considering the risk cohort of preterm and term neonates, a freshly isolated cell line, ideally of neonatal origin, is desirable and should be employed in further studies. Moreover, HBMEC used in this study originated from cortex tissue. Although a recent study found an equal distribution of typical BBB markers in parenchymal and meningeal vessels [67], cortical HBMEC may still differ from the meningeal endothelial cells also contributing to the BBB, and results should not be unconditionally transferred from one to the other.

Conclusions
While BBB breakdown has generally been acknowledged as being a pathophysiological basis of neuroinflammatory diseases, the impact of ACKR3 has only just begun to be appreciated. Whereas, to date, ACKR3 has been adjudged a key role primarily in autoimmune neuroinflammatory disorders, our results suggest its additional involvement in infectious diseases of the CNS. Ureaplasma spp. and other pathogens might be able to initiate a cascade beginning with ACKR3 induction, proceeding to BBB breakdown, and ultimately cumulating in inflammatory brain injury. This may be mediated by CXCL12 loss and CXCL11 increase, as some of our results suggest.
This study is the first to introduce an in vitro model of Ureaplasma meningitis. Our findings of Ureaplasma-induced ACKR3 expression demonstrate a proinflammatory capacity of Ureaplasma spp. in HBMEC. Data from co-stimulated HBMEC indicate an immunomodulatory capacity of Ureaplasma spp. in the event of co-infection, which may be of particular clinical relevance.