More than just inflammation: Ureaplasma species induce apoptosis in human brain microvascular endothelial cells

Background Ureaplasma species (spp.) are commonly regarded as low-virulent commensals but may cause invasive diseases in immunocompromised adults and in neonates, including neonatal meningitis. The interactions of Ureaplasma spp. with host defense mechanisms are poorly understood. This study addressed Ureaplasma-driven cell death, concentrating on apoptosis as well as inflammatory cell death. Methods Human brain microvascular endothelial cells (HBMEC) were exposed to Ureaplasma (U.) urealyticum serovar 8 (Uu8) and U. parvum serovar 3 (Up3). Resulting numbers of dead cells as well as mRNA levels and enzyme activity of key agents in programmed cell death were assessed by flow cytometry, RNA sequencing, and qRT-PCR, respectively. xCELLigence data were used for real-time monitoring of changes in cell adhesion properties. Results Both Ureaplasma isolates induced cell death (p < 0.05, vs. broth). Furthermore, Ureaplasma spp. enhanced mRNA levels for genes in apoptosis, including caspase 3 (Up3 p < 0.05, vs. broth), caspase 7 (p < 0.01), and caspase 9 (Up3 p < 0.01). Caspase 3 activity was increased upon Uu8 exposure (p < 0.01). Vice versa, Ureaplasma isolates downregulated mRNA levels for proteins involved in inflammatory cell death, namely caspase 1 (Uu8 p < 0.01, Up3 p < 0.001), caspase 4 (Uu8 p < 0.05, Up3 p < 0.01), NOD-like receptor pyrin domain-containing 3 (Uu8 p < 0.05), and receptor-interacting protein kinase 3 (p < 0.05). Conclusions By inducing apoptosis in HBMEC as main constituents of the blood-brain barrier, Ureaplasma spp. may provoke barrier breakdown. Simultaneous suppression of inflammatory cell death may additionally attenuate host defense strategies. Ultimate consequence could be invasive and long-term CNS infections by Ureaplasma spp. Electronic supplementary material The online version of this article (10.1186/s12974-019-1413-8) contains supplementary material, which is available to authorized users.


Background
Commonly colonizing the adult urogenital tract, the two human Ureaplasma species (spp.) Ureaplasma (U.) urealyticum and U. parvum are generally regarded as low-virulent commensals [1]. Nonetheless, vertical transmission in pregnancy occurs frequently and appears to be inversely related to maturity [2]. Intra-amniotic Ureaplasma infections increase the risk for chorioamnionitis, premature rupture of membranes, and preterm birth [3][4][5]. Despite ongoing research, however, the implications of a postnatal Ureaplasma colonization or infection remain unresolved and appear to be underestimated so far [6]. As well as provoking invasive infections in immunocompromised adults [7][8][9], Ureaplasma spp. may cause pneumonia and sepsis in preterm and term neonates [10,11]. Furthermore, a growing number of case reports describe Ureaplasma spp. as causative agents in neonatal meningitis [12,13]. Considering typical sequelae of meningitis like cerebral palsy or neurodevelopmental impairment [14,15], potentially bearing long-term health implications, Ureaplasma spp. may have to be regarded of considerable relevance particularly in preterm and term neonates.
Nonetheless, in vitro data addressing the pro-inflammatory capacity of Ureaplasma spp. are scarce [16][17][18]. We recently established a cell culture model of Ureaplasma meningitis [19], using human brain microvascular endothelial cells (HBMEC), important constituents of the blood-brain barrier (BBB) and among the first cells to encounter pathogens seeking entry into the central nervous system (CNS) [20]. Having detected Ureaplasma-induced responses of atypical chemokine receptor 3, which may ultimately mediate BBB breakdown, we were the first to provide in vitro evidence of Ureaplasma-driven neuroinflammation [19].
Recent studies propose a close association of inflammation and cell death [21,22]. Induction of cell death in HBMEC is a mechanism some pathogens employ, presumably to gain entrance into the CNS by an impairment of BBB integrity [23][24][25][26][27]. Inflammatory forms of cell death, such as pyroptosis and necroptosis, can be distinguished from the so-called immunologically silent process of apoptosis [28]. As described in Fig. 1, meticulous cascades are involved in either pathway and different forms of cell death are furthermore closely interlinked. Caspases are key mediators among all three of them, with caspases 4 and 5 being involved in inflammatory cell death and caspases 3, 7, and 9 primarily mediating apoptosis [28][29][30]. Caspases are produced as inactive pro-enzymes which have to be activated by cleavage as a part of the cascades illustrated in Fig. 1 [28,29].
We used our newly established cell culture model of Ureaplasma meningitis to assess induction of cell death with particular focus on caspase levels upon exposure of HBMEC to Ureaplasma spp.

Stimulation assays
For qRT-PCR, RNA sequencing, and flow cytometry, HBMEC were seeded in gelatin-coated 6-well culture plates (Greiner Bio-One) at a density of 2 × 10 5 cells/well and cultivated for 48 h. Confluent monolayers were washed, and 1 ml fresh growth medium was added per well. As described previously [19], 250 μl broths containing 10 9 -10 10 CCU Ureaplasma were inoculated per milliliter of HBMEC medium. One hundred nanograms per millilter bacterial lipopolysaccharide (LPS, Escherichia (E.) coli serotype 055:B5, Sigma-Aldrich) was added to a subgroup of HBMEC. Cells were stimulated for 4 and 30 h for mRNA analysis and 24 and 48 h for flow cytometry.
For impedance-based real-time monitoring of transendothelial resistance (xCELLigence), HBMEC were transferred to gold electrode-coated plates (Omni Life Science, Bremen, Germany) at a density of 1.25 × 10 4 cells/well and cultivated in 200 μl growth medium for 48 h. At this point, cells were stimulated in duplicates as described above or left without stimulation.
Inocula and incubation periods had been determined in preliminary experiments [19] analogous to previous approaches [17,[31][32][33][34]. Unstimulated HBMEC accounted for negative controls. To adjust for potentially confounding broth effects, cells exposed to Ureaplasma isolates were additionally compared to broth control throughout the experiments. In selected experiments, heat-killed Ureaplasma isolates (60°C for 15 min) were additionally used as negative controls.

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 instructions. 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 protocol. First strand cDNA was diluted 1:10 with deionized, nuclease-free H 2 O (Sigma-Aldrich) and stored at − 20°C until further processing.

RNA sequencing
After the extraction of total RNA (NucleoSpin® RNA Kit, Macherey-Nagel), samples were stored at − 80°C until analysis. Experiments were repeated three times (n = 3). Library preparation was conducted by the Core Unit Systems Medicine, University of Wuerzburg, Germany, using the Illumina TruSeq stranded mRNA Kit (Illumina, San Diego, CA, USA) according to the manufacturer's instructions, with 700 ng of input RNA and 13 PCR cycles. Thirteen to fourteen pooled libraries were sequenced on a NextSeq 500 (Illumina) with a read length of 75 nucleotides, producing~34-40 million raw reads per library. These were assessed for read quality, amount of duplicates, and presence of adapter sequences employing FastQC 0.11.5 [36]. The Illumina TruSeq adaptors were cleaved using cutadapt (version 1.14) [37], and reads were trimmed keeping a quality drop value below a mean of Q20. Processed sequences were mapped to the human genome using the short read aligner STAR (version 2.5.2b) [38], retrieving genome and annotation files from GENCODE (version 25-March 2016 freeze, GRCh38). The proportion of reads mapped to the human reference genome ranged from 76 to 90% in total for all samples. Sequences aligning to specific genes were quantified with the help of bedtools subcommand intersect (version 2.15.0) [39]. Differentially expressed genes were identified using DESeq2 (version 1.16.1) [40], and differences with a Benjamini-Hochberg corrected p value < 0.05 were considered as significant. For the comparison between different groups, reads per kilo base per million mapped reads (RPKM) were calculated employing DGEList and RPKM function from edgeR [41].

Flow cytometry
Cells were harvested, separated by centrifugation, and stained with Fixable Viability Dye eFluor TM 780 (eBioScience, Thermo Fisher), a dye labelling dead cells. After centrifugation, cells were resuspended in phosphatebuffered saline (PBS, Sigma-Aldrich) and fixed using fixation buffer (BioLegend, San Diego, CA, USA). Centrifugation and permeabilization in permeabilization wash buffer (BioLegend) were followed by staining with antibodies to cleaved caspase 3 (Alexa Fluor 647 conjugated, Cell Signaling Technology, Danvers, MA, USA), caspase 8 (unconjugated, Abcam, Cambridge, UK), and cleaved caspase 9 (PE conjugated, Cell Signaling Technology). Cells were separated by another centrifugation step and were afterwards stained with an Alexa Fluor 405-conjugated secondary antibody (Life Technologies, Thermo Fisher Scientific). After centrifugation, cells were resuspended in PBS containing 1% human serum (Biochrom GmbH) and specimens were read on a FACSCanto™ II flow cytometer (BD Biosciences). A minimum of 10,000 events were acquired and analyzed with FACSDiva v6.1.3 software (BD Biosciences). For viability analysis, all events were included and viability dye positive cells were considered dead. The exact gating strategy for caspase analysis is described in Additional file 1. Experiments were repeated three times (n = 3).
xCELLigence real-time cell monitoring The xCELLigence system was used to continuously monitor cell adhesion properties. Cells are cultivated in culture plates equipped with microelectrodes, which allow a computer to measure electrical impedance [42]. The xCELLigence software (version 1.2.1.1.002) converts collected data to a cell index. For xCELLigence measurements, HBMEC on gold electrode-coated plates (Omni Life Science) were placed in an ACEA xCELLigence DP system (Omni Life Science) and real-time monitoring of transendothelial resistance was initiated. When forming confluent monolayers, as indicated by a plateau of the cell index, cells were stimulated as described above.
Real-time monitoring was continued for 48 h after stimulation. To compare independent experiments, the cell index was normalized to the untreated control and the time point of stimulation. Experiments were repeated five times (n = 5).

Statistical analysis
qRT-PCR, RNA sequencing, and flow cytometry results were analyzed by a one-way ANOVA followed by Tukey's multiple comparisons test employing Prism® 6 software (GraphPad Software, San Diego, CA, USA). xCELLigence data were assessed by a two-way ANOVA and subsequent Bonferroni test. The significance threshold for p values was set at < 0.05. Data are shown as means ± standard deviation (SD).

Ureaplasma-driven cell death in HBMEC
Numbers of viability dye positive cells, considered dead, were determined by flow cytometry (Fig. 2) (Fig. 2). LPS did not have any significant effect on cell viability at 24 h, but significantly enhanced numbers of dead cells after 48 h (1.24-fold ± 0.3, p = 0.0085, vs. control; Fig. 2).
Caspase 8 mRNA expression in HBMEC was generally rather low, as deducible from RPKM values no higher than 0.25 (Fig. 3d). No significant effects of Ureaplasma stimulation on mRNA or protein levels could be detected compared to control cells and broth control (Fig. 3d-f).
A shorter stimulation period of 4 h did not provoke any significant Ureaplasma-driven mRNA effects if compared to broth and control. Ureaplasma isolates did mostly not differ significantly from one another, and broth usually did not have any effect itself.

Discussion
This study is the first to provide in vitro insights into strategies Ureaplasma spp. use to differentially interfere with programmed cell death in HBMEC. Programmed cell death is an immune defense mechanism host cells employ to eliminate particularly intracellular pathogens [28]. On the less favorable side, however, cell death may also benefit the pathogen by reduction of immune cell numbers or impairment of physiological barriers. For the first time, we could provide evidence for Ureaplasma-driven induction of apoptosis and simultaneous suppression of inflammatory forms of cell death in HBMEC.
Whereas caspases are important contributors to all forms of programmed cell death (Fig. 1), caspases 3, 7, and 9 seem to be primarily involved in apoptosis [29]. Our data demonstrate an Ureaplasma-driven upregulation of caspase 3 and caspase 9 mRNA after 30 h as well as an increase in enzyme activity (Fig. 3a-c, g-i). Similarly, caspase 7 mRNA expression was enhanced after 30 h of Ureaplasma exposure (Fig. 3j)  furthermore observe a trend towards elevated mRNA levels for BAK, FOS, and APAF1 (Fig. 3k, n, o) as additional pro-apoptotic proteins. These results suggest an ability of Ureaplasma spp. to induce apoptosis. Acknowledging the role of the individual proteins in the apoptotic cascade ( Fig. 1), Ureaplasma spp. appear to primarily activate the so-called intrinsic apoptotic pathway in HBMEC.
Ureaplasma-driven upregulation of FOS mRNA furthermore indicates an activation of the extrinsic pathway as well, although FASL, FADD, and caspase 8 as downstream effectors do not seem to be involved on mRNA levels. This Ureaplasma-driven induction of apoptosis is reflected in a significant increase in cell death upon 24 and 48 h exposure of HBMEC to Ureaplasma spp. (Fig. 2). Whereas induction of apoptosis in HBMEC has been described for other pathogens or bacterial components, including listeriolysin O as well as E. coli shiga toxin and hemolysin [25][26][27], we are the first to provide evidence for Ureaplasma-driven apoptosis in HBMEC. As main components of the BBB, apoptosis of HBMEC may consecutively result in BBB breakdown. We could indeed verify Ureaplasma-driven impairment of HBMEC barrier properties by continuously monitoring cell adhesion properties using the xCELLigence technique. Of note, other authors using the xCELLigence system described loss of cell adhesion in apoptotic but not in necrotic cell death [43].
Recently having demonstrated the upregulation of atypical chemokine receptor 3 in HBMEC by Ureaplasma spp., potentially causing BBB impairment [19], induction of apoptosis appears to be yet another method these pathogens employ to disturb the protective CNS barrier. Numerous neuroinflammatory diseases are facilitated by a compromised BBB integrity [20,44]. Ureaplasma-driven BBB impairment may (1) allow Ureaplasma entry into the CNS, (2) predispose for invasive CNS infections with other pathogens, and (3) facilitate inflammatory cell influx into the CNS, causing chronic intracerebral inflammation. In line with this, Ureaplasma spp. have not only been identified as causative pathogens in neonatal meningitis, but have also been associated with cerebral palsy or intraventricular hemorrhage in preterm infants [6,12,13,45,46]. Of note, white matter injury, in particular, has also been associated with endothelial cell apoptosis and BBB breakdown [47]. Ureaplasma-driven induction of apoptosis and BBB impairment could thus be the common pathological feature in several neuroinflammatory diseases of prematurity.
We could recently demonstrate Ureaplasma-induced downregulation of apoptotic caspase mRNA in pulmonary epithelial cells, but a converse upregulation of apoptotic caspase protein or activity in pulmonary microvascular endothelial cells (submitted manuscript). In line with the findings presented in this study, Ureaplasma spp. may thus specifically induce cell death in microvascular endothelial cells, facilitating their own invasion from the blood stream into the respective tissue. Once there, pathogens may then reciprocally inhibit apoptosis to prevent eradication and establish chronic infections with corresponding long-term sequelae.
In line with this theory, Ureaplasma spp. appear to suppress pyroptosis and potentially necroptosis in HBMEC. For several involved genes in both pathways, we could observe a downregulation of mRNA levels after 30 h of Ureaplasma exposure. This includes caspases 1 and 4, NLRP3, and gasdermin D for pyroptosis (Fig. 4) as well as the necroptotic kinase RIPK3 (Fig. 5b). Only for MLKL, the executioner of necroptosis, Ureaplasma exposure of HBMEC resulted in a distinct upregulation of mRNA expression (Fig. 5c). However, regardless of its mRNA amount, MLKL needs to be phosphorylated by RIPK3 to fulfill its role in necroptosis [48]. Other authors demonstrated that a lack of RIPK3 mRNA resulted in absent MLKL phosphorylation and inhibition of necroptosis [49]. We speculate that, by suppressing key participants in inflammatory cell death, Ureaplasma spp. may concurrently attenuate several important immune defense mechanisms. First of all, pyroptosis and necroptosis are considered effective mechanisms to eliminate intracellular pathogens [24,29,[50][51][52]. Furthermore, particularly caspase 4 has been shown to trigger not only pyroptosis, but also the production of the pro-inflammatory cytokine interleukin (IL)-1β [53,54], thus interlinking cell death and inflammation. Therefore, Ureaplasma-driven downregulation of caspase 4 may additionally suppress secretion of IL-1β and thus attenuate important pro-inflammatory cytokine cascades. The exact mechanisms of Ureaplasma-driven modulation of programmed cell death are yet to be determined. Evaluating virulence factors of Ureaplasma spp. in the context of well-known factors initiating or inhibiting apoptosis, pyroptosis, and necroptosis ( Fig. 1), it seems reasonable to speculate an involvement of TLR signaling, Ureaplasma-driven TNF-α protein release (recently demonstrated for human monocytes [17]), and potentially cell invasion and ammonia production by Ureaplasma spp.
Interestingly, LPS, a component of bacterial membranes used to mimic bacterial infections, induced effects widely differing from Ureaplasma-evoked responses. Following LPS exposure of HBMEC, we could observe an early upregulation of apoptotic cascade mRNA levels after only 4 h (Fig. 3), whereas Ureaplasma spp. generally evoked responses solely over longer stimulation periods. This fierce and immediate LPS effect is reflected in the fulminant clinical course of a typical bacterial meningitis, whereas the ramifications of an infection with Ureaplasma spp. may be confined in the beginning, but increase over time. Studies describe cases of chronic Ureaplasma meningitis with a history of several months [12,13]. Furthermore, unlike the downregulation we could observe upon Ureaplasma exposure, LPS enhanced caspase 1, caspase 4, NLRP3, and gasdermin D as well as RIPK1 mRNA expression in HBMEC (Figs. 4 and 5), thus apparently triggering pyroptosis and necroptosis. Nonetheless, LPS did not increase the overall number of dead HBMEC upon a 24-h stimulation period and only began to do so after 48 h (Fig. 2), with effects less pronounced than the ones evoked by Ureaplasma spp. It is therefore likely that LPS simultaneously enhances protective factors which confine cell death. Ureaplasma spp., on the other hand, appear to disturb the pro-and anti-apoptotic equilibrium and cause extensive cell damage. In line with this speculation, we were able to demonstrate an increase in mRNA and protein expression of anti-inflammatory IL-10 induced by LPS, but not by Ureaplasma isolates in primary human monocytes [17]. IL-10 has been shown protective against apoptosis in HBMEC [55].
Simultaneous strength and limitation of this study is the use of a non-immortalized cell line. Primary cell lines closely resemble in vivo cells, but also have a limited in vitro life span. This is reflected in a relevant number of dead control cells particularly after 48 h (96 h after seeding) and in a decrease in transendothelial resistance in control cells over time. Aggravating factors may be lack of substrate and lack of space after confluency.
This study provides a thorough assessment of complex cascades involved in different forms of programmed cell death, correlating the basic cellular processes with actual ramifications on cell viability and, by monitoring barrier properties, even physiological functioning. We did, however, concentrate primarily on mRNA levels and can provide protein or enzyme activity data only for a few of many participants. Under in vivo conditions, activation processes and complex interactions are involved in the regulation of caspases and other enzymes [30] and in vitro settings cannot be fully representative. Further studies functionally analyzing the individual cascades, assessing associated signaling pathways, and including additional apoptosis assays are therefore essential to improve our understanding of Ureaplasma-driven cell death.

Conclusions
Due to their high prevalence, difficult cultural detection, and contradictory clinical data, the relevance of Ureaplasma spp. is still subject of controversial discussion. Having recently demonstrated a pro-inflammatory capacity of Ureaplasma spp. in HBMEC, the present study reveals that the injurious effects of Ureaplasma spp. go far beyond an ability to "merely" cause inflammation. Our in vitro findings suggest Ureaplasma-driven apoptosis in HBMEC, which, in vivo, may ultimately cause BBB breakdown with resulting CNS inflammation. A potential additional suppression of inflammatory forms of cell death by Ureaplasma spp. may further impair host immune defense mechanisms and ultimately facilitate long-term cerebral infections. We appear to be confronted with a meticulously acting pathogen of still underestimated clinical relevance.

Additional file
Additional file 1: Gating strategy for analysis of caspase flow cytometry results. Events were gated via forward and side scatter, and a SSC-height versus FSC-width dot plot was used to exclude doublets. Single cells were then gated for viability dye negative, caspase positive cells and numbers were depicted in the respective figures (Fig. 3c, f, i). CASP: caspase. (TIF 2881 kb)