Reduced internalization of TNF-ɑ/TNFR1 down-regulates caspase dependent phagocytosis induced cell death (PICD) in neonatal monocytes

Phagocytosis-induced cell death (PICD) is diminished in cord blood monocytes (CBMO) as compared to cells from adults (PBMO) due to differences in the CD95-pathway. This may support a prolonged pro-inflammatory response with sequels of sustained inflammation as seen in neonatal sepsis. Here we hypothesized that TNF-α mediated induction of apoptosis is impaired in CBMO due to differences in the TNFR1-dependent internalization. Monocytes were infected with Escherichia coli-GFP (E. coli-GFP). Monocyte phenotype, phagocytic activity, induction of apoptosis, and TNF-α/TNF-receptor (TNFR) -expression were analysed. In the course of infection TNF-α-secretion of CBMO was reduced to 40% as compared to PBMO (p<0.05). Neutralization of TNF-α by an αTNF-α antibody reduced apoptotic PICD in PBMO four-fold (p < 0.05 vs. infection with E. coli). PICD in CBMO was reduced 5-fold compared to PBMO and showed less responsiveness to αTNF-α antibody. CBMO expressed less pro-apoptotic TNFR1, which, after administration of TNF-α or infection with E. coli was internalized to a lesser extent. With similar phagocytic capacity, reduced TNFR1 internalization in CBMO was accompanied by lower activation of caspase-8 (p < 0.05 vs. PBMO). Stronger caspase-8 activation in PBMO caused more activation of effector caspase-3 and apoptosis (all p < 0.05 vs. PBMO). Our results demonstrate that TNFR1 internalization is critical in mediating PICD in monocytes after infection with E.coli and is reduced in CBMO.


Introduction
Pre-and postnatal infections are triggers for a variety of diseases later in life, involving different organ systems. The common link between infection and its secondary sequels, organ damage, evidently is inflammation, and typical diseases of the preterm infant, such as bronchopulmonary dysplasia (BPD) [1], necrotising enterocolitis (NEC) [2], retinopathy of prematurity (ROP) [3], periventricular leucomalacia (PVL) [4], and others have been closely tied to pre-or postnatal infection [5]. Lately, in neuropsychiatric diseases, where one would not necessarily consider a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Mononuclear cell cultures. Peripheral blood mononuclear cells from adults and cord blood mononuclear cells (PBMC and CBMC) were isolated by density gradient centrifugation on Ficoll cushions (Amersham, Chalfont St. Giles, UK) as described previously [14]. Washed cells were resuspended in VLE RPMI-1640 (Biochrom, Germany). For analysis of post-phagocytic reactions, cells were counted in an ultraplane Neubauer hemocytometer, placed at 2x10 6 cells/ml in flat bottom 24 well cell culture plates (Greiner, Solingen, Germany), containing 10% heat-inactivated fetal calf serum (FCS, Biochrom, Germany) and gentamycin (Sigma, Munich, Germany, 200μg/ml), and were incubated at 37˚C.
To compare post-phagocytic reactions in whole blood and leucocytes, whole blood (50 μl) was stained with anti-CD14 antibodies for 15 min on ice. Via FACS analysis the percentage of monocytes was determined and the appropriate volume of whole blood subjected to the invitro-infection-assay.
TNFR1 internalization assays utilizing anti-TNFR antibodies were performed as described before [10]. Biotin-labelled TNF-α for monitoring internalization was from eBiosciences (Frankfurt, Germany). In brief, 1x106 mononuclear cells were resuspended in PBS and Biotinlabelled TNF-α added in a final concentration as recommended by the manufacturer. Mononuclear cells were kept for the indicated time interval under standard culture conditions, before streptavidin-FITC was administered. After washing in PBS specimen were subjected to FACS analysis.

Apoptosis assays
Annexin-V assay. Annexin V was provided by Immunotools (Friesoythe, Germany). 2μl of annexin V solution was added to 10 6 cells in CaCl 2 supplemented PBS (2.5 mmol final concentration) for 20 min on ice in the dark as described by the manufacturer.
TUNEL assay. Cells were stained with CD14 mAb for 15 minutes at RT before fixation in paraformaldehyde (2% v/v in PBS). Subsequent steps of the TUNEL assay were performed according to the manufacturer's recommendations (Roche, Mannheim, Germany). Fixed, permeabilized and DNAse I treated mononuclear cells served as positive controls.
Detection of hypodiploid nuclei. DNA fragmentation was assessed according to Nicoletti and previously described [10]. In brief, washed cells were slowly resuspended in 2ml of -20˚C ethanol 70% with continuous vortexing and stored for four hours at -20˚C. Cells were washed twice, resuspended in 50μl PBS containing 13 units RNAse (DNAse free; Sigma, Germany) and incubated for 15 minutes at 37˚C. 180μl of PI (70μg/ml) was added, incubated for 20 minutes and analysis was performed immediately. Alternatively, mononuclear cells were stained with CD14 antibody for 15 minutes at RT to identify monocytes. A fixation with paraformaldehyde (2% v/v in PBS) for 2 hrs at RT replaced the ethanol fixation. Afterwards, cells were permeabilized by incubation in PBS-T (PBS, Triton X-100 0,1% v/v) for 20 minutes at RT, washed twice in PBS, resuspended in PBS-PI (PBS, 70μg/ml PI and 13 units RNAse) and incubated for 10 minutes at RT before analysis by flow cytometry. Cell-doublets were discriminated by assessment of PI-width/PI-area as described [15]. For further dissection of apoptotic and non-apoptotic DNA degradation, living cells were incubated with Vybrant dye (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's recommendations. FACS analysis allows detection of sub-G1 populations representing apoptotic cells. Cells in G1 phase which were also positive for PI were designated as cells undergoing non-apoptotic cell death. The latter method was used to determine TNF-α concentrations inducing apoptotic cell death only. We found that TNF-α concentration below 5 ng/ml did not induce apoptosis. A TNF-α concentration higher than 5 ng/ml induced also non-apoptotic cell death and was not suitable for this study.
ELISA. The TNF-α enzyme-linked immunosorbent assay (ELISA) was purchased from Ebiosciences (Ebiosciences-Natutec, Frankfurt, Germany) and used according to the manufacturer's recommendations. The read-out was executed in a spectra max 340PC ELISA reader (molecular devices, Sunnyvale, CA, USA) with a sensitivity from 4-500pg/ml.
Immunoblot. For the immunoblot analysis 6x10 6 cells were washed twice with PBS and lysed in Laemmli-lysis buffer. After heating to 96 C o for 5 min contaminating DNA was destroyed by pipetting samples through a syringe (gauge 23) before they were subjected to gel electrophoresis (SDS-PAGE). SDS-PAGE was performed according to standard protocols. Antibodies were diluted according to the manufacturers'recommendations. For imaging a LAS 3000 (Fujifilm, Düsseldorf, Germany) was used. Signal strength was quantified with the multi-gauge software (Fujifilm, Düsseldorf, Germany).

Statistical analysis
Results are expressed as mean +/-standard deviation. Error bars represent standard deviations. Values of p < 0.05 were considered significant. Analyses were done with statistical software (performing student's t-test, one-way and two-way ANOVA adjusted according to Bonferroni-Holm for multiple group comparisons as provided by GraphPad Software Statistical Package, La Jolla, CA,USA). Data distribution was assessed by a Shapiro-Wilk normality test.

E.coli induces reduced TNF-α secretion in CBMO which is functional in PICD
We infected monocytes with E.coli-GFP for four hours and analysed their phagocytic activity. As shown before [16], phagocytic indices did not differ between PBMO and CBMO (Panel A in S1 Fig Infection with E.coli for four hours raised the concentration of soluble TNF-α up to 2 ng/ml in supernatants of infected PBMO. Soluble TNF-α from infected CBMO was approximately 2.5-fold lower (Fig 1A, p < 0.05). The TNF-α secretion in infected whole blood samples was comparable to that found after preparation of PBMO and CBMO (Panel B in S1 Fig).
To test if TNF-α caused apoptosis in CBMO and PBMO we added TNF-α into the cultures. We observed a significant induction of apoptosis in both groups. However, the apoptosis rate of PBMO was 3.7 times higher than CBMO (Fig 1B). Whereas the combination of TNF-α and infection with E.coli did not increase apoptotic rates in PBMO, it in contrast significantly enhanced the CBMO apoptosis rates, compared to infection alone (p < 0.05, Fig 1B).
Employing the E.coli-GFP in-vitro infection model, we assessed the induction of PICD by three independent methods and observed reduced apoptotic rates in CBMO (Fig 1B and 1C and Table 1). PICD-induced apoptosis was stronger in PBMO compared to CBMO.
Monocytes were infected with E.coli-GFP for 4 hours and free bacteria were removed. Apoptosis was detected by annexin V to detect phosphatidylserine in the outer leaflet of the plasmamembrane (n = 7), which is an initial phase of apoptosis. Hypodiploid nuclei were detected by utilizing the method of Nicoletti et al. (see material and methods)n = 6).DNA strand breaks, which occur in the late phase of apoptosis were detected by the TUNEL assay ((n = 12), all assays, ÃÃ p < 0.01 non-infected vs. infected (# p < 0.005 ANOVA PBMO vs. CBMO)).
Cells were infected with E. coli-GFP for 4 hours with an MOI of 25.
To shed light on the question, whether PICD is caused predominantly by CD95L/CD95 or by TNF-α/TNFR1 signalling, we added αTNF-α and CD95L neutralizing antibody ZB4 in two concentrations to infected monocytes ( Fig 1C). Addition of the lower concentration of either αTNF-α or ZB4 reduced PICD to 70% ( The results indicate that induction of PICD is not restricted to CD95L/CD95 signalling. This effect could not be observed for CBMO, were ZB4 failed to reduce PICD ( Fig 1C, 2 nd vs. 3 rd column). The αTNF-α antibody in turn could also block the already reduced PICD in infected CBMO.

Fig 1. E. coli induces TNF-α secretion which is involved in induction of apoptosis in PBMO and CBMO.
Assessment of secreted TNF-α 4 h p.i. from supernatants of PBMO and CBMO (A, n = 7; ***p < 0.005, clamped bars, student's t-test; ***p < 0.005, blunt-ended bars, two-way ANOVA). Induction of apoptosis 4 h p.i. assessed by Nicoletti assay in CD14+ monocytes after indicated treatment (B, n = 8; *p < 0.05, clamped bars, student's t-test; blunt-ended bars, two-way ANOVA). PICD assay as shown in (B), but with addition of blocking antibodies ZB4 and αTNF-α (final concentration given below the chart) as indicated (C, n = 3; *p < 0.05, clamped bars, student's t-test; blunt-ended bars, two-way ANOVA). As shown for phagocytic indices and TNF-α secretion, apoptosis was induced in adult and neonatal whole blood to comparable extent to that observed for PBMO and CBMO (Fig 1 and Panel C in S1 Fig).

Diminished TNFR1 expression and -internalization after infection leads to reduced apoptosis in CBMO
In a recent publication we demonstrated the effect of TNFR1 internalization with regard to pro-apoptotic signalling [10]. Therefore, it seemed reasonable to compare this process in PBMO and CBMO.
Tracking of TNFR1 internalization after addition of TNF-α showed a rapid uptake of TNFα /TNFR1 in PBMO, but a delay in CBMO. Over a period of 240 min, CBMO exhibited significant TNFR1 internalization only at 120 min p.i. which dropped again at 240 min p.i. (Fig 2A). We further tracked biotinylated TNF-α on monocytes and, in contrast to PBMO, observed TNF-α retention on the plasma-membrane of CBMO ( Fig 2B).
The PBMO that internalized TNFR1 after addition of TNF-α showed a significant induction of apoptosis from 60 min to 120 min p.i. with a peak 120 min p.i. (Fig 2C). In contrast, the percentage of TNFR1 internalizing CBMO undergoing apoptosis was diminished compared to PBMO ( Fig 2C).
Additionally, we determined induction of apoptosis in PBMO and CBMO which did not internalize TNFR1. No significant amounts of apoptotic monocytes could be detected (S2 Fig). Altering the TNFR1 surface expression can pilot PICD in monocytes As shown in Fig 1, phagocytosis of E.coli induced a stronger pro-apoptotic potential compared to TNF-α stimulation. Reduction of surface TNFR1 expression could reflect internalization or Assessment of internalized TNFR1 in PBMO and CBMO after indicated intervals (A, n = 8; *p < 0.05, ***p < 0.005, clamped bars, student's t-test; ***p < 0.005, blunt-ended bars, two-way ANOVA). Detection of TNF-α on the plasma-membrane of monocytes (B, n = 3; p < 0.05; clamped bars, student's t-test; blunt-ended bars, twoway ANOVA). Apoptosis induced by addition of TNF-α was measured in monocytes exhibiting internalized TNFR1 (C, n = 7, *p < 0.05, **p < 0.01, ***p < 0.005, clamped bars, student's t-test; blunt-ended bars, two-way ANOVA). shedding of the receptor. We addressed the question whether cycling of TNFR1 from the surface to the cytoplasm during E.coli infection has an influence on the induction of PICD.
In order to prevent the TNFR1 from being shed from the surface [17], we administered the metalloprotease inhibitor GM6001 to infected and non-infected monocytes. The infectioninduced down-regulation of TNFR1 on PBMO could be entirely blocked by GM6001; on CBMO we found higher densities than on infected monocytes only (Fig 3A, 3 rd columns).
Due to the fact that GM6001 could block shedding of TNFR1 and/or shedding of its ligand TNF-α resulting in lower soluble TNF-α (Fig 3B) which is a prerequisite for internalization and induction of apoptosis [13], we added TNF-α into the cultures. To further evaluate the role of TNFR1 expression in pro-apoptotic signalling, we compared the induction of apoptosis in infected PBMO and CBMO in the absence or presence of GM6001 (Fig 3C). Addition of GM6001 to infected monocytes increased TNFR1 surface expression and led to lower apoptotic rates in PBMO but not in CBMO. Infected and GM6001-treated CBMO showed apoptosis rates similar to infected PBMO (Fig 3C).
After infection with E. coli some groups received an αTNF-α mAb, a pancaspase inhibitor (zVAD; Fig 4B), or the specific caspase-8 inhibitor (IETD; Fig 4C). The latter was given in order to demonstrate the dependency of caspase-3 induction on caspase-8 activation.
Infection with E. coli led to increased cleavage of caspase-8 in PBMO (p<0.01; Fig 4B), while this effect was strongly reduced in CBMO 120 min p.i. Caspase-8 activation was timelimited, because 240 min p.i., both, PBMO and CBMO did not show caspase-8 activation significant different to the non-infected controls (Fig 4B). Addition of αTNF-α mAb to infected monocytes significantly reduced activities of caspase-8 and caspase-3 in PBMO and CBMO (Fig 4B and 4C, 3 rd columns).
Additionally, we checked whether TRADD and RIP; proteins forming a signalling complex engaging caspase-8, were expressed differentially in PBMO and CBMO. Both TRADD and RIP were expressed at comparable levels in PBMO and CBMO either in non-treated and infected or TNF-α treated monocytes (Panels A and B in S3 Fig). However, in E.coli-infected PBMO RIP was processed to the smaller cleaved form RIPc (Fig 4C). The same was found for TNF-α administration. Significant, E.coli or TNF-α induced RIPc formation could not be found in CBMO and was diminished in PBMO, treated with caspase-8 inhibitor IETD (Fig 4D).
Next, we addressed the question whether we could inhibit PICD by blocking TNF-α-mediated caspase-8 signalling in monocytes. Therefore we added the caspase-8 inhibitor IETD to infected monocytes (Fig 5). Compared to the group of infected monocytes without IETD treatment, IETD blocked the induction of apoptosis in PBMO and CBMO, suggesting that in both, PBMO and CBMO caspase-8 activation is a prerequisite for the induction of apoptosis after infection (Fig 5, 2 nd vs. 3 rd column; p < 0.05).
Reduced levels of TNF-α in neonatal monocytes after infection or challenge with bacteria ( Fig 1A) or bacterial PAMPs (such as LPS) are a matter of debate with most studies indicating a bias towards reduced TH1 cytokine production in CBMO. In response to most TLR ligands, most reports found neonatal innate immune cells, including monocytes, conventional and plasmacytoid dendritic cells producing less IL-12p70 and IFN-ɣ, moderately less TNF-α but as much or even more IL-1ß, IL-6, IL-23, and IL-10 than adult cells [18].In concurrence with our study, reduced extracellular TNF-α levels in E. coli-infected purified monocytes have been described [19] [20][21][22]. Dampening inflammation is thought to be essential for the fetal immune system to avoid rejection by the mother and termination of pregnancy [23]. Therefore reduced secretion of TNF-α could be beneficial for the fetal environment, albeit bearing the risk of reduced capacity to fight of infection and of prolonged activation of immune effector cells by due to missing pro-apoptotic signaling. In this regard, the isolation of PBMO and CBMO via depletion of granulocytes and removal of serum factors could be limiting factors of our model. However, comparison of our published results to the whole blood assays (S1 Fig), showed that basic results, such as phagocytic activity, TNF-α secretion and induction of apoptosis were similar in both experimental setups.
We found that the susceptibility towards PICD of infected CBMO increased, when exogenous TNF-α was added (Fig 1B, 2 nd vs. 4 th columns). These data, together with the observation that PICD of PBMO, but not CBMO could be greatly diminished by blocking TNF-α (Fig 1C   Fig 5. Inhibition of caspase-8 leads to reduction of cleaved caspase-3 and mounts for diminished PICD. Detection of cleaved caspase-3 4 h p.i. Groups which received the α-TNF-α mAb,andIETD were indicated (A; n = 10, *p < 0.05, ***p < 0.005, clamped bars, student's t-test; blunt-ended bars, two-way ANOVA). Detection of hypodiploid nuclei after indicated treatment and administration of caspase inhibitors 4 h p.i. (B; n = 10, *p < 0.05, **p < 0.01 ***p < 0.005, clamped bars, student's t-test; blunt-ended bars, two-way ANOVA). https://doi.org/10.1371/journal.pone.0182415.g005 and Table 1) pointed towards a crucial role of the TNF-α pathway in these processes. Addition of exogenous TNF-α led to rapid internalization of TNFR1 in PBMO but not CBMO (Fig 2A). Since apoptosis mainly occurred in TNFR-1-internalizing monocytes, the difference in TNFR-1 internalization could account for the difference in apoptosis observed between PMBO and CBMO ( Fig 2C). This is in line with the observation that CBMO expressed lower densities of TNFR1 on the surface of infected cells (Fig 3A, 1 st vs. 2 nd column) and that TNF-α remained on the surface of CBMO (Fig 2B). Downstream signalling of internalized TNF-α is regulated by the recruitment of adaptor proteins like TRADD and RIP forming two competing protein complexes driving either apoptosis or pro-inflammatory NFkappaB activation [13,[24][25][26]. TRADD can recruit caspase-8 which cleaves RIP1 [27]. This truncated RIPc contains the death domain enabling the binding of FADD. In turn, the anti-apoptotic activation of the NFkappaB pathway is reduced. Since blockage of caspase-8 by IETD (Fig 5) abrogates apoptosis it is feasible to argue that reduced TNFR1 internalization in CBMO causes reduced apoptosis. Furthermore, we observed cleavage of RIP in infected or TNF-α treated PBMO (Fig 4B and 4C). This process was caspase-8 dependent and could not be observed in CBMO (Fig 4B and 4C). The biological significance of the rather modest effect of TNF-α on E.coli-induced PICD in cord blood (CBMO) remains to be elucidated; although the administration of anti-TNF-α antibody inhibited PICD of CBMO significantly (Fig 1C).
Like TNF-α, CD95L of which reduced production in CBMO was recently published to cause diminished apoptosis [9], recruits an adaptor protein FADD [24]. CD95L activated FADD can directly engage caspase-8 and, unlike internalized TNF-α/TNFR1, does not require further signals. This is in line with our findings that supplementation with bioactive CD95L increased the level of endogenous CD95L in CBMO and enhances apoptosis to levels observed in PBMO [9]. CBMO which are challenged by both E.coli and TNF-α died more frequently compared to infected only or TNF-α treated only CBMO (Fig 1B). This minor but significant effect, which is lacking in PBMO cannot be attributed to higher expression of RIP and TRADD in CBMO per se, because these proteins were shown to be equally expressed (Panels A and B in S3 Fig). Therefore, processing of RIP and activation of caspase-8 could be a consequence of TNFR1 internalization, which is reduced in CBMO. RIP However, some methodical limitations cannot rule out a coincidence of the observed processes. Future experiments should be conducted to purify internalized TNFR1 signalosomes which can be tested for caspase and RIP activation. As pointed out before TNF-α/TNFR1 ligation is potentially an inducer of an apoptosis modulating signal cascade. TNF-α secretion could induce CD95L expression suggesting that CD95L/CD95 signalling is a predominant pro-apoptotic pathway without requirement of TNFR1 internalization. Evidence for phagocytosis-triggered activation of the pro-apoptotic cascade involving caspases -8 and -3, was described for Staphylococcus aureus [28]. Phagocytosis also regulates the expression of small Bcl-2 like proteins [23]. For the BCL-2 like protein t-Bid a connecting function to the extrinsic and intrinsic apoptosis cascade could be shown [29], defining caspase-8 as a hinge in an auto-activating loop accelerating apoptosis.
In this regard, numerous bacterial pathogens, among them E.coli, Salmonella and Shigella produce and secrete effector proteins which impair the signalling machinery of TNF-family ligands and their receptors by inhibition of downstream mediators such as FADD, TRADD and caspases [24,30].
In previous publications we found caspases-8 and -9 activated after infection with GFP-E. coli [31]. Caspase-8 and -9 are functional in the intrinsic-and extrinsic apoptosis pathway and in turn activate effector caspases, e.g. caspase-3.
In this manuscript, we also addressed the question, whether shedding of TNF-α and its receptor TNFR1 interferes with pro-apoptotic signalling. Blockage with the metalloproteaseinhibitor GM6001 (Fig 3) resulted in increased densities of TNFR1 on infected monocytes ( Fig  3A 2 nd vs. 3 rd columns), and decreased the number of infected apoptotic PBMO (Fig 3C, 2 nd vs. 3 rd columns). Adding TNF-α into this scenario (Fig 3C, 4 th columns) had no synergistic effect on the PICD of PBMO, but led to a significant increase of PICD of CBMO to the extent as comparable to infected PBMO.
Besides TNF-α/TNFR1/TNFR2 and CD95L/CD95 other ligands/receptors may play a role in regulation of apoptosis in PBMO and CBMO. TRAIL receptors can be activated and also activate the extrinsic caspase signalling pathway. The TRAIL receptor expression was not found reduced in CBMO compared to maternal PBMO [32], but expression of TRAIL ligand has not been examined in detail yet. To date, the model of sustained inflammation is favored to explain fatal outcome of neonatal sepsis. However, our data emphasizes the possibility that the reduced capacity of the neonatal immune system could contribute to pathology. These monocytes could have harmful effects on cells in the surrounding tissue, by inducing further apoptosis and tissue damage as it was shown for the lung [33]. To what extent this may influence the course of neonatal sepsis, remains to be elucidated in vivo.
Supporting information S1 Fig. Periphagocytic reactions are similar in ex-vivo and in-vitro assays. Whole blood cells of adults (white bars) and neonates (grey bars) and MNC were infected with E.coli (MOI 25) and the PI (A, calculated as CD14+GFP+ monocytes: CD14+ monocytes, see the material and method section) and the TNF-α secretion (B) measured. Induction of apoptosis in noninfected and infected whole blood of adults (white hatched bars) and neonatal (grey hatched bars) samples was also determined (n = 3, ÃÃ p < 0.01 ÃÃÃ p < 0.005, clamped bars, student's ttest; blunt-ended bars, two-way ANOVA).