Helicobacter pylori Cytotoxin-Associated Gene A Impairs Human Dendritic Cell Maturation and Function through IL-10–Mediated Activation of STAT3

Helicobacter pylori colonizes the gastric mucosa, where it induces a chronic inflammatory response, which is associated with peptic ulcer disease and an increased risk of gastric cancer. This inflammatory response toward H. pylori is characterized by the recruitment of different immune cells, mainly dendritic cells (DCs), neutrophils, macrophages, as well as B and T lymphocytes (1) to the site of infection. In fact, human DCs can enter H. pylori–infected gastric epithelium to take up bacteria and their virulence factors (2). DCs not only play a crucial role as primary responders to microbial infection because of their ability of capturing and transferring Ags, but they also determine the type of T cell–mediated response to be mounted (3). Moreover, in the gastrointestinal tract, CD11c⁺CD8α⁺ and CD103⁺ DCs have been shown to promote regulatory T cell (Treg) responses and mediate tolerance. Several lines of evidence suggest that H. pylori may induce a response by Tregs, leading to H. pylori persistence. In this regard, the expression levels of the Treg marker Foxp3 have been described to be higher in the gastric epithelium of H. pylori–infected subjects (4–8). In addition, chronic exposure of murine DCs to H. pylori has been reported to have a detrimental effect on the ability of DCs to induce a Th1 response (9). Moreover, H. pylori has been shown to alter the DC-polarized Th17/Treg balance toward a Treg-biased response, resulting in a suboptimal Th17 response and a subsequent failure to eradicate the pathogen (10). Furthermore, mice neonatally infected with H. pylori fail to mount local and systemic immune responses toward H. pylori infection and present higher H. pylori colonization of the gastric mucosa. In these mice, immunologic tolerance toward H. pylori was developed in the neonatal period because of a biased Treg/T effector cell ratio (11). However, the molecular mechanisms by which H. pylori infection induces tolerogenic DCs and a subsequent Treg response remain unclear.

Although differentiation of DCs into immunogenic or tolerogenic DCs has not been fully characterized, it is generally accepted that DC maturation in addition to DC lineage determine the functionality of DCs (12). Different signaling cascades are involved in DC maturation, predominantly the NF-ĸB pathway. In this context, RelB has been reported to be a critical factor involved in DC maturation, and tolerogenic DCs have been generated by RelB silencing using short hairpin RNA (13). Other signaling networks involved in programming DCs into a tolerogenic state include Wnt-β-catenin and Jak/STAT3. In intestinal DCs, β-catenin was described to be required for Treg induction and suppression of inflammatory effector T cells (14), whereas other studies have demonstrated an important role of STAT3 in DC differentiation and function under physiological conditions and cancer. For instance, hyperactivation of STAT3 was shown to lead to an abnormal differentiation of DCs in cancer (15), whereas inhibition of Jak2/STAT3 or STAT3 silencing induced a dramatic improvement of DC differentiation (16, 17). Moreover, in different mouse models, including the IL-6 knockout and gp130 knockin mice, the IL-6/STAT3 signaling pathway was shown to regulate DC differentiation in vivo (18). Furthermore, DC-specific STAT3 deletion in mice was associated with impaired mucosal tolerance and revealed STAT3 as a negative regulator of DC function (19). Cytokines and growth factors as IL-10 induce phosphorylation of STAT3, which is followed by its dimerization and translocation into the nucleus, where it modulates the expression of target genes. Although H. pylori has been described to induce phosphorylation of STAT3 in gastric cancer cells (20), no data on STAT3 activation upon H. pylori uptake in DCs have been reported to date.

In the current study, we demonstrate that H. pylori leads to semimaturation of human DCs by activation of the transcription factor STAT3. These semimature DCs have a tolerogenic phenotype and induce Treg expansion that might contribute to H. pylori persistence.

DCs were generated from fresh blood samples taken from healthy volunteers for the purpose of this study as well as for serological analysis. The study was approved by the ethics committee of the Technische Universität München. All volunteers gave written informed consent.

The H. pylori strains G27 (21), PMSS1 and SS1 (22), J99 (ATCC number 700824), P1 (23), and 60190 (ATCC number 49503) were grown on Wilkins–Chalgren blood agar plates under microaerobic conditions (10% CO₂, 5% O₂, 85% N₂, and 37°C).

G27 isogenic mutant strains G27ΔCagA, G27ΔBabA, G27ΔVacA, G27ΔgGT, and G27ΔUreA/B were grown on Wilkins–Chalgren blood agar plates containing 50 μg/ml kanamycin for selection under the same conditions.

Escherichia coli strains K12 (provided by Prof. T. Miethke, Institute of Medical Microbiology, Immunology, and Hygiene, Munich, Germany) were cultured in Luria–Bertani medium at 37°C.

Bacteria was fixed in 4% paraformaldehyde (PFA) for 4 h at 4°C and washed several times with cold PBS before addition to the cells.

PBMCs were isolated from H. pylori–negative healthy donors, after informed consent, by density gradient centrifugation with Biocoll (Biochrom).

Monocytes were isolated from PBMCs by magnetic cell labeling (MACS) with the Monocyte Isolation Kit II (Miltenyi Biotec). Their purity was determined by flow cytometry staining (anti-CD14 and anti-CD45).

Immature DCs were generated by culturing monocytes in RPMI 1640 medium with glutamine (Invitrogen), 10% heat-inactivated FCS GOLD (PAA) and 1% penicillin/streptomycin (Invitrogen), 20 ng/ml human rIL-4 (Miltenyi Biotec), and 20 ng/ml human rGM-CSF (Miltenyi Biotec) for 6 d. Afterwards, cells were harvested and analyzed by FACS for CD14 and CD11c expression (Supplemental Fig. 1A).

Immature DCs were coincubated with H. pylori at multiplicity of infection (MOI) 5 for 24 h.

For dose dependency assays, cells were infected at MOI 2, 5, 10, 20, and 50 for 24 h, whereas for time-course experiments, the MOI was 5 and DCs were coincubated up to 72 h.

For IL-6 and IL-10 neutralization experiments, DCs were incubated with 5 μg/ml anti-human IL-6 (R&D Systems) or 5 μg/ml anti-human IL-10 (R&D Systems) neutralizing Abs for 1 h before H. pylori infection. To control the neutralization efficiency, recombinant cytokines were used for DCs stimulation in parallel. Therefore, immature DCs were incubated with 90 ng/ml recombinant human IL-6 (Miltenyi Biotec) or 10 ng/ml recombinant human IL-10 (eBioscience) plus the neutralizing Abs as described previously.

STAT3 activation was blocked by incubating the cells with 0.5 μM of the specific inhibitor Stattic (Calbiochem) for 1 h before H. pylori infection.

Cells were stained with ethidium monoazide bromide (Anaspec) for 30 min on ice for live/dead discrimination, and then resuspended in cold FACS buffer (PBS-1% BSA). Fluorescence-labeled Abs recognizing CD14, CD11c, HLA-DR, CD80, CD83, CD86, CD4, CD25, and Foxp3 (eBioscience) were added following manufacturer’s instructions for 30 min at 4°C in the dark. After incubation, cells were washed and resuspended in FACS buffer before analysis or staining of intracellular Foxp3 expression. Intracellular staining was performed by incubating the cells with Foxp3 Ab diluted in permeabilization buffer (eBioscience), according to the manufacturer’s protocol. After 30 min incubation at room temperature cells were washed, resuspended in FACS buffer and analyzed. Analysis was performed with the FACS CyAn (Beckman Coulter) and the FlowJo software.

For determination of cytokine secretions, culture supernatants were harvested, and cytokine concentrations were assessed by sandwich ELISA according to the manufacturer’s instructions (eBioscience).

Cells were harvested at different times post infection as indicated, and lysed in SDS sample buffer (62.5 mM Tris-HCl [pH 6.8], 2% w/v SDS, 10% glycerol, 50 mM DTT, and 0.01% w/v bromophenol blue). Equal amounts of lysate were loaded on 8% SDS-PAGE gels and separated proteins were transferred to nitrocellulose membrane (Protran). After blocking, anti–p-IkBα, anti–β-catenin, anti-STAT3, anti–p-STAT3 (Cell Signaling Technology), or anti–β-actin (Sigma-Aldrich) Abs were used to detect protein expression according to manufacturer’s instructions. To detect CagA translocation, a rabbit anti-CagA serum, provided by Dr. R. Vogelmann (University Hospital Mannheim II, Mannheim, Germany), and a p-Tyr Ab (Millipore) were used. Quantification of band density was performed using Quantity One software (Bio-Rad).

CD4⁺ T cells were separated from PBMCs by using the CD4⁺ T cell Isolation Kit II (Miltenyi Biotec). Their purity was determined by flow cytometry (anti-CD45 and anti-CD4).

Immature DCs were infected with H. pylori at MOI 5. After 24 h, freshly isolated naive CD4⁺ T cells were added to the infected DCs at 2:1 ratio, as described previously (24). DCs and CD4⁺ T cells were cocultured for 3 d at 37°C in a humidified incubator containing 5% CO₂. Foxp3 expression was analyzed by FACS, and the released cytokines were measured by ELISA.

ANOVA test with Bonferroni’s correction for multiple comparisons was performed. For single comparisons, two tailed t test was used.

Maturation of DCs is characterized by high surface expression levels of costimulatory molecules such as CD80 and CD86, the maturation marker CD83, and the MHC class II receptor HLA-DR. Thus, the effect of the H. pylori strain G27 on human DC maturation was assessed by analyzing the expression of these molecules 24 h postinfection of immature DCs (Supplemental Fig. 1A, 1B) by flow cytometry. Induction of CD80, CD86, and CD83 expression was observed in H. pylori–infected cells, whereas lower levels of MHC-II were detected. However, the levels observed for all markers were significantly lower compared with those induced by PFA-fixed H. pylori (Fig. 1A) or E. coli (Supplemental Fig. 1C), suggesting that only live H. pylori leads to incomplete maturation of DCs. Moreover, this semimaturation was not time or dose dependent, because no significant differences in the expression of CD86 and CD83 were observed after infecting the cells for longer times (data not shown) or at higher MOI; however, it should be noted that loss of cell viability was detected when increasing the MOI (Supplemental Fig. 1D). In addition, all H. pylori strains tested induced DC semimaturation, indicating a strain-independent effect (Supplemental Fig. 1E).

To determine whether live H. pylori was actively dampening DC maturation, cells were coincubated with live and PFA-fixed H. pylori at the same time. Lower DC maturation was detected when compared with dead bacteria alone (Fig. 1B), suggesting that bacterial virulence factors actively induce semimaturation of DCs, thereby preventing full DC activation triggered by PFA-fixed H. pylori. Comparable results were obtained when live H. pylori was added in combination to PFA-fixed E. coli (Supplemental Fig. 1F). Thus, we next investigated which H. pylori virulence determinants might be involved.

DCs were infected with H. pylori strains deficient for either of the virulence factors CagA, BabA, UreA/B, γ-glutamyl transpeptidase (gGT), or VacA, and the expression of CD86 and CD83 was again analyzed after 24 h infection. The amount of CD86/CD83 double-positive cells was significantly higher when DCs were infected with CagA-deficient bacteria, whereas deletion of the other factors did not influence the maturation of the DCs (Fig. 1C). Comparable results were obtained when cells were coincubated with SS1 strain (Supplemental Fig. 1G), which—although being CagA positive—has a nonfunctional CagY (25) and does not translocate CagA into host cells (Supplementary Fig. 1H), indicating that CagA translocation is mainly responsible for H. pylori–induced DC semimaturation.

Taken together, these results indicate that the translocation of CagA plays a crucial role in inducing semimaturation of DCs during H. pylori infection.

To depict the H. pylori–induced cytokine secretion by DCs which, in turn, will influence the subsequent immune response, immature DCs were incubated with H. pylori for 24 h. The levels of the proinflammatory cytokine IL-12p70, necessary for Th1 differentiation, significantly increased upon H. pylori stimulation. However, the levels observed were lower than the ones detected when cells were incubated with PFA-fixed bacteria (Fig. 2). Similarly, IL-6, which promotes Th17 polarization, was secreted at higher levels by cells stimulated with dead bacteria compared with live H. pylori. Opposite results were obtained for IL-10, a cytokine driving Treg differentiation, because IL-10 levels were significantly higher in the supernatants from live H. pylori–infected DCs. It should be noted that high SDs observed reflect high variability among donors. Reduced cytokine secretion was detected when cells were infected at higher MOI (Supplemental Fig. 1D), which was due to increased cell death as observed by FACS analysis.

Because CagA was the main virulence factor inducing DC semimaturation, the influence of CagA on cytokine secretion also was evaluated. Interestingly, the absence of CagA reverted this tolerogenic phenotype. Higher levels of IL-12p70 were accompanied with decreased IL-10 secretion by DCs infected with the CagA deletion mutant, whereas no differences in the levels of IL-6 were detected. Similar results were observed when infecting the cells with the H. pylori strain SS1 deficient in CagA translocation (Supplemental Fig. 2) compared with the translocation-proficient PMSS1 strain. Moreover, a comparison of anti-inflammatory (IL-10) and proinflammatory (IL-12p70) cytokine ratios indicated that the presence of CagA favored an anti-inflammatory–biased cytokine secretion (Fig. 2).

These results suggest that CagA is the main H. pylori virulence factor responsible for inducing tolerogenic DCs.

To examine the functional relevance of DC semimaturation upon H. pylori infection, coculture experiments with allogeneic naive CD4⁺ T cells, which are the main T cell subset infiltrating the gastric mucosa of H. pylori–infected subjects and play an important role in immunity against H. pylori, were performed. Because CagA was proven to influence the maturation status of human DCs after H. pylori infection, we also examined the effect of CagA on the subsequent T cell response. Interestingly, we observed a significant inhibition of T cell proliferation, when CD4⁺ T cells were coincubated with H. pylori–primed DCs. This effect was independent of the presence of CagA (Supplemental Fig. 3) but induced by the virulence factor gGT, as reported previously (26). Next, we characterized T cell differentiation induced by H. pylori–primed DCs by measuring cytokine release in the supernatant of the cocultures. Specifically, the levels of IFN-γ, IL-17A, and IL-10, as markers for Th1, Th17, and Treg response, respectively, were analyzed. Increased levels of IFN-γ were observed when T cells were coincubated with CagA-deficient H. pylori-stimulated DCs (Fig. 3A) compared with wild-type H. pylori, whereas no differences were observed in the levels of IL-17A, as expected, because the absence of CagA highly influenced the secretion of IL-12p70 but had hardly any effect on the release of IL-6 by DCs. Interestingly, the absence of CagA induced lower levels of IL-10 produced by cocultured T cells. In addition, less Foxp3⁺ Tregs were observed (Fig. 3B), indicating that semimaturation of DCs induced by CagA skews the subsequent T cell response toward a Treg phenotype. Similar results to the observed for the CagA-deficient bacterium were obtained when T cells were coincubated with DCs primed with the SS1 strain (data not shown).

To decipher the molecular mechanism by which H. pylori was inducing DC semimaturation, β-catenin, NF-ĸB, and STAT3 signaling pathways, which have been previously reported to influence DC maturation, were examined. No changes in β-catenin or p-IĸBα were detected, whereas a clear activation of STAT3 was observed in H. pylori–stimulated DCs regardless of the H. pylori strain used (Fig. 4A).

The implication of the H. pylori virulence factors CagA, BabA, UreA/B, gGT, and VacA in STAT3 activation was further analyzed. When comparing STAT3 phosphorylation induced by the respective mutant strains, we did not observe any differences compared with the parental strain, except for the CagA mutant (Supplemental Fig. 4A). This CagA dependency was substantiated in two different genetic backgrounds, G27 and its respective ΔCagA mutant, and PMSS1 and SS1 (Fig. 4B). Significantly lower levels of phosphorylated STAT3 were observed in DCs infected with CagA-deficient bacteria. The lower levels of activated STAT3 observed after infecting DCs with H. pylori CagA–deficient strains suggested that CagA might be involved in STAT3 phosphorylation, as it has been previously described in gastric cancer epithelial cells (20). In addition, high levels of IL-6 and IL-10 were secreted by H. pylori–infected DCs (Fig. 1B), which could be mediating STAT3 activation.

To determine whether STAT3 activation was due to cytokines secreted, cells were incubated with supernatants obtained from control and H. pylori–infected DCs. STAT3 phosphorylation was indeed detected in cells incubated with supernatants from infected DCs (Fig. 4C), albeit at lower levels compared with direct infection, indicating that cytokines released upon infection contributed to STAT3 activation.

To further elucidate the specific effect of IL-6 and IL-10 on STAT3 activation, DCs were incubated with anti–IL-6 and/or anti–IL-10–neutralizing Abs prior to infection. Reduced levels of p-STAT3 were observed when IL-6 was blocked (Fig. 4D). However, this effect was higher when cells were incubated with the anti-IL-10 Ab or the combination of anti–IL-6 and anti–IL-10 Abs, indicating that IL-10 is mainly responsible of activating STAT3 in DCs after H. pylori infection. This is also supported by the observation that CagA-deficient strains induced significantly lower levels of IL-10 (Fig. 2A), which correlate with lower levels of p-STAT3.

To determine whether STAT3 activation in DCs upon H. pylori infection was responsible for DC semimaturation, we blocked STAT3 activation by incubating cells with anti–IL-6 and anti–IL-10–neutralizing Abs and/or with the specific inhibitor of STAT3 activation and dimerization Stattic (25). Higher levels of CD86/CD83 double-positive cells were only observed when DCs were incubated with anti–IL-10–neutralizing Ab (Fig. 5A), indicating that STAT3 activation is hampering DC maturation and confirming that IL-10 is the main cytokine mediating this effect. Unexpectedly, Stattic alone had almost no effect. This result could be explained by the observation that complete inhibition of STAT3 phosphorylation was only achieved when cells were incubated with Stattic in combination with the anti–IL-10–neutralizing Ab (Fig. 5B), supporting the fact that STAT3 is indeed mainly activated by IL-10 released by DCs after H. pylori infection. Interestingly, inhibition of STAT3 induced a significant increase in the release of IL-12p70 by DCs upon H. pylori infection (Fig. 5C), suggesting that activation of STAT3 negatively regulates IL-12p70 secretion. Similarly, no effect of Stattic alone was detected, but higher concentrations of the inhibitor-induced cell death (Supplemental Fig. 4B). Thus, the following experiments to prevent STAT3 signaling were performed by blocking IL-10 activity with neutralizing Abs.

When analyzing T cell differentiation under these conditions, we observed that restoring DC maturation by blocking IL-10–mediated activation of STAT3 induced the secretion of higher levels of IFN-γ from cocultured CD4⁺ T cells, whereas no effects were detected in the release of IL-17A (Fig. 5D). Furthermore, hindering STAT3 activation in DCs induced lower Treg expansion (Fig. 5E).

Taken together, these results indicate that STAT3 activation upon H. pylori infection confers a tolerogenic phenotype on DCs that favors a Treg response.

The adaptive immune response to H. pylori, characterized by recruitment of Th1 effector cells that secrete large amounts of IFN-γ (27, 28), is not efficient enough to control the infection, allowing lifelong persistence of H. pylori. Bacterial persistence can be enabled through the presence of high levels of Tregs infiltrating the gastric mucosa of infected subjects (4), which are critical for maintaining immunologic tolerance. DCs represent a major link between innate and adaptive immunity and play a crucial role in shaping the adaptive immune response to H. pylori. We show in this study that H. pylori induces a semimaturation of human DCs conferring a tolerogenic phenotype that elicits expansion of Tregs. Our results are in line with former studies in murine models showing impaired DC maturation and tolerogenic reprogramming of DCs by H. pylori (24, 29, 30); however, they provide a novel molecular mechanism involving the H. pylori virulence factor CagA and the activation of the transcription factor STAT3.

The role of H. pylori virulence factors in DC maturation and function has been extensively investigated in mice but revealed conflicting results. In this context, CagA has been shown to suppress DC maturation and function (31), whereas in a recent report VacA and gGT have been pointed out as the main H. pylori virulence determinants involved in murine DC tolerization in vitro and in vivo (24). In the current study, we identified CagA to induce a tolerogenic phenotype in human DCs by influencing cytokine release. Specifically, CagA-proficient H. pylori induced a predominant secretion of IL-10, whereas in the absence of CagA, lower levels of IL-10 were observed. IL-10 is fundamental to maintain DCs in their immature state, even in the presence of maturation stimuli, and at the same time contributes to Treg differentiation, whereas proliferation, cytokine secretion, and migratory capacities of effector T cells are reduced (32). Recently, H. pylori was found to induce IL-10 production via p38 MAPK and NF-ĸB activation downstream DC-SIGN, TLR2, and TLR4 signaling in human DCs (33), and our own unpublished data also indicate that TLR4 and to a lesser extent TLR2 are important for regulating IL-10 secretion by DCs. Furthermore, increased levels of IL-10 expression have been detected in CagA-positive H. pylori–infected subjects (34); however, it remains unclear by which molecular mechanism CagA is able to influence IL-10 secretion. Tanaka et al. (31) proposed a model in which phosphorylation of CagA in murine DCs would inhibit the MyD88-independent pathway via SHP-2 activation, thus reducing the production of IFNs and thereby suppressing host immune response. Nevertheless, further experiments need to be conducted to elucidate whether blocking of TRIF signaling by CagA is also responsible for the enhanced IL-10 secretion we observe in human DCs.

IL-10 is known to signal predominantly via STAT3 activation, and especially in DCs, STAT3 has been described to play an essential role in directing the anti-inflammatory activity of IL-10 (19). Indeed, we observed phosphorylation of STAT3 upon H. pylori infection, which was reduced in the absence of CagA correlating with lower levels of IL-10. Furthermore, the use of an IL-10–neutralizing Ab impaired STAT3 activation, indicating that IL-10 is the main cytokine responsible for STAT3 phosphorylation in DCs after H. pylori challenge. This observation was further supported by the fact that impairment of STAT3 activation by the specific STAT3 inhibitor Stattic was only achieved in few samples obtained from donors secreting low levels of IL-10 upon H. pylori challenge. Several reports have shown STAT3 to negatively regulate DC differentiation and function in vivo (19, 35). Thus, we hypothesize that STAT3 activation upon H. pylori infection is responsible for DC tolerization. This hypothesis is supported by data in Fig. 5 showing that blocking of STAT3 phosphorylation confers a less tolerogenic DC phenotype favoring a Th1 immune response.

Increased levels of IL-10 after incubating DCs with H. pylori were concomitant to low levels of IL-12 and depended on the presence of CagA. This observation is in agreement with previous reports describing reduced IL-12 secretion by H. pylori–primed DCs (29, 36, 37) and the involvement of CagA in the suppression of IL-12p40 expression (31); however, the possible signaling pathways implicated remain unclear. We speculate that activation of STAT3 might be also involved in preventing IL-12 release, because STAT3 was found to negatively regulate the IL-12p40 promoter (38), whereas in tumor-associated DCs, STAT3 inhibits NF-ĸB/c-Rel–dependent IL-12p35 expression (39). Indeed, we observed increased levels of IL-12p70 after blocking STAT3. Nevertheless, additional experiments are required to further characterize the H. pylori virulence determinants as well as the downstream signaling cascades engaged in IL-12 suppression during H. pylori infection.

Previous reports showed that DCs exposed to H. pylori fail to induce an effector T cell response of the Th17 type in vitro and in vivo (10, 40, 41). In line with these observations, we did not detect enhanced secretion of IL-17A from CD4⁺ T cells cocultured with H. pylori–primed DCs. IL-6 is one of the most important cytokines associated with the differentiation of human Th17. We detected low levels of IL-6, which were not changed in the absence of CagA, whereas dead bacteria caused a significant increase in the release of IL-6 that correlated with higher IL-17A secretion from T cells (data not shown), indicating that loss of bacterial viability is sufficient to deeply alter the cytokine secretion profile of DCs and the subsequent T cell response.

The presence of CagA has been linked to enhanced gastric inflammation (42, 43), which is attributed to its potent induction of innate immune responses in epithelial cells. However, our data suggest an important role for CagA in inducing tolerance. Remarkably, a decline in the ratio of Th1/Th2-derived cytokines was found from asymptomatic gastritis, and other preneoplastic lesions as gastric atrophy or intestinal metaplasia, to gastric adenocarcinoma. This decrease was associated with a concomitant increase in the number of Tregs in peripheral blood and the presence of CagA-positive strains (44), indicating that, indeed, H. pylori favors a Treg-mediated chronic inflammation and consequently the persistence of strains bearing CagA.

Although H. pylori has been shown to reprogram DCs toward a tolerogenic phenotype, previous reports failed to establish the molecular mechanisms involved. To our knowledge, in this paper, we provide for the first time evidence that activation of STAT3 in human DCs upon H. pylori infection might be a key event instructing a Treg-biased adaptive immune response by which H. pylori can evade immune clearance.

We thank Dr. Anne Krug and Prof. Dirk Busch for helpful comments and discussion.

The authors have no financial conflicts of interest.