T cell-derived extracellular vesicles are elevated in essential HTN

Abstract

Extracellular vesicles (EVs) are novel mediators of cell-to-cell communication and appear to mediate the pathogenesis of hypertension (HTN). However, the mechanisms underlying the involvement of EVs in HTN remain unclear. The adaptive and innate immune systems play an important role affecting the kidney and vasculature in animal models of HTN. Evolving evidence shows that immune cell-derived EVs can modulate the immune system in a paracrine fashion and therefore may mediate the effects of inflammation in the pathogenesis of HTN. Therefore, we aimed to understand if specific subtypes of leukocyte/immune cell-derived EVs are altered in essential HTN using an in vivo model of angiotensin II (ANG II)-induced HTN. After 4 wk of ANG II treatment, EVs were isolated from the blood and kidney. EV origin and counts were characterized with Imaging Flow Cytometry, antibody panels targeting platelets, endothelial cells, and leukocytes including B and T cells, monocytes, and neutrophils. Leukocyte-derived EVs (CD45⁺) were elevated in the circulation and kidney tissue in ANG II-induced HTN. Subgroup analysis depicted T cell-derived EVs (CD3⁺) to be significantly elevated in ANG II-induced HTN (3.50e⁺⁵ particles/mL) compared with control groups (9.16e⁺⁴ particles/mL, P = 0.0106). T cell-derived EVs also significantly correlated with systolic blood pressure levels (r² = 0.898, P = 0.0012). In summary, leukocyte-derived EVs, and more specifically T cell-derived EVs (CD3⁺), are elevated in ANG II-induced HTN in the circulation and kidney tissue and correlate well with blood pressure severity. EVs from the circulation and kidney may be sensitive biomarkers for HTN and end-organ damage and may lead to new mechanistic insights in this silent disease.

INTRODUCTION

Hypertension (HTN), a silent disease affecting over 100 million Americans, is a major risk factor for cardiovascular disease (16). Yet, the cause of HTN in the majority of adults is still incompletely understood. The adaptive and innate immune systems are thought to play an important role affecting the kidney and vasculature in HTN (7). In particular, T and B cells, macrophages, inflammatory cytokines, and reactive oxygen species are involved in development and maintenance of HTN (7). However, the interactions between these immune cells and how their actions are mediated in target organs such as kidney and vasculature to influence blood pressure are not well understood.

Extracellular vesicles (EVs) are potential novel communicators, biomarkers, and bioactivators in health and disease such as cardiovascular diseases (2, 10). Less than 1 µm in size, these vesicles carry markers of their parent cells that can be used for their detection (27). They also carry a specific cargo including lipids, proteins, and nucleic acid, which can be delivered to neighboring cells or distal organs. This can lead to changes of phenotype and function in their target cells. EVs have been found to be major players in regulating the immune system (27). Depending on their origin and cargo (RNA/protein), EVs can augment or dampen the immune responses to infections with viruses/microbes and to cancer. In particular, antigen-presenting cells (APCs) can release EVs with a similar palette of surface immune regulatory proteins. This allows APCs to modulate T cell and other immune cell responses nearby or at a distant site (1, 25).

In essential HTN, few studies have focused on the immune regulatory role of EVs. Macrophage-derived EVs (CD68⁺) have been detected in the circulation of a rat model of ANG II-induced HTN where macrophages infiltrate the heart (19). Endothelial cells in culture, when exposed to EVs deriving from a human monocyte cell line THP-1, display enhanced expression of inflammatory markers including inducible nitric oxide (NO) synthase (iNOS), IL-12, and IL-10 mRNA levels compared with controls (19). EVs derived from T cell culture have been shown to reduce endothelium-dependent vasodilation in resistance arteries from mice, likely via a NO-dependent mechanism (14). Our own group found that the levels of leukocyte-derived EVs, and not endothelium-derived EVs, are elevated in spontaneously hypertensive rats (SHRs) and correlate well with the severity of HTN (5). Therefore, we aimed to understand if specific subtypes of leukocyte/immune cell-derived EVs are altered in essential HTN using an in vivo model of ANG II-induced HTN.

METHODS

Animals.

ANG II was delivered via osmotic minipumps (500 ng·kg⁻¹·min⁻¹, model 2004, Alzet) to 12-wk-old 129S6 male mice for 4 wk. A subgroup of animals was also treated with candesartan (10 mg·kg⁻¹·day⁻¹, angiotensin receptor blocker) or a combination of hydralazine (vasodilator), hydrochlorothiazide (diuretic), and reserpine (α-adrenergic inhibitor) (HHR; 30 mg·kg⁻¹·day⁻¹) in drinking water for 4 wk. Systolic blood pressure (SBP) was measured for 2 wk using the tail-cuff method after 2 wk of training daily (31). After 4 wk, mice were euthanized and citrated blood was collected from a cardiac puncture.

EV isolation.

EVs were enriched using differential centrifugation. Platelet-poor plasma was generated with an initial centrifugation step at a relative centrifugation force (RCF) of 5,000 g (centrifuge: 524/5424 R-Rotor FA-45-24-11, Eppendorf) at room temperature for 15 min using 1.5-mL microcentrifuge tubes (Axygen). The supernatant was centrifuged at maximum speed (15,000 rpm, RCF: 21,130 g) for 1 h. The resulting pellet (P21) was resuspended with filtered 0.1 µM HEPES buffer [10 mM HEPES (pH7.4) + 0.15 M NaCl] and centrifuged again at an RCF of 21,130 g.

Isolation of leukocyte-derived EVs from kidneys.

Kidneys were flushed in vivo with 30 mL PBS through the heart and then removed and subsequently incubated with 2 mL of 1 mg/mL collagenase type A plus 5 mM CaCl₂ plus 1 mg/mL DNase I for 30 min at 37 °C. After being minced with a razor blade, digested kidney material was then centrifuged at 1,000 g for 10 min, and the supernatant was spun twice at 10,000 g for 30 min followed by a centrifugation at 21,130 g (28, 33). Thus, an EV pellet (P21) and EV supernatant (SN21000) were generated and stored at −80 °C until EV phenotyping with Imaging Flow Cytometry was performed.

Characterization of EV morphology and size by cryoelectron microscopy and Nanoparticle Tracking Analysis.

EV morphology was analyzed using cryoelectron microscopy. Purified samples were verified by standard methods for cryoelectron microscopy (cryoEM) to determine EV morphology (32). EVs were then imaged using a Tecnai F20 Twin emission electron microscope at the Molecular Electron Microscopy Core at the University of Virginia (https://med.virginia.edu/molecular-electron-microscopy-core/).

All images were recorded with a Gatan 4K × 4K pixel charge-coupled device camera. EV size was assessed with Nanoparticle Tracking Analysis (NTA) using the Zetaview PMX 110 multiple parameter particle tracking analyzer (Particle Metrix, Meerbusch, Germany) in size scatter mode using Zetaview software (version 8.02.28). All samples were diluted in PBS to a final volume of 4 mL. Ideal measurement concentrations were found by pretesting the ideal particle per frame value (20–100 particles/frame).

The system was calibrated using 100-nm polystyrene beads (Nanosphere Size Standard catalog no. 3100A, Thermo Scientific), and plasma vesicle profiles were then recorded and analyzed at 11 camera positions with a 2-s video length, a camera frame rate of 30 frames/s, and a temperature of 21°C. Interassay variation was assessed by analyzing the same sample aliquots with the same settings and on the same day and operator. All samples were taken two times, and an average in size or concentration was determined after acquisition.

SDS-PAGE and Western blot analysis.

Western blot analysis was performed by SDS-PAGE. Proteins on Western blots were detected using the following antibody raised against tumor susceptibility gene 101 [TSG101; Clone EPR7130(B), catalog no. 125011, Abcam]. Acquisition of the fluorescent signal was performed by the Odyssey infrared imaging system with resolution set at 169 lm (LI-COR Biosciences) (17). We used 1.0 μg/mL mouse anti Tsg101 antigen overnight as the primary antibody. Dye-coupled secondary antibody (0.1 μg/mL, LI-COR Biosciences) in Odyssey blocking solution was diluted at 1:1 with PBS and 0.15% (vol/vol) Tween-20 and added for 2 h at room temperature. Acquisition of the florescent signal was performed by the Odyssey infrared imaging system with the resolution set at 169 µm (LI-COR Biosciences). Image studio software (version 2.1, LI-COR Biosciences) was used to analyze and export images.

Imaging Flow Cytometry: EV labeling and phenotyping.

EVs were labeled for Imaging Flow Cytometry using two antibody panels directed toward 1) platelet/endothelial and leukocyte surface markers and 2) B and T cell markers, monocytes, and neutrophils. The antibodies used were as follows: CD3-FITC for T cell-derived EVs (clone 145-2C11, catalog no. 100306, BioLegend), CD11b-PE for indicating monocyte/macrophage-derived EVs (clone M1/70, catalog no. 1012082, BioLegend), CD105-PE for indicating endothelium-derived EVs (clone MJ7/18, catalog no. 120408, BioLegend), acknowledging that CD105 is found on a heterogeneous group of cells including activating and proliferating endothelial cells, CD45-PE/DAZZLE for indicating leukocyte-derived EVs (clone 30-F1, catalog no. 103146, BioLegend), Ly6G-Violet 421 for indicating neutrophil-derived EVs (clone 1A8, catalog no. 127628, BioLegend), CD19-APC for indicating B cell-derived EVs (clone 6D5, catalog no. 115530, BioLegend), and CD41-APC (clone MWReg, catalog no. 133914, BioLegend).

Single staining was performed adding 1 µL of each individual antibody to P21 while the antibody mix was diluted to 20 µL before being added to P21. In both cases, single staining and antibody mix were precentrifuged at 21,000 g for 1 h to avoid antibody aggregation. Samples were incubated 1 h in darkness and at room temperature. After dilution with 1 mL HEPES (10 mM plus 0.15 M NaCl) buffer, labeled EVs were centrifuged at 21,000 g and the final pellet was resuspended in 50 µL HEPES buffer before Imaging Flow Cytometry acquisition. Phenotyping of specific surface antigens on EVs was detected using Imaging Flow Cytometry (Amnis Image-StreamX Mark II, Luminex) according to our previous publications (4, 17). The flow cytometer Amnis Image-StreamX Mark II was used. Data acquisition was performed in the flow cytometry facility at the University of Virginia (https://med.virginia.edu/flow-cytometry-facility/). Briefly, all lasers required for the fluorochromes used were adjusted to full power, including the 758-nm laser for scatter. A collection gate was established based on scatter intensity that eliminates the speed beads. All controls were applied to avoid false events. Precisely, in the following order: unstained EVs, stained EVs, buffer with antibody mixes, and compensation control for each fluorescent used (Fig. 1D) (4, 17). The acquired raw data were then analyzed using IDEAS software (version 6.02, Amnis/Luminex) and FCS Express7 DeNovo software (https://denovosoftware.com/) to create the histograms and dotplots.

Fig. 1.

Rigorous extracellular vesicle (EV) characterization. A: cryoelectron microscopy images of EVs from mouse platelet-poor plasma after 21,130-g centrifugation (P21). Left: example of a normotensive EV image (control); right: example of a hypertensive EV image [ANG II-induced hypertension (HTN)]. B: EV size distribution (P21) by Nanoparticle Tracking Analysis in ANG II-induced HTN compared with control mice. No significant differences were found (range = 0–1,000 nm, P = 0.62). We used ordinary one-way ANOVA based on Sidak's multiple-comparison test to perform this analysis. C: positive detection of the tumor susceptibility gene 101 (TSG101) marker in Western blot in the P21 EV pellet. D: enumeration and phenotyping of EVs using Imaging Flow Cytometry (IFC; Amnis Image-StreamX Mark II). The gate strategy was set on scatter without speed beads and using following samples/controls: unstained EVs, stained EVs with antibodies, and buffer with reagents. Standardized fluorescence intensity units (MESF) were calculated for the CD3 marker using Bangs Laboratories Beads (see Ref. 12), showing levels typical for EVs (MESF value up to several 1,000 for EVs) and not cells or debris (typical MESF value up to 100,0000 for cells).

Statistics.

All data were analyzed using Student’s t test two-tailed, unpaired, and ordinary one-way ANOVA based on Sidak’s multiple-comparisons test. Graph Pad-Prism (version 8.4.2) was applied for this analysis, and P < 0.05 indicates statistical significance.

RESULTS

A rigorous and basic characterization of EVs isolated and enriched from platelet-poor plasma by differential centrifugation from 129S6 normotensive mice is shown in Fig. 1. Cryoelectron microscopy of the P21 EV pellet of normotensive and hypertensive mice (Fig. 1A) confirmed enrichment for vesicles with a double lipid bilayer but also demonstrated heterogeneity in size and density. NTA was used for particle size and count analysis (Fig. 1B), confirming that the majority of EVs were 100- to 200-nm size in diameter (average EV size: 210 ± 64.80 nm). There were no statistical differences in number and sizes of EVs between EVs from normotensive and hypertensive mice (n ≥ 10). General EV protein markers such as TSG1010 were also found in EVs by Western blot analysis of the P21 EV pellet (Fig. 1C). Measurement of the molecules of equivalent soluble fluorescence (MESF) values for CD3-derived EVs (T cell-derived EVs) confirmed that our analysis with Imaging Flow Cytometry (Fig. 1D) detected EVs (typical MESF value up to sveral 1,000 for EVs) and not cells or debris (typical MESF value up to 100,000 for cells).

After characterizing our enriched EV preparation as outlined above, we tested EVs in the circulation of 129S6 mice after 4 wk of treatment with ANG II (see Fig. 2). ANG II-infused mice had significantly higher SBP levels (Fig. 2A) compared with controls after 4 wk [173.79 ± 4.9 mmHg (ANG II-induced HTN) vs. 129.8 ± 6.2 mmHg (controls), P < 0.054]. Levels of endothelium-derived (CD105⁺; Fig. 2B)- and platelet-derived EVs (Fig. 2C) did not increase significantly; however, levels of leukocyte-derived EVs (CD45⁺) were significantly increased after 4 wk (ANG II-induced HTN: 1.36e⁺⁶ particles/mL and controls: 4.10e⁺⁵ particles/mL, P = 0.035; Fig. 2D). In addition, endothelium-derived EVs (CD105; Fig. 2B) or platelet-derived EVs (CD41; Fig. 2C) did not correlate with blood pressure severity while levels of leukocyte-derived EVs (CD45⁺) correlated well with the degree of SBP after 4 wk (r² = 0.7861, P = 0.001; Fig. 2D).

Fig. 2.

Extracellular vesicles (EVs) after 4 wk in the circulation and from kidneys. Systolic blood pressure (SBP) levels were significantly increased after 4 wk of treatment with ANG II (n ≥ 10; A). B and C: endothelium- (B) and leukocyte-derived (C) EVs. Leukocyte-derived EVs increased significantly after 4 wk (*P < 0.035), whereas platelet- and endothelium-derived EVs did not. Endothelium-derived EVs did not correlate with SBP; in contrast, leukocyte-derived EVs (CD45⁺) correlated positively with SBP (r² = 0.7861, P = 0.0001). Student’s t test two-tailed, unpaired, was performed for this analysis. HTN, hypertension.

We also tested EVs isolated from kidney tissue from mice after 4 wk of treatment with ANG II. Levels of leukocyte-derived EVs (CD45⁺) enriched from the kidney tissue were significantly higher in the kidneys from hypertensive mice compared with normotensive mice (ANG II-induced HTN: 7.98e⁺⁵ particles/mL and controls: 4.13e⁺⁵ particles/mL, P = 0.0257; Fig. 3A), and we observed a positive correlation with SBP levels (r² = 0.5247, P = 0.0656; Fig. 3B).

Fig. 3.

Kidney-derived extracellular vesicles (EVs) in ANG II-induced hypertension (HTN). Leukocyte-positive (CD45⁺) kidney-derived EVs were significantly increased in ANG II-induced HTN (A; P = 0.0257) and correlated with the severity of systolic blood pressure (SBP; B; r² = 0.5247, P = 0.0656, n ≥ 3). Student’s t test two-tailed, unpaired, was used for this analysis.

To determine whether EV levels and subtypes are influenced by angiotensin type 1 receptor (AT₁R) activation, we treated mice with either candesartan, an angiotensin receptor antagonist, or a combination of hydrochlorothiazide (diuretic), hydralazine (vasodilator), and reserpine (α-adrenergic inhibitor) (HHR). Blood pressure levels were equally normalized with both drug regimens [176.30 ± 4.31 mmHg (ANG II-induced HTN) vs. 126.98 ± 5.49 mmHg (controls), P < 0.0054; 133.94 ± 6.94 mmHg susceptibility (ANG II-induced HTN + HHR), P < 0.0026; 136.21 ± 8.34 mmHg (ANG II-induced HTN + candesartan), P < 0.034; n ≥ 5 each group; Fig. 4A].

Fig. 4.

Subgroups of leukocyte-derived extracellular vesicles (EVs) after treatment with antihypertensive agents in the circulation. A: candesartan (Cand) and hydralazine/hydrochlorothiazide/reserpin (HHR) treatment groups had equal systolic blood pressure (SBP) lowering when treated simultaneously with ANG II (HHR: P = 0.0026 and Cand: P = 0.034, n ≥ 10). B and C: EV counts in leukocyte-derived EVs decreased after normalization of SBP (P = 0.0054) and correlated significantly (r² = 0.40, P = 0.0095) with SBP (n ≥ 3). D: T cell-derived EVs (CD3⁺) were significantly elevated in ANG II-induced hypertension (HTN) compared with control mice (P = 0.0106), ad they also strongly correlated with SBP (r² = 0.898, P = 0.0012, n ≥ 5). E-G: no significance was reported for all other immune markers including monocytes, neutrophils, and B cells. Results of B cell-derived EVs (CD19⁺) are provided as an example; n reflects number of animals tested. We used ordinary one-way ANOVA based on Sidak's multiple-comparison test to perform this analysis.

Compared with the control group (ANG II-induced HTN), leukocyte-derived EV (CD45⁺) levels were significantly lower in the HHR-treated group and numerically lower in the candesartan treated-group (ANG II-induced HTN: 9.11e⁺⁵ particles/mL and HHR: 2.22e⁺⁵ particles/mL, P = 0.0054; ANG II-induced HTN: 9.11e⁺⁵ particles/mL and candesartan: 4.90e⁺⁵ particles/mL, not significant; Fig. 4B). Again, leukocyte-derived EV (CD45⁺) levels correlated well with SBP (r² = 0.40, P = 0.0095; Fig. 4C). These data suggest that EV levels are not determined by AT₁R activation but rather by the level of blood pressure.

We next queried which subgroups of leukocyte-derived EVs (CD45⁺) might be involved in ANG II-induced HTN. As shown in Fig. 4D, levels of T cell-derived EVs (CD3⁺) were significantly elevated in ANG II-induced HTN (controls: 9.16e⁺⁴ particles/mL vs. ANG II-induced HTN: 3.50e⁺⁵ particles/mL, P = 0.0106) and correlated significantly with SBP level (r² = 0.898, P = 0.0012). In contrast, no differences were observed in B cell, neutrophil, or monocyte-derived EVs (Fig. 4, E–G). Levels of T cell-derived EVs (CD3⁺) were numerically lower in mice treated with HHR or candesartan compared with untreated mice but did not reach statistical difference.

DISCUSSION

EVs have gained significant attention as potential new surrogate biomarkers for endothelial dysfunction and vascular damage. In clinical studies, EV levels have been found to be higher in patients with essential HTN and correlate with the severity of HTN (22, 23). Earlier studies primarily examined only endothelium-derived EVs in hypertensive patients using several endothelial markers including CD62E⁺ and CD31⁺/CD42⁻ EVs (8, 9, 18, 22, 23, 29, 34). However, Zu et al. (34) also showed that leukocyte-derived EVs (CD45⁺) were elevated in hypertensive patients compared with controls, whereas there was no difference in platelet-derived EVs (CD41⁺). Interestingly, their study demonstrated that endothelium-derived EVs (CD144⁺) were lower in hypertensive patients (34), in contrast to our findings that showed that endothelium-derived EV (CD105⁺) levels were not different in ANG II-induced HTN in mice after 4 wk. The difference could be explained by use of a different endothelial cell marker. However, similarly to Zu et al., using the SHR model of essential HTN (5), we recently reported that leukocyte-derived EVs (CD31⁺ and CD45⁺) were significantly elevated in SHRs at 12 wk of age compared with age-matched normotensive Wistar-Kyoto rats. In this model of essential HTN, we also observed that circulating levels of EVs correlated with blood pressure severity (5).

In the present study, we followed the guidelines from the EV Flow Cytometry Working Group (30) and Minimal Information for Studies of Extracellular Vesicles (MISEV2018) (26) to rigorously characterize circulating EVs in normotension and ANG II-induced HTN. We confirmed by cryoelectron microscopy that our enriched EV preparation contained a heterogeneous group of lipid bilayered vesicles with an average size of 210 nm and expressing the EV-specific marker TSG101 (26). However, we observed no statistical difference in EV sizes between normotensive or hypertensive animals. To our knowledge, size differences were not analyzed in previous studies (13, 20).

Similar to our observation in the SHR HTN model (5), leukocyte-derived EVs (CD45⁺) were increased in ANG II-induced HTN in mice and also correlated significantly with blood pressure severity. Taken together, both SHR essential HTN and ANG II-induced HTN models and the human data (34) suggest that circulating leukocyte-derived EVs (CD45⁺) are potentially clinical biomarkers of the severity of HTN that might therefore reflect early end-organ damage in HTN and could be used to guide therapy. In this regard, we found that kidneys from hypertensive animals had higher leukocyte-derived EV (CD45⁺) counts. These kidney-isolated EVs also correlated with SBP levels. As T cells have been implicated in sodium retention in the kidney during HTN and infiltrating the kidney in hypertensive animals (15, 21), it remains to be determined whether leukocyte-derived EVs (CD45⁺) in the kidney come from the circulation or are generated in the kidney from infiltrating immune cells (24). Furthermore, our observation raises the question whether EVs in tissue are simply biomarkers or have direct functional role in mediating tissue injury or regulation of blood pressure. EVs have also been described to elicit immune cell responses nearby or at a distant site, suggesting that they may play a role in organ-to-organ communication in hypertension.

We used two different antihypertensive regimens in ANG II-induced HTN to determine whether different treatments have different effects on circulating EV levels and more specifically whether their levels are dependent on AT₁R activation. Our data suggest that blood pressure lowering influences EV levels, independent of AT₁R, as HHR combination lowered EV levels similarly as candesartan (angiotensin receptor blocker) and was statistically significant compared with the ANG II-induced HTN group. We also showed that leukocyte-derived EVs (CD45⁺) were numerically lower after treatment with HHR. This finding suggests that normalization of HTN leads to reduced numbers of EVs, indicating a possible mechanistic role of EVs in HTN pathogenesis.

Interestingly, further subgroup analysis of leukocyte-derived EVs showed that CD3⁺ or T cell-derived EVs were significantly higher in ANG II-induced HTN and numerically lower after blood pressure treatment. CD3⁺ EVs also correlated with blood pressure levels. This was not observed in B cell (CD19⁺)-, monocyte (Ly6g⁻/CD11b⁺)-, and neutrophil (Ly6g⁺/CD11b⁻)-derived EVs. This finding is in line with the paradigm that T cells play a dominant role in ANG II-induced HTN (6). To our knowledge, T cell-derived EVs (CD3⁺) have only been previously studied in pulmonary HTN, where T cell-derived EVs (CD3), and not macrophage/monocyte-derived EVs (CD14 and CD68), were able to distinguish patients with pulmonary HTN from healthy patients (11).

In summary, leukocyte-derived EVs (CD45⁺), and more specifically T cell-derived EVs (CD3⁺), are elevated in ANG II-induced HTN in the circulation and kidney tissue and correlate well with blood pressure severity. EVs have been described to be regulators of immune processes and are messengers in the vascular and renal tubular structure in HTN (3). EVs may play a significant functional role in HTN pathogenesis besides being a biomarker reflecting HTN severity. Further analysis of circulating and tissue EVs and their cargo and origin can lead to new mechanistic insights in this silent disease and exploration of new diagnostic and treatment targets.

GRANTS

This work was supported by American Heart Association Postdoctoral Fellowship 18POST34060071 (to S.L.) and National Heart, Lung, and Blood Institute Grant K23-HL-126101 (to U.E.). The data described in this study were gathered Amnis ImageStream funded by National Center for Research Resources Grant 1S10-RR-03163301.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

S.L., T.H.L. and U.E. conceived and designed research; S.L. performed experiments; S.L. analyzed data; S.L., L.M., J.L., J.G., T.H.L. and U.E. interpreted results of experiments; S.L. prepared figures; S.L., T.H.L. and U.E. drafted manuscript; S.L., L.M., J.L., J.G., T.H.L. and U.E. edited and revised manuscript; S.L., L.M., J.L., T.H.L. and U.E. approved final version of manuscript.