Membrane dynamics and interactions in measles virus dendritic cell infections


For correspondence. E-mail; Tel. (+49) 931 201 49895; Fax (+49) 931 201 49553.


Viral entry, compartmentalization and transmission depend on the formation of membrane lipid/protein microdomains concentrating receptors and signalosomes. Dendritic cells (DCs) are prime targets for measles virus (MV) infection, and this interaction promotes immune activation and generalized immunosuppression, yet also MV transport to secondary lymphatics where transmission to T cells occurs. In addition to MV trapping, DC-SIGN interaction can enhance MV uptake by activating cellular sphingomyelinases and, thereby, vertical surface transport of its entry receptor CD150. To exploit DCs as Trojan horses for transport, MV promotes DC maturation accompanied by mobilization, and restrictions of viral replication in these cells may support this process. MV-infected DCs are unable to support formation of functional immune synapses with conjugating T cells and signalling via viral glycoproteins or repulsive ligands (such as semaphorins) plays a key role in the induction of T-cell paralysis. In the absence of antigen recognition, MV transmission from infected DCs to T cells most likely involves formation of polyconjugates which concentrate viral structural proteins, viral receptors and with components enhancing either viral uptake or conjugate stability. Because DCs barely support production of infectious MV particles, these organized interfaces are likely to represent virological synapses essential for MV transmission.


Danger signals such as pathogen encounter launch a maturation programme in immature myeloid dendritic cells (DCs), which dramatically alters their phenotype and activity to enable acquisition of a migratory phenotype, and thereby exit from peripheral tissues and homing to secondary lymphoid tissues where they present antigens to scanning T cells. The ability of DCs to initiate and shape adaptive immune responses is largely determined by the pathogen which triggers and modulates their maturation programme (Steinman et al., 2003; Schneider-Schaulies and Dittmer, 2006; Pohl et al., 2007a). Inadequate DC maturation and inhibitory rather than activatory signalling to T cells are strategies widely exploited in viral immunomodulation and measles virus (MV) has been analysed as an efficient effector in this context (reviewed in Schneider-Schaulies et al., 2003; Servet-Delprat et al., 2003; Schneider-Schaulies and Dittmer, 2006). Their migratory activity has, however, also coined DCs as Trojan horses mediating viral transport to secondary lymphatics where transmission to conjugating T cells can occur (Pohl et al., 2007a; de Swart et al., 2007; Lemon et al., 2011). MV interaction with receptors at the DC plasma membrane essentially determines both their activation and viral trapping and uptake, while viral transmission, signalling or presentation of viral components to T cells involves formation of organized membrane interfaces [virological (VS) or immunological synapses (IS)] segregating receptors at the target cell. Given the limited ability of DCs to support MV replication in the absence of CD40 ligation (Servet-Delprat et al., 2000), MV transmission from these cells as crucial for subsequent dissemination most likely is strictly dependent on the formation of a tight interaction of target cells with DC having trapped MV to attachment receptors only (referred to as ‘trans-infection’) or supported de novo synthesis of viral proteins and assembly of viral ribonucleoprotein complexes or particles at defined membrane compartment (cis-infection).

Measles virus interaction with DC surface receptors

A variety of plasma membrane receptors has been implicated in MV interaction which can be divided into those promoting (bona fide revealed by conferring susceptibility to MV infection in Chinese hamster ovary cells) or those enhancing uptake by attachment (DC-SIGN) or eventually supporting fusion (substance P receptor, moesin) and others mediating recognition and signalling (e.g. TLR2, FcγRII, yet also DC-SIGN). Tissue resident immature DCs are equipped with a plethora of pattern recognition receptors (PRR) also including C-type lectin receptors such as DC-SIGN, which endocytoses ligands, yet also signals to modulate TLR-driven gene activation (Gringhuis et al., 2007). Although not yet confirmed for DCs, wild-type MV activated signalling via the TLR2/CD14 complex in monocytes and this ability segregated with residue 481 within the viral receptor binding haemagglutinin (H) protein (Bieback et al., 2002). MV-induced TLR2 signalling might be involved in DC maturation, e.g. by induction of IL-1β, yet also causes upregulation of CD150, the MV entry receptor, within 12 h (see below; Bieback et al., 2002). Although the ability of MV to induce DC maturation may be limited (Servet-Delprat et al., 2000; Abt et al., 2009), it is apparently sufficient to mobilize these cells and to promote induction of virus-specific immunity. As for other viruses also including human immunodeficiency virus (HIV), DC-SIGN serves as MV attachment receptor on DCs and enhances viral uptake (de Witte et al., 2006). Except for phleboviruses (Lozach et al., 2011), DC-SIGN does not act as viral entry receptor, yet rather may act to capture and at least partially route its ligands into degradative or storage compartments from where they can be re-localized and surface exposed for transmission to T cells (Dale et al., 2011; Chen, 2012). This has been suggested to occur for MV as well, and in line with this, MV-loaded DCs were able to transmit virus to both CD4+ and CD8+ T cells and MV-derived antigens were efficiently presented to CD4+ T lymphocytes after antigen uptake (de Witte et al., 2008).

Among the wild-type MV entry receptors, nectin-4 confers susceptibility to infection of epithelial cells, which is crucial for MV exit from the host (Muhlebach et al., 2011; Noyce et al., 2011). Expression of CD150 (also referred to as signalling lymphocyte activating molecule, SLAM) is confined to cells of the haematopoietic lineage, and is the major MV entry receptor into cells of this compartment essentially defining MV tropism in vitro and in vivo (de Swart et al., 2007; Lemon et al., 2011). Homotypic interactions that regulate T and monocyte responses (Veillette et al., 2007) are clearly outcompeted by MV H protein that binds with much higher affinity (Hashiguchi et al., 2011). Surface expression of CD150, however, requires cellular activation, and thus, expression levels of CD150 are very low on immature DCs in vitro. CD150 was not co-detected on DC-SIGN+ cells in respiratory tissue blocks of healthy individuals, which led to the suggestion that MV trapping by DCs rather than infection would be key to transmission to T cells (de Witte et al., 2008).

Upregulation of CD150 on DCs in response to PRR signalling

Although TLR2 signalling can promote CD150 surface display (Bieback et al., 2002), it is unclear whether this also applies to DCs, and if so, if MV would be trapped sufficiently long and co-clustering of CD150 with trapped virions would occur. We were able to show that DC-SIGN ligation by MV, yet also mannan or antibodies, caused transient vertical surface recruitment of CD150 from intracellular stores, and co-segregation of attachment and entry receptor within ceramide-enriched membrane microdomains (Avota et al., 2011). Local accumulation of ceramide at the plasma membrane as a result of consumption of sphingomyelin (SM) by sphingomyelinase (SMases) induces coalescence of raft microdomains into extended ceramide-enriched platforms. These trap and cluster receptor molecules and are platforms where signalling is initiated and certain pathogens including viruses can enter (Gulbins and Kolesnick, 2003). At the plasma membrane, ceramide accumulation often results in response to activation of acid SMase (ASMase) catalysed by, for instance, signalling via TNF-R family members, yet also viruses (Gulbins and Kolesnick, 2003). Ceramide-enriched domains were also found to catalyse inward budding of a subset of intralumenal vesicles (ILVs) into multivesicular bodies. There, sorting of cargo occurred independently of the endosomal sorting complex required for transport (ESCRT) machinery but rather relied on raft-based microdomains for lateral segregation and the activity of neutral sphingomyelinase (NSMase; Trajkovic et al., 2008). Interestingly, ILVs destined for release as exosomes were dependent on ceramide generation, while the ESCRT machinery was only required for ILVs destined for lysosomal sorting.

Rationales to study if MV could possibly modify the DC plasma membrane to create an optimal environment for uptake by DC-SIGN ligation and subsequent ceramide generation were that: (i) DC-SIGN traps a variety of viruses (including MV) to promote their endocytic uptake or transmission to T cells (de Witte et al., 2006; 2008) and (ii) endocytic uptake of some viruses into non-APCs was linked to local generation of ceramides and thereby, local receptor concentration (Riethmuller et al., 2006). MV transiently activated both ASMase and NSMase and induced plasma membrane display of ASMase and ceramides in a MV glycoprotein (gp) and DC-SIGN-dependent manner (Avota et al., 2011; Fig. 1). This was accompanied by transient vertical surface transport of CD150 from intracellular stores which co-stained for ASMase indicating that it is co-transported with this glycosylated enzyme. Thus, MV attachment induces formation of a plasma membrane environment transiently concentrating its receptors and ceramides, which, given the general supportive role of those for membrane fusion (Utermohlen et al., 2008), explains DC-SIGN-mediated enhancement of infection. These findings support the role of MV cis-infection of DCs in vivo and thereby the role of DC-SIGN+ DCs as early targets for infection. DC-SIGN-dependent SMase activation was also found essential for its membrane proximal and distal signalling and, thereby, PRR cross-talk and subsequent NF-κB activation which may essentially contribute to magnitude and shaping of T-cell responses (Gringhuis et al., 2007; Avota et al., 2011).

Figure 1.

Receptor-stimulated surface display of ceramides. Surface display of ceramides (red) on DCs prior to (upper row) and 20 min following ligation of DC-SIGN (green) (middle and bottom rows, in this example, by mannan).

Sorting of MV envelope proteins and MV particle formation

As evident from their limited ability to support production of infectious MV observed by us and others (Servet-Delprat et al., 2000), DCs may restrict MV replication to retain their integrity during DC migration. As a prerequisite to understand potential DC-specific regulations of viral protein trafficking to relevant assembly sites, assembly and/or budding, for MV, these needed to be first characterized in fully permissive cells. For obvious reasons, these analyses focused on envelope proteins and, within those, especially MV M protein which, in analogy to its functional orthologues, is considered as major regulator of MV particle assembly and budding. Thus, ablation of M protein expression or attenuation of its interaction with the cytoplasmic domain of the viral H gp is associated with diffuse distribution of M protein and enhanced fusion activity at the expense of particle formation (Tahara et al., 2007). In intact virions, M protein coats the viral ribonucleoprotein complex rather than the membrane, and the helical symmetry of the M protein cage differs from that of the nucleocapsid cargo (Liljeroos et al., 2011). Encagement of ribonucleoprotein particles (RNPs) by M protein occurs in the cytosol consistent with its role as transcriptional repressor (because it blocks access of the active transcriptase complex) and its requirement for efficient RNP transport to the budding sites (Runkler et al., 2007).

Measles virus M protein associates with medium- and high-density membranes in fractionated extracts of infected or transfected cells, yet detection of high molecular weight complexes is confined to plasma membrane fractions (Vincent et al., 2000; Salditt et al., 2010). M protein of either MV strain oligomerized also upon plasmid-driven expression (Pohl et al., 2007b; Runkler et al., 2007), and occasionally migrates as double band in SDS-PAGE (Pohl et al., 2007b). About 20% of total M protein (both species detected) co-floatated with detergent-resistant membrane microdomain (DRM) fractions and this was enhanced upon coexpression of the F protein (Pohl et al., 2007b). Both M and F protein promoted virus-like particle (VLP) production, which was not enhanced upon their coexpression. Moreover, release of N protein in VLPs required coexpression of stable M protein able to associate with the plasma membrane (Runkler et al., 2007). In general, relative amounts of M proteins released in the VLP assays corresponded to those determined for virus particles indicating that MV budding is overall inefficient.

Its co-floatation with intermediate-sized membrane fractions and transient ubiquitination suggested that trafficking and VLP/budding promoting activity of MV M proteins could involve the ubiquitin/ESCRT system. However, M protein did not detectably redistribute YFP-tagged Vps4, AIP-1 or Tsg101 to the plasma membrane (Salditt et al., 2010). Release of M protein-containing material or infectious virus was reduced by about 70% or 40%, respectively, upon overexpression of dnVps4, while that of dnTSG101, dnAIP-1 or dnCHMP3 had little or no effect (neither of these proteins affected release of F protein-driven VLPs). Since MV M proteins with putative candidate L domain motifs (PSVP, YMFL and FKVL) replaced by alanine stretches also differed neither with regard to subcellular distribution nor VLP release from the authenic M protein (Salditt et al., 2010), MV M protein trafficking and VLP production apparently do not essentially involve interaction of known L domains with ESCRT components. In contrast to that of the putative canonical L domains, ablation of the PIQP motif (aa 20–23) (which is part of a PIQPTTY motif resembling the PTAPPEY motif in VP40, also with regard to its N-terminal position) caused complete loss of plasmid-driven VLP release.

Membrane microdomains in MV budding

Members of the tetraspanin (tspan) family also organize membrane microdomains (tetraspanin-enriched microdomains, TEMs). TEM organization requires tspan palmitoylation which self-interacts into primary and secondary complexes (the latter co-recruiting cellular molecules). TEMs are implicated in cell activation, migration and fusion and, in DCs, in formation of functionally relevant membrane microdomains and thereby, T-cell activation (Jones et al., 2011; Rubinstein, 2011). TEMs also serve as constituents of a universal egress gateway for HIV-1, but not, for instance, for influenza A virus, although both viruses exit through DRMs (Booth et al., 2006; Welsch et al., 2007). In contrast, MV M protein marginally colocalized with CD63 and CD82 (and Lamp-1), moderately with CD9 and strongly with CD81, while interestingly, Ebola virus VP40 seemingly preferred domains enriched in CD9 (which redistributes upon VP40 expression) rather than CD81 (Salditt et al., 2010). Whether these differences are reflected selection of specific membrane microsegments for budding and a differential sensitivity of particle formation to tspan-specific antibodies as recently confirmed for HIV and MV's close relative, canine disemper virus, remains to be determined. Interestingly, a prominent role of CD9 yet not other tspans in canine distemper virus-mediated membrane fusion became evident, and this related to CD9 clustering into net-like structures and induction of zippering microvillar structures at the cell–cell interface (Singethan et al., 2008).

Raft microdomains were implicated as sites of particle production for many enveloped viruses also including MV, and this mainly based on co-floatation of raft-associated cellular proteins with viral components, their colocalization with raft components such as GM1 or lack of particle production upon cholesterol extraction. Viral lipidomes will reveal the role of defined lipids in viral particle production yet so far are only determined for hepatitis C virus, HIV and human cytomegalovirus (Brugger et al., 2006; Liu et al., 2011; Merz et al., 2011). At least for HIV, the importance of sphingolipids, as especially reflected by selective enrichment of SM and dihydro-SM as virion constituents in the viral replication cycle, has thus been clearly revealed (Brugger et al., 2006). Whether SMase activity is also involved in budding of viral particles from internal or plasma membrane compartments is as yet unknown, yet the low abundance of ceramides in HIV and hepatitis C virus virions seems to argue against a role of SMases at least for these viruses. Unfortunately, the MV lipidome is not available, nor has particle production from cells compromised in this enzymatic activity been analysed.

Accumulation of viral proteins at DC/T cell interfaces and functional consequences

Formation of organized DC/T cell interfaces in MV pathogenesis serves at least two, if not three, functions, two of which involve (immune synapse, IS) to activate or suppress T-cell activation, while a transmission interface between MV-infected or -loaded DCs and T cells may mediate MV transfer to T cells and subsequent dissemination (virological synapse, VS).

Activation of both humoral and cytotoxic effector functions which essentially limit and clear the infection argues that functional IS as required for priming and cytotoxicity are established (de Witte et al., 2008). MV is, however, also known for its ability to cause transient immunosuppression, and viral regulation of DC maturation, viability and T-cell stimulatory activity has been considered as central in this regard (reviewed in Kerdiles et al., 2006; Schneider-Schaulies and Schneider-Schaulies, 2009). MV interaction with PRRs can regulate downstream NF-κB-dependent gene transcription, and while this may be involved in cytokine imbalance in MV-induced immunosuppression, it does not explain the inability of MV-infected DCs to promote T-cell expansion in vitro. Their ability to restrict MV replication and partial activation on MV exposure may be prerequisite to preserve DC morphology and induction of motility (Shishkova et al., 2007). In conjugates involving MV-infected DCs and allogenic T cells, MV H protein clusters appeared as polarized towards the contact interface which, given the inability of MV–DCs to promote T-cell activation, revealed a surprisingly normal IS architecture with regard to CD3 and MHC class II distribution. On live cell analysis, it became, however, evident that the majority of MV–DC interactions with T cells were short-lived and did not elicit sustained Ca2+ mobilization as required for T-cell activation (Shishkova et al., 2007). Suggesting an active role of the MV gps in loss of IS stability and function, both parameters were partially improved when a recombinant MV encoding vesicular stomatitis virus G protein was used for DC infection. On surface interaction with an unknown receptor on T cells, MV abrogates T-cell receptor-driven activation of the PI3 kinase thereby targeting key molecules involved in cell cycle progression. At a cellular level, dynamic reorganization of the actin cytoskeleton as required of IS formation and receptor segregation there proved to be substantially paralysed which was recently linked to sequential activation of neutral and acid SMases in these cells (Muller et al., 2006; Gassert et al., 2009). MV gp signalling may thus impede cytoskeletal remodelling at the IS and thereby, its function, and this may be further enhanced by abrogation of recruitment of IS stabilizing and activating receptors. We confirmed the importance of plexinA1 on T cells for their efficient expansion by allogenic DCs (Eun et al., 2006; Tran-Van et al., 2011), and it appeared that actin-dependent IS recruitment of this molecule was strongly impaired if T cells were exposed to MV prior to conjugating to mature DCs or stimulatory beads. Neuropilin-1, which can act to stabilize the IS eventually by plexinA1 interaction in turn, was found downmodulated on MV-infected DCs (Tran-Van et al., 2011). Strikingly, MV–DCs readily secreted repulsive plexinA ligands, semaphorins 3A and 6A, which caused transient loss of F actin and microvillar extensions, chemotactic motility and expansion upon exogenous supplementation. Thus, interference with actin cytoskeletal remodelling by MV gp interaction and production of repulsive rather than adhesive plexinA ligands may collectively act to physically destabilize the IS thereby compromising the T-cell activation within the conjugates.

Measles virus transmission in DC/T cell conjugates: repatterning of viral proteins and cellular components

Efficient cell-associated acquisition has proven a general feature of T-cell infection in vitro and in vivo which may be particularly important in tissues where diffusion of particles rather than concentration would limit transfer to recipient cells (Felts et al., 2010; Dale et al., 2011; Chen, 2012). For HIV, cell-associated viral transfer can exceed that by viral particles up to 100-fold (Chen, 2012), and similarly, MV transmission to T cells may involve DCs because virus-loaded DCs were found in proximity to T cells in secondary lymphatics (Lemon et al., 2011). Moreover, there was evidence of eGFP transfer from infected DCs to conjugating T cells in vitro and this involved both extended interfaces and long protrusions connecting potential donors and targets (de Witte et al., 2008). In line with these observations, formation of polyconjugates between MV-infected DCs and autologous T cells and – albeit rarely – actin-rich protrusions co-staining for MV H protein was seen in cocultures (Koethe et al., 2012) indicating that MV may also exploit organized structures for transmission to T cells. In addition to H protein, accumulation of MV P protein, a structural component of the viral RNP, was also seen at the interfaces. On direct comparison, MV acquisition by T cells by cis-infected DCs exceeded that by DCs loaded with the virus or cell-free infection up to threefold. Although it has not been clarified, the comparatively still low transmission rate may at least partially lie in destabilizing signals given by the MV gps which accumulate at the DC/T cell interface (Koethe et al., 2012). Pre-activated T cells acquire MV from infected DCs much more efficiently and this may relate both to induction of CD150 and to enhancement of viral replication upon T-cell activation. In line with observations made in other VS, CD150 on T cells segregated detectably into the DC/T cell interface and latrunculinB sensitivity revealed the importance of actin dynamics in this recruitment. Transmission was partially abrogated by pre-exposure of DCs to MV H-specific and T cells to CD150-specific antibodies, while blocking of CD150 on DCs had no effect which is explained by the low surface levels of this protein on these cells (see above). Capping of its receptor CD46 upon cell-to-cell transmission of attenuated MV was observed earlier in cocultures of persistently infected monocytic U937 cells and uninfected HeLa cells as was formation of long microfilamental processes emanating from the donor U937 cells (Firsching et al., 1999). Because in both systems, receptor (CD150/CD46) or H protein neutralizing antibodies or fusion inhibitory peptide supplementation only partially prevented transmission, an active role of these extensions in viral transfer was suggested. This has recently been confirmed for HIV (Aggarwal et al., 2012), yet it is as yet unknown whether this also applies to MV transmission.

In common to that characterized for HIV, the potential MV VS was found to selectively concentrate activated LFA-1, DC-SIGN and CD81 all of which may act to stabilize the structure or to concentrate nascent viral particles (Felts et al., 2010; Chen, 2012; Koethe et al., 2012). Interestingly, phosphorylated ERM proteins also partitioned there which, in addition to their activity to link actin to PIP2-enriched membrane domains, may have an enhancing effect on MV uptake as revealed for moesin in non-T cells earlier (Schneider-Schaulies et al., 1995). In contrast, this process proved to be supported in neuronal and T cells by substance P receptor (Harrowe et al., 1992; Makhortova et al., 2007), and strikingly, this molecule was also enriched within the MV–DC/T cell interface (Koethe et al., 2012; Fig. 2). Thus, by repatterning receptors which stabilize the structure or have been implicated in MV uptake or enhancement thereof and viral proteins, the interface organized between MV-infected DCs and pre-activated T cells shares important similarities to the HIV VS to efficiently support MV entry into target T cells.

Figure 2.

Model for MV transmission from DCs to T cells.

A. Schematic representation of transfer of MV particles via filopodial extensions or synapse-like interfaces (bottom graph) with representative examples of both structures (first panel, MV H in red), and concentration of CD150 on T cells (transfected to express CD150-HA), or that of MV structural proteins (P or H) at the interfaces between pre-activated T cells and eGFP–MV-infected DCs (second to fourth panels).

B. Schematic model of the interface between MV-infected DCs and activated T cells (upper graph). Examples of cellular components (each in red) accumulating between eGFP–MV-infected DCs and T cells are shown below.

A and B. Nuclei are counterstained with DRAQ5; quantitative assessments of redistribution are shown in Koethe et al. (2012).

Conclusion and outlook

Interaction with DCs is the key to the understanding of major aspects of viral pathogenesis which extend from the control of magnitude and quality of immune responses to immune evasion or suppression, yet also trafficking to sites where they are passed to other motile cells for successful dissemination. At a cellular level, formation of organized membrane microdomains segregating receptors and receptor-proximal signalling complexes in a defined lipid/protein environment is crucial in regulating access to and compartmentalization within DCs, yet also for formation of stabilized interfaces with T cells to convey positive (antigen presentation) or negative (T-cell silencing) signals or virus (as particles or RNPs). Moreover, dynamic coupling to the actin (and presumably, tubulin) cytoskeleton are key elements in dynamic interaction of viruses with membranous compartments. As revealed for MV, viruses may actively regulate the lipid and thereby protein environment required for entry into DCs by activating cellular SMases via their attachment receptor DC-SIGN and, as a result, promoting vertical recruitment of CD150 which clusters together with DC-SIGN in ceramide-enriched domains. Because these tend to prevent co-recruitment of CD4 and chemokine receptors, extrafacial ceramide generation in response to DC-SIGN ligation may rather prevent HIV from accessing the cytosol and this may contribute for preferential trapping by these cells. It will be exciting to determine the importance of DC-SIGN-dependent SMase activation on the mode of uptake of its viral ligands directly, and if this could be modified on interference with this particular pathway. Because SMase activation can be accomplished on ligation of TNF-R family members also including CD40 or certain cytokine receptors, the micromilieu of pathogen–DC encounter might additionally impact on the formation of membrane microdomains important in trapping or uptake. Evidently, SMase activation has a direct impact on actin cytoskeletal remodelling and it will be highly interesting to see whether it, when catalysed by ligation of the respective receptors, takes an active part in regulating viral uptake and compartmentalization.

Similar to that for HIV, the interface between the MV-infected DC and autologous T cells appears to be a perfect transmission structure concentrating viral components, receptors mediating and possibly supporting entry, and stabilizing elements also including those regulating association with the actin cytoskeleton. It is particularly the latter whose role in transmission has been highlighted. Thus, it may regulate lateral segregation of receptors (as analysed in detail for the IS), or their containment by fencing, yet also provide surface extensions (filopodia, filopodial bridges, nanotubes or cytonemes) and thereby transfer structures enabling cell-to-cell transfer of virus particles. Whether this structure serves as transmission site and if, as likely the case, particles or RNPs are transmitted remain to be determined in high-resolution analyses as will be the role of substance P receptor and moesin in this setting.


We apologize to all our colleagues whose exciting work on this topic which due to space constraints we could not include into this review. We thank Jürgen Schneider-Schaulies, Bert Rima, Paul Duprex, Teunis Geijtenbeek, Stephan Becker, Larissa Kolesnikova and Erich Gulbins for helpful discussion and fruitful collaboration on the subject, and the Deutsche Forschungsgemeinschaft for financial support of our laboratory work.