EurALMF Program Report

Investigation of TNF-receptor GFP-chimera and other components of the NF-kB signalling pathway in living cells

DI. Dr. Johannes A. Schmid

Dept. of Vascular Biology and Thrombosis Research

Univ. Vienna

and Competence Center Bio-Molecular Therapeutics

Brunnerstr. 59, A-1235 Wien

Tel. (*43)-1-4277-62573, Fax: (*43)-1-4277-62550

Mail: Johannes.Schmid@univie.ac.at

Internet : http://www.univie.ac.at/VascBio/schmid/

Introduction

My interest was focussed on the investigation of the signalling pathway, which is initiated by the cytokine TNFa (tumour necrosis factor a). This cytokine exists both as membrane-bound and as soluble trimeric protein, serving as ligand for two different receptors (TNF-receptors 1 and 2, TNFR1, TNFR2), which have similar extracellular but distinct intracellular domains. Upon binding it is assumed to induce a trimerization or oligomerization of receptors, thereby initiating a signalling cascade (Fig. 1) resulting in the activation of the transcription factor NF-kB, as well as components of MAPK-pathways or apoptotic protease cascades. NF-kB and MAPK-dependent transcription factors induce the expression of inflammatory genes, like adhesion molecules that are essential for binding of leukocytes to the endothelium and transmigration from the blood circulation to the site of inflammation (for review see Mantovani et al., 1997). It is generally accepted that ligand induced clustering of TNF-receptors results in a corresponding clustering of adapter molecules on the cytoplasmic side of the receptor (Vandenabeele et al., 1995).

The intracellular domain of TNFR1 comprises a so called death domain (DD) that binds to corresponding adapter molecules with similar domains. One important adapter protein for TNFR1 is FADD (Fas-associated death domain protein), which leads to activation of caspase 8 and initiation of the apoptosis signalling cascade. Other important adapter molecules are RIP (Receptor interacting protein kinase) or TRADD (TNFa-receptor associated death domain protein), which are also able to activate the NF-kB signalling pathway. The pathway triggered by the adapter molecule TRADD is assumed to involve the interaction with other adapter proteins, such as TRAF´s (TNFa-receptor associated factors), which also serve as direct adapters for the second TNF-receptor, TNFR2. The current model assumes that oligomerization of adapter molecules (initiated by trimerization of receptors) leads to activation of serine/threonine kinases such as NIK (NF-kB inducing kinase) or MEKK1 (MAPK/ERK kinase kinase 1) and that these kinases initiate MAPK- and NF-kB pathways (Eder, 1997). The key event in activation of the transcription factor NF-kB is the phosphorylation, ubiquitinylation and degradation of the inhibitory molecule IkB, a process, which is triggered by specific kinases (IkB kinases, IKK1 and IKK2) that are components of a multimeric signalling complex (Malinin et al., 1997; Nemoto et al., 1998, Mercurio et al., 1997; Zandi et al., 1997; O´Conell et al., 1998). After degradation of IkB, NF-kB is released and translocates to the nucleus, where it activates inflammatory genes. Besides, it also upregulates the synthesis of its inhibitor, which represents an auto-regulatory feedback-inhibition of the NF-kB activity (de Martin et al., 1993).

General Aims

Although the major components that are involved in the course of TNF-receptor signalling from the cell surface to the nucleus have been identified, we still lack a detailed understanding of the dynamics and interactions of these signalling molecules in vivo.

An important aim of our group is to elucidate the dynamics of signalling events in living cells using fusion proteins of signalling components with distinguishable variants of GFP (cyan, green and yellow fluorescent proteins, CFP, GFP and YFP, respectively) and to investigate their interactions by the means of fluorescence resonance energy transfer (FRET) microscopy.

Specific Aims for the EurALMF Project

In the first stage, we wanted to investigate, whether the clustering of receptors (TNFR1 and TNFR2) and a consecutive clustering of the corresponding adapter proteins (e.g. TRAF1, TRAF2) can be observed in response to TNFa. If so, the dynamics of the association process should be investigated, as well as the question whether the signalling occurs exclusively on the cell surface, or also in endocytic compartments like early or recycling endosomes, after internalisation of receptor-ligand complexes. Since, it is not completely resolved how the signalling is shut off on the receptor and the adapter protein level, for instance whether it occurs by an internalisation and degradation of receptor or by a shedding of the extracellular domain from the cell surface, we aimed to use fluorescently-tagged agonistic antibodies against the ligand binding domain to study the de-activation process. In addition, we were interested to investigate the distribution of the receptors between cell surface, secretory and endocytic compartments and to test whether this distribution changes in response to TNFa. Moreover, potential functional interactions between TNFR1 and TNFR2 should be investigated by using CFP- and YFP-chimera of both receptors and FRET microscopy.

Since important new data on the intracellular dynamics of NF-kB and IkB were found by us just briefly before the start of the EurALMF program, we decided to slightly modify the specific aims  of the research program so that we could use the EMBL light microscopy facilities for these purposes, as well. The additional aim was to investigate the nucleocytoplasmic shuttling of different components of the NF-kB signalling pathway in living cells. Due to this additional interest, I had to reduce the experimental plan for the TNF-receptor project.

Work done in preparation for the EurALMF program

As a prerequisite for the project to be done at the EMBL Heidelberg, we cloned CFP-, GFP- or YFP-fusion proteins of various components of the NF-kB signalling pathway (TNFR1, TNFR2, TRAF1, TRAF2, NIK, RIP, IKK1, IKK2, p65-NF-kB and IkBa.). Moreover, we generated stable 293 or HeLa cell lines expressing different GFP-chimeras of p65-NF-kB, IkBa, TRAF1, TRAF2, IKK2, TNFR1 and TNFR2 or combinations thereof. The GFP-moiety of fusion proteins is usually not affecting the functionality of the protein, as shown in a large number of reports on different GFP-chimeras (Cubitt et al., 1995; Tsien, 1998). In fact, different functional assays (like reporter gene analysis) already proved the functional integrity of the GFP-chimera that we generated. The investigation of various GFP-fusion proteins of components of the signalling cascade  (TRAF1, TRAF2, NIK and IKK2) revealed already interesting findings on their localization and interactions in living cells.

Results

1. Multimerisation of TNFR- and TRAF-molecules by different stimuli

1.1. Studies with TNF-receptors

One of our major aims was to investigate whether a trimerization or multimerisation of GFP-linked TNFa-receptors can be visualized in living cells. For that purpose, we generated constructs containing a GFP-, CFP- or YFP-tag on the C-terminus of the receptor (at the intracellular domain). By that means the extracellular, ligand-binding domain should remain unaffected. However, it could not be excluded that the GFP-moiety interferes either with the oligomerization of the receptor or with binding of adapter proteins to the intracellular domain.

Nevertheless, reporter gene assays with TNFR1 or TNFR2-chimeras comprising a GFP tag indicated that they are still functional with respect to NF-kB activation. Moreover, overexpression of TNFR1-GFP fusion proteins clearly induced apoptosis, thus supporting the anticipated pro-apoptotic function. Morphological analysis of the chimeras revealed that a major part of the TNFR-proteins is located intracellularly in compartments of the secretory pathway as already previously described for wild type TNFR. Using high numerical aperture objectives (60x and 100x), we could clearly demonstrate incorporation of TNFR2-GFP chimeras into the cytoplasmic membrane, exhibiting a fine uniform lining of the cellular membrane (Fig. 2). Addition of TNFa to the culture medium  lead to the formation of cluster-like fluorescent dots in the cytoplasmic membrane within about 20 min (Fig. 3). These clusters indicate oligomerization of TNFR2 molecules on the cell surface. Based on the size of the dots, we assume that multimeres larger than trimeres are built, which might be the result of crosslinking of TNFa-induced receptor trimeres by intracellular adapter proteins. After prolonged incubation time (4 h), these clusters had disappeared and moreover not even the fine lining of the plasmamembrane was significantly visible, implying that TNFa-receptors might have been internalised or shed, after the ligand induced oligomerisation (Fig. 4). However, this TNFa-induced aggregation of TNFR2 was not always clearly detectable. This can be explained by the rather low affinity binding of soluble TNFa to the TNFR2, which binds only membrane-bound TNFa with higher affinity. Thus, at a later stage of our investigations, we used antibodies against TNFR2 to induce the crosslinking of receptors, resulting in a more pronounced multimerisation at the cell surface  (Fig. 5). For TNFR1 it was much more difficult to obtain clear morphological data, because overexpression of the receptor induced apoptosis in a majority of transfected cells. Moreover, the number of TNFR1 molecules per cell is very low (in the range of a few thousand), and the major part of the receptors resides in Golgi-compartments (Fig. 6). These general limitations prevented unambiguous morphological analysis of TNFa-induced oligomerisation of TNFR1.

 

1.2. Studies with TRAF1 and TRAF2

Based on the general model of TNFa induced signalling, we assumed that ligand induced receptor oligomerisation results in a corresponding oligo- or multimerisation of adapter molecules within the cell. Important members of these adapter molecules are TRAF´s (TNFa-receptor associated factors), which bind directly to TNFR2 and indirectly to TNFR1 (via TRADD, TNFR-associated death domain protein). Overexpression of TRAF1 or TRAF2 after transient transfection leads to the formation of large intracellular aggregates (Fig. 7). This inherent tendency to form clusters was more pronounced for TRAF2, which showed distinct multimeres even at the lowest DNA-amounts that we tested for transient transfection. To achieve more physiological levels we generated stable transfectants expressing either TRAF1, TRAF2 or both adapter molecules together. Under conditions of stable low-level expression TRAF1 exhibited a uniform cytosolic fluorescence (Fig. 8), whereas TRAF2 still showed in many cases distinct aggregates, although there was a diffuse cytosolic distribution visible, as well (Fig. 9). In some cells this diffuse distribution was predominant. Addition of TNFa to stable transfectants (expressing CFP-TRAF1 and YFP-TRAF2) and subsequent microscopy resulted in the formation of additional small fluorescent clusters within about 20 min (Fig. 10), which contained both TRAF2 and TRAF1 molecules. However, further detailed analysis showed that these clusters are also formed, when evaporation from the medium occurs during the imaging at 37°C. Subsequent experiments under conditions, which exclude evaporation problems showed aggregation events of TRAF molecules after addition of TNFa, as well. However, I could observe that similar aggregates occurred after prolonged imaging under CFP-excitation conditions (Fig. 11) meaning that either the UV-component of the excitation light induced the clustering of TRAF-molecules (which would be reasonable given that UV-light is able to activate the NF-kB signalling pathway) or that the excitation of CFP lead to a CFP-dependent aggregation. Since these observations were made rather at the end of the short term project at the EMBL, I could not clarify, which effect is finally inducing the aggregation of TRAF molecules. However, I think that the general property of TRAF molecules to form multimeres after a variety of different stimuli appears meaningful based on the fact that a number of completely different stimuli (TNFa, IL-1, UV-irradiation, virus-infection, cellular stress…) is able to activate the NF-kB pathway.

 

2. Nucleocytoplasmic shuttling

Preliminary experiments in our lab indicated that NF-kB and IkBa accumulate in the nucleus after addition of Leptomycin B, which blocks nuclear export processes. This indicates that NF-kB and IkBa shuttle constitutively between cytosol and nucleus and are transported to the nucleus in the absence of any NF-kB activating stimulus. This observation has a significant importance for our general understanding of the regulation of NF-kB activity. The classical model of NF-kB regulation suggests that this transcription factor is kept inactive by cytosolic sequestration due to binding to its inhibitor IkB, which masks the nuclear localization sequence of NF-kB. Upon signal-induced phosphorylation and ubiquitinylation of IkBa, the inhibitor is degraded and NF-kB is released so that it can translocate to the nucleus and bind to its cognate promoter sequences. The observation of a constitutive shuttling of NF-kB and IkBa between cytosol and nucleus implies a more dynamic model, where IkBa inhibits NF-kB activity primarily by preventing its binding to DNA. To further elucidate the dynamics of NF-kB, IkBa and other components of the NF-kB signalling pathway, we aimed to investigate the behaviour of these proteins as GFP-chimeras in living cells.

Using the EurALMF equipment, which is optimised for microscopy of living cells, I could

verify the nuclear accumulation of NF-kB/IkBa in the presence of Leptomycin B (Fig. 12). More detailed time lapse studies and quantification of nuclear versus cytosolic fluorescence were used to define the kinetics of this process (Fig. 13).

Moreover, studies using inhibition of nuclear export with Leptomycin B were extended to upstream signalling molecules. Interestingly, the NF-kB inducing kinase NIK showed nucleocytoplasmic shuttling with kinetics similar to that of NF-kB (Fig. 14), whereas other signalling molecules, such as IKK2 or TRAF2, did not show any significant nuclear accumulation within the time investigated.

Another aim was to verify the nuclear import of shuttling proteins without the use of Leptomycin B (and thus without addition of a chemical drug). For that purpose, I wanted to apply a bleaching strategy using the confocal laser microscope facility. The rationale was that repetitive bleaching of a nuclear area of a cell expressing a GFP-fusion protein, which appears cytoplasmic at steady state, but which shuttles continuously between cytosol and nucleus, would result in the loss of cytoplasmic fluorescence with time.

 

 

 

 

 

 

Indeed, experiments with NF-kB/IkBa-expressing cells and bleaching of a nuclear area resulted in the expected loss of cytosolic fluorescence. However, investigation of non-shuttling proteins under the same conditions caused a comparable loss of cytosolic fluorescence. This effect could be attributed to the unwanted bleaching of the cytosolic rim above and below the focal plane. Although, the signal, which generates the image with confocal microscopy, originates only from the focal plane, the exciting laser beam (and thus the bleaching effect) penetrates the whole height of the cell (Fig. 15B). Thus, this approach to study the nucleocytoplasmic shuttling of signalling molecules turned out to be inappropriate and was not further pursued. However, the problem might be overcome by applying 2-photon microscopy, where excitation-effective photon-pairs are just generated at the focal plane.

 

 

 

 

 

 

 

 

3. Optimisation of FRET microscopy techniques with CFP- and YFP-chimeras

Our group is generally interested in defining and characterising protein-protein interactions. One interesting and powerful technique to study molecular interactions is to make use of a quantum physical phenomenon named “Fluorescence Resonance Energy Transfer (FRET)”, which is only detectable, when two macromolecules with appropriate fluorophores are in close proximity to each other (in the range of 0 to 10 nm). In the last few years, GFP-variants were generated (e.g. CFP and YFP) that have fluorescence properties, which make them suited for FRET analysis, because the emission spectrum of CFP (the energy donor) overlaps with the excitation curve of YFP (the energy acceptor, see Fig. 16). Moreover, the inherent fluorescence of GFP-variants allows studying their behaviour in living cells. By that means, protein-protein interactions can be visualised and localised in vivo, which opens a wide field of potential scientific applications. There are several possibilities to monitor FRET by means of microscopy:

 

In the following section I will describe some of the arguments for the different FRET microscopy techniques:

a)      Imaging with FRET filter sets (acceptor emission at donor excitation) using limited exposure times for image acquisition (principle: see Fig. 17).
This techniques has several problems and limitations:
1. Signal cross talk from donor fluorescence
2. Some direct excitation of the acceptor at the donor excitation wavelength
3. Only sensitized acceptor fluorescence is used for the FRET signal
4. Expression-level dependent signal
5. Camera exposure time and gain are crucial
6. Valid controls are nearly impossible.

b)      Ratio imaging: Imaging of both the donor fluorescence and the acceptor fluorescence at donor excitation and division of the acceptor image by the donor image (Fig. 18).
The advantages of this method compared to the previous one are:
1. Better use of the FRET signal (donor fluorescence decrease and acceptor fluorescence increase)
2. Equal camera settings for both images give a black ratio image for negative controls, but a positive ratio image for clearly interacting proteins (using Omiga Optical filter sets XF88 and XF114).
However there are still important limitations:    
1. The FRET signal is expression-level dependent
2. Only external controls are possible
3. CFP- and YFP-chimeras have to have the same intracellular distribution, so that ratio images can be meaningfully calculated

c)      Measuring the donor bleaching kinetics: The kinetics of donor fluorophor destruction by photobleaching is slower in the presence of a FRET acceptor, onto which a part of the excitation energy is transferred (Fig. 19).
The advantages of this technique are:  
1. Results are concentration- or expression level independent
2. It is not necessary that donor and acceptor colocalize completely
3. FRET efficiencies can be calculated and thus distances between proteins.
The remaining limitations are:   
1. External control is necessary
2. Illumination has to be as constant as possible
3. Obtaining FRET-images is difficult

d)      Donor recovery after acceptor bleaching: Imaging of the donor fluorescence in the presence of a FRET acceptor, followed by (partial) destruction of the acceptor fluorophore and imaging of the donor fluorescence again shows an increase in donor emission (donor recovery, see Fig. 20).
This technique is quite powerful and has many advantages:      
1. It is concentration- or expression level independent
2. It is not necessary that donor and acceptor colocalize completely
3. Internal control is included
4. FRET-images can be obtained (ratio images or differential images before and after acceptor photobleaching).
However, calculation of FRET efficiencies is only possible, when the acceptor can be bleached completely without damage of the donor fluorophore.

e)      Fluorescence Lifetime Imaging Microscopy (FLIM): is based on the fact tat the donor fluorescence lifetime decreases significantly in the presence of a FRET acceptor. It is probably the most powerful technique to measure and visualize FRET in living cells. However, it requires complex and expensive equipment, which is usually not easily available.

 

Before starting the EurALMF project, I had already experience in applying the ratio imaging technique and I had done preliminary experiments with the donor bleaching technique. Knowing the limitations of the ratio imaging technique, which is primarily the question of fluorophore concentration (expression level of CFP- and YFP-fusion proteins), I aimed to work out better methods to monitor FRET between CFP- and YFP-fusion proteins. However, initial attempts to use the technique, which is based on the donor fluorescence recovery after acceptor bleaching, were not successful. Using paraformaldehyde-fixed positive control cell, expressing covalently linked CFP and YFP (eCYFP), I could not observe a significant increase in donor fluorescence after bleaching of YFP with the 514 nm line of the Argon-laser. Therefore, I intended to apply the other method, which uses the kinetics of donor bleaching to determine the FRET effect. (Just recently, I managed to employ and optimise the donor recovery technique for CFP/YFP-pairs and I noticed that it is necessary to use living, unfixed cells for that method).

However, the technique to determine FRET by measuring the kinetics of donor bleaching turned to work both with living and with fixed cells (after 3.5% paraformaldehyde-fixation), although it was obvious that the bleaching kinetics was much faster for fixed cells.

Donor bleaching was done either with a conventional illumination system using the CFP-filter set or with a LSM510 confocal laser scanning microscope (and the 458 nm line, which is used for CFP-excitation). Time series of about 30 images were taken and the decrease in donor fluorescence was quantified for selected regions of the cell (using the “Measure All” algorithm of the NIH-Image “Measure” macro).

I could verify that the donor bleaching kinetics approach can be carried out with the confocal microscope, as well. Since bleaching of regions of interest are easily done with the LSM510 equipment, I intended to optimise the donor bleaching method by including an internal control, where the acceptor is bleached in a control region before starting the donor bleaching time series.

The rationale was that a destruction of the acceptor fluorophore in a region of interest should result in a faster kinetics of the subsequent donor bleaching in that region compared to the surrounding areas, where the intact FRET acceptor would slow down the donor bleaching process. This concept could be experimentally confirmed with samples of fixed cells expressing the CFP-YFP-fusion protein (eCYFP) or CFP and YFP separately (Fig. 21). However, a more detailed analysis of the bleaching kinetics obtained with the confocal LSM510 equipment showed that it could not be fitted by a single exponential decay algorithm (as with normal conventional bleaching (Fig. 22A), but rather by a double exponential algorithm (Fig. 22B). Exact analysis of the fluorescence in the acceptor-bleached and the non-bleached region revealed that the acceptor bleaching in the region of interest was partially reversible, which resulted in the gradual re-appearance of FRET acceptor fluorophores (and thus probably in a change of the donor bleaching kinetics in that area). Another problem is that the wavelength used for excitation of CFP is suboptimal with the normal LSM510 (458 nm instead of 436 nm). At that wavelength a co-excitation (and bleaching) of the acceptor occurs also in the regions, where only donor bleaching should be achieved (Fig. 23). This might result in an alteration of the donor bleaching kinetics, as well. Thus, at least for fixed cells, it is not feasible to use an internal acceptor-bleached control region, when doing donor bleaching kinetic analysis. However, using living cells and a better excitation of CFP (e.g. with conventional illumination and specific CFP-filters), it might be possible to incorporate an internal control (with bleached acceptor fluorophores) in donor bleaching kinetic analysis.

In addition to experiments for image acquisition in FRET microscopy, I worked out a strategy (in collaboration with Jens Rietdorf from the EurALMF) to convert donor bleaching kinetics data to an apparent FRET image on a single pixel basis. The idea was to perform a simple algorithm for a single exponential decay fitting for every single pixel and to calculate an image from these data. Jens Rietdorf created a macro (for NIH-Image) for this purpose, which might be further optimised to generate real FRET efficiency images. It is planned to proceed with this project, because analysis of donor bleaching kinetics gives the most accurate estimations for FRET efficiencies.

 

Some general considerations when planning a EurALMF project

 

References

·        Cubitt, A. B., R. Heim, S. R. Adams, A. E. Boyd, L. A. Gross, and R. Y. Tsien. 1995. Understanding, improving and using green fluorescent proteins. Trends-Biochem-Sci. 20:448-55.

·        de Martin, R., B. Vanhove, Q. Cheng, E. Hofer, V. Csizmadia, H. Winkler, and F. H. Bach. 1993. Cytokine-inducible expression in endothelial cells of an I kappa B alpha-like gene is regulated by NF kappa B. EMBO J. 12:2773-9.

 

·        de Martin, R.; Schmid, J.A. and Hofer-Warbinek, R. 1999. The NF-kB/Rel family of transcription factors in oncogenic transformation and apoptosis. Mutation Res. 437, 231-243.

·        Eder, J. 1997. Tumour necrosis factor alpha and interleukin 1 signalling: do MAPKK kinases connect it all? Trends-Pharmacol-Sci. 18:319-22.

·        Malinin, N. L., M. P. Boldin, A. V. Kovalenko, and D. Wallach. 1997. MAP3K-related kinase involved in NF-kappaB induction by TNF, CD95 and IL-1. Nature. 385:540-4.

·        Mantovani, A., F. Bussolino, and M. Introna. 1997. Cytokine regulation of endothelial cell function: from molecular level to the bedside. Immunol-Today. 18:231-40.

·        Mercurio, F., H. Zhu, B. W. Murray, A. Shevchenko, B. L. Bennett, J. Li, D. B. Young, M. Barbosa, M. Mann, A. Manning, and A. Rao. 1997. IKK-1 and IKK-2: cytokine-activated IkappaB kinases essential for NF-kappaB activation [see comments]. Science. 278:860-6.

·        Nemoto, S., J. A. DiDonato, and A. Lin. 1998. Coordinate Regulation of IkB kinases by Mitogen-Activated  Protein Kinase Kinase Kinase 1 and NF-kB-Inducing Kinase. Mol-Cell-Biol. 18:7336-43.

·        O´Connell, M. A., B. L. Bennett, F. Mercurio, A. M. Manning, and N. Mackman. 1998. Role of IKK1 and IKK2 in lipopolysaccharide signalling in human monocytic cells. J-Biol-Chem. 273:30410-4.

·        Schmid, J.A.; Birbach, A; Hofer-Warbinek, R.; Pengg, M.; Burner, U.; Furtmüller, P.G.; Binder, B.R.; and de Martin, R. Dynamics of NF-kB and IkBa studied with GFP-fusion proteins. Investigation of the GFP-binding to DNA by fluorescence resonance energy transfer (FRET). J. Biol. Chem. in press

·        Tsien, R. Y. 1998. The Green Fluorescent Protein. Annu-Rev-Biochem. 67:509-44.

·        Vandenabeele, P., W. Declercq, R. Beyaert, and W. Fiers. 1995. Two tumour necrosis factor receptors: structure and function. Trends-Cell-Biol. 5:392-9.

·        Zandi, E., D. M. Rothwarf, M. Delhase, M. Hayakawa, and M. Karin. 1997. The IkappaB kinase complex (IKK) contains two kinase subunits, IKKalpha and IKKbeta, necessary for IkappaB phosphorylation and NF-kappaB activation. Cell. 91:243-52.