In order to determine at what stage of the differentiation process relocalization of centrosome proteins occurs, we performed cell culture of mouse C2C12 myoblasts and triggered differentiation by serum withdrawal. Alternatively, we used H-2Kb
-tsA58 mouse myoblasts, carrying a thermolabile T-antigen and allowing differentiation upon temperature shift from 33 to 37°C [11
]. In the undifferentiated state, the centrosomal proteins of both cell lines were found in a single focus within the pericentriolar material adjacent to the nucleus, while the protein PCM-1 ("pericentriolar material protein 1") localized to multiple 'centriolar satellites', as described by [13
] and [14
]. One day after differentiation of myoblasts was triggered, we observed centrosome proteins at the nuclear periphery already at the single cell stage, prior to fusion into myotubes (Fig. ).
Figure 1 Relocalization of centrosome proteins to the nuclear surface in differentiating myoblasts. Cultures of C2C12 myoblasts were induced to differentiate by serum starvation for one day. At this stage, the culture contains undifferentiated myoblasts (u), as (more ...)
We investigated the pattern of various proteins and found that significant amounts of pericentrin and PCM-1 accumulated at the nuclear periphery, whereas the centrosome protein cdk5rap2 showed partial relocalization. Only minor amounts of gamma-tubulin were found at the nuclear periphery, consistent with data from [10
], and the protein ninein was not found to relocalize to the nuclear periphery in our experiments.
A closer look at differentiating C2C12 cells revealed that at the single cell stage, proteins such as pericentrin and PCM-1 started to accumulate at the nuclear periphery in the proximity of the centrosome, whereas areas of the nuclear envelope distal from the centrosome showed less centrosome protein enrichment (Fig. , first two rows). At the same time, remnant pericentrin and cdk5rap2 were still visible at the centrosome, and PCM-1 was still partly localized in pericentriolar satellites (Fig. ). The early relocalization of these proteins from the centrosome to nearby sites of the nuclear surface raises the question as to whether direct transfer from the centrosome is involved, or whether the proteins are recruited to the nuclear surface from a soluble cytoplasmic pool. Both scenarios seem possible: direct transfer from the centrosome to the nucleus might occur by diffusion. Alternatively, at the onset of differentiation, proteins such as pericentrin might be recruited from a soluble pool and localize to the nuclear surface next to the centrosome due to minus-end directed transport along microtubules [15
], since the microtubule-organizing centre in these myoblasts is strongly focused at the centrosome and would thus favour deposit of pericentrin in the surrounding area (Fig. , cell marked with 'u'). It is noteworthy, however, that already at the single cell stage microtubules were reorganized into a sun-like array radiating from the nuclear periphery within those cells displaying perinuclear relocalization of centrosome proteins (Fig. , cell marked with 'd').
Figure 2 Reorganization of microtububules and centrosome protein in cells expressing myogenic differentiation markers. Culture of mouse myoblasts containing undifferentiated (u) cells, and cells that started to differentiate (d). The centrosome protein PCM-1 is (more ...)
To verify that these cells had indeed entered the differentiation programme, we performed immunofluorescence microscopy with differentiation markers. We noticed that all cells in which the centrosome protein PCM-1 had relocalized to the nuclear surface showed cytoplasmic expression of embryonic myosin, an isoform of myosin that is specifically expressed upon onset of muscle cell differentiation (Fig. ). Furthermore, these cells expressed the differentiation marker myogenin, a transcription factor that localizes to the nucleus of differentiating cells (Fig. ). This marker was absent from cells showing pericentriolar PCM-1 staining. Conversely, these cells with pericentriolar PCM-1 were the only ones that stained positively for the proliferation marker Ki-67 in the nucleus, whereas nuclei in cells with relocalized PCM-1 did not contain Ki-67 (Fig. ). Once relocalization of PCM-1 occurred, it persisted until the final stages of differentiation: We detected PCM-1 around the nuclear surface after fusion of multiple C2C12 cells into myotubes (Fig. ). Moreover, nuclei in muscle from adult mice showed comparable staining of PCM-1 at their surface (Fig. ).
Figure 3 Perinuclear localization of centrosome proteins persists after fusion of myoblasts into myotubes, and in adult mouse muscle. (A) Myotube of fused C2C12 cells in culture. Immunofluorescence of PCM-1 (red); DNA is stained in blue. (B) Cryosection through (more ...)
We next sought to determine whether centrosome proteins that had assembled around the nucleus were bound at the inside or at the outside of the nuclear envelope. For this purpose, we transfected mouse myoblasts with GFP-tagged lamin A, a protein of the inner nuclear envelope. Deconvolution microscopy revealed that the centrosome protein PCM-1 localized in a rim slightly outside of lamin A (Fig. ), suggesting that PCM-1 associates with the outer surface of the nuclear envelope. Because the PCM-1 staining at the nuclear surface was visible in small clusters, we tested whether these sites co-localized with nuclear pores. However, double immunofluorescence with an antibody against a family of nuclear pore complex proteins indicated that most of the PCM-1 clusters were distinct from nuclear pores (Fig. ). To reveal further structural details, we performed immuno-electron microscopy of ultra-thin cryosections of differentiated C2C12 cells. PCM-1 was detected with antibodies and gold-coupled protein A. This technique preserved cellular membranes and showed that PCM-1 localized mainly outside the nucleus, away from the outer nuclear membrane (Fig. ). Consistent with our immunofluorescence data, immunogold labelling of PCM-1 was often seen in clusters (Fig. , top right). The gold was mostly seen along grey electron-dense material surrounding the nuclear surface (Fig. , arrows). This electron-dense material had a thickness between 30 and 40 nm, suggesting that was part of a tight matrix.
Figure 4 The centrosome protein PCM-1 localizes to dense structures on the cytoplasmic site of the nuclear envelope. (A) Deconvolved image of a nucleus from a differentiated H-2Kb-tsA58 cell, expressing GFP-lamin A (green), and stained for PCM-1 (red) and DNA (more ...)
To test for the biochemical behaviour of perinuclear PCM-1-containing material, we purified nuclei from differentiated C2C12 cells that had fused into myotubes. These were separated from non-differentiated myoblasts by short treatment of cultures with trypsin, leading to selective enrichment of the differentiated cells. Following fractionation of cells, we performed extraction of purified nuclei with buffer containing various concentrations of salt, detergents, or urea. We found that perinuclear PCM-1 localization was largely resistant to 1.5 M NaCl, to treatment with 1% Triton X-100, or even to extraction with 6 M urea (Fig. ). However, PCM-1 was efficiently removed from the nuclear surface in 8 M urea, or in buffer containing the denaturing detergent SDS (sodium dodecyl sulfate) (Fig. ). Nuclei treated with SDS lost their integrity even after short times of treatment, and could therefore not be quantified reproducibly. Altogether, these results indicate that PCM-1 is part of an insoluble matrix in differentiated muscle cells. Taking into account that several centrosome proteins form a fibrous meshwork surrounding the centrioles in undifferentiated cells [16
], it seems plausible to assume that a similar fibrous meshwork forms around the nucleus during differentiation. Proteins such as pericentrin that have large predicted alpha-helical domains, and the potential to form coiled-coil interactions, might contribute to the formation of such a meshwork.
Figure 5 The centrosome protein PCM-1 is part of a detergent-resistant perinuclear matrix in differentiated muscle cells. (A) Purified nuclei from differentiated C2C12 cells were incubated for 10 minutes in buffer, containing as indicated, 0.1% Triton X-100, or (more ...)
To further characterize the nature of the perinuclear meshwork of centrosome proteins, we designed an in-vitro-approach to determine whether its formation was reversible in differentiated C2C12 cells upon re-entry into the cell cycle. Because muscle cells become postmitotic after differentiation, as shown by the lack of Ki-67 staining (Fig. ), cultures of differentiated C2C12 cells could not simply be driven into mitosis. To circumvent this problem, we isolated nuclei from differentiated cells and incubated them in cycling egg extracts from Xenopus laevis (Fig. ). These extracts are capable of mimicking cell cycle events such as S-phase and mitosis in vitro, and assemble mitotic spindles around various sources of DNA, including exogenously added nuclei (Fig. ).
Figure 6 The perinuclear matrix of centrosome proteins disassembles in mitotic egg extract. Purified nuclei from differentiated C2C12 cells were incubated in Xenopus egg extract and driven through the cell cycle in vitro. (A) Nuclei in extract in S-phase. Left: (more ...)
We noticed that upon entry of these extracts into a prometaphase-like state, C2C12 nuclei began to disassemble. DNA condensed into mitotic chromosomes, and the perinuclear 'matrix' of PCM-1 disintegrated into a system of interwoven fibres, and finally disassembled into protein aggregates of varying sizes (Fig. ). We performed deconvolution microscopy on nuclei, stained with various markers, in S-phase and prometaphase. We found that PCM-1, pericentrin, gamma-tubulin, as well as the nuclear envelope protein nesprin 1 localized to patches on the nuclear envelope that were closely apposed, but without fully co-localizing with each other (Fig. , left column). In samples that had entered prometaphase, we found that PCM-1, pericentrin, and nesprin 1 localized to spots and interconnected fibres that co-localized partly, whereas gamma-tubulin was found more diffusely distributed (Fig. ). We therefore believe that PCM-1 and pericentrin form distinct fibrillar structures at the outer nuclear surface that are connected at various contact sites, thus constituting a tight matrix. This matrix may be structurally equivalent to the fibrous pericentriolar material in undifferentiated cells, and it can be disassembled or at least loosened upon entry into mitosis. Consistently, many centrosome proteins, including PCM-1 and pericentrin, are seen in undifferentiated cells during mitosis in a wide crescent-shaped area at the spindle poles or diffuse in the cytoplasm, whereas in interphase they are more focused at the centrosome [14