In the present work we show that p38 is activated by mitogens and cellular stress in the same cell line, but that the signalling pathways differ. We suggest that p38 plays a dual role in cell cycle control in fibroblasts mediating cell cycle progression or cell cycle arrest depending on the extracellular stimulus. In NIH3T3 cells, p38-dependent cell cycle arrest either due to cellular stress or constitutive activation by overexpression of the kinase itself or an upstream activating kinase has been demonstrated in various publications [7–10]. The fact that in the present study in the same cell line stimulation with serum or growth factors results in phosphorylation and activation of p38, and - vice versa - FCS- and growth factor-induced DNA-synthesis is blocked by the p38-specific inhibitor SB203580 or siRNA-mediated knock-down of p38 clearly indicates that p38 is also required for proliferation and points to a dual role of p38 in cell cycle regulation. Our data suggest that a key element might be the duration and/or amount of activation which then leads to different downstream signalling. While anisomycin-exposure leads to a strong and sustained activation of the p38 and MK2 kinases, thereby increasing the phosphorylation of CREB [27, 28], mitogen-induced activation of p38 is weaker and transient, but is required for cyclin D1 expression. In addition, we demonstrate that p38 is differentially regulated in response to anisomycin and mitogens with respect to the involvement of small GTPases (Figure 6).
Dual regulation of a kinase depending on the extracellular stimulus has been proposed in the neuronal cell line PC12 for ERK [for review see ]. In these cells, EGF leads to a transient activation of ERK with cytoplasmic retention resulting in a proliferative cellular response. In contrast, NGF-induced ERK activation in the same cells is sustained and accompanied by a nuclear translocation causing cell cycle arrest and neuronal differentiation.
To better understand the role of p38 in mitogen-induced proliferation, we studied cell cycle proteins, i.e. phosphorylation of pRB and expression of the cyclins D1 and A. It is generally accepted that during G1-phase, cyclin D1/Cdk4 and downstream cyclin E/Cdk2 phosphorylate pRB which then dissociates from the transcription factor E2F allowing transcription of S-phase specific genes, such as cyclin A, and thereby entry into S-phase . Since the mitogen-induced expression of cyclin D1 was strongly reduced in the presence of SB203580 we conclude that expression of cyclin D1 requires the activity of p38. An expected consequence of a decrease in cyclin D1 is less activity of the cyclin D1/Cdk4 complex and in turn less phosphorylation of pRB. Hence, the observed attenuation of pRB phosphorylation in the presence of SB203580 is very likely due to decreased cyclin D1 expression. Downregulation of cyclin A by SB203580 was observed at a later time point, i.e. 14 h after mitogenic stimulation. According to the kinetics of pRB phosphorylation we assume that this time point correlates with early S-phase. We therefore conclude that the decrease in cyclin A is not directly mediated by p38, but rather a consequence of inhibition of pRB phosphorylation by SB203580.
The mechanism of p38-dependent expression of cyclin D1 in our cell system is not known so far. In melanoma cells, p38-ATF-2-dependent expression of cyclin D1 in response to hepatocyte growth factor/scatter factor has been described . Induction of cyclin D1 by pp60v-src
is also mediated via the p38/JNK-ATF-2/CREB pathway in human breast cancer cells . Since we did not detect p38-dependent phosphorylation of ATF-2 nor CREB, an involvement of ATF-2 or CREB in cyclin D1 expression in our cell system is unlikely. One possible explanation comes from the observation, that ERK1/2 phosphorylation is blocked from 6 h on after FCS-stimulation, very likely as a secondary effect of p38 inhibition. Hence, p38 activity seems to be required for sustained ERK1/2 phosphorylation. In fibroblasts, sustained ERK1/2 activity is required for cyclin D1 expression, especially during mid-G1-phase . The underlying mechanism of this cross-talk between p38 and ERK1/2 remains to be elucidated.
To our surprise, inhibition of p38 function by SB203580 did not only block mitogen-induced G0/G1-S transition, but also attenuated continuous proliferation of NIH3T3 cells. This observation is in contrast to data obtained in BJ primary fibroblasts and WI-38 fibroblasts. In these cells, SB203580 does not alter proliferation in exponentially growing cultures, which show doubling rates comparable to our NIH3T3 [36, 37]. However, cell type specific differences might explain different functions of p38 in NIH3T3 cells. Very recently, it was shown that the transcription factor FoxM1 acts downstream from the Ras-MKK3-p38 pathway in NIH3T3 cells . Importantly, FoxM1 is also known to regulate a number of proliferative genes [38; manuscript in preparation]. Although we have not tested, another explanation for the discrepancies could be p53 function. While BJ and WI-38 fibroblast express wild-type p53 [39, 40], the NIH3T3 cells we used are p53-deficient (unpublished observations).
Our observation of a sustained p38 activation after anisomycin- or sorbitol-treatment is in accordance with other reports showing persistent activation of p38 after cellular stress, e.g. in C3H10T1/2 cells in response to anisomycin or UV  or in several cell lines in response to γ-irradiation, genotoxic compounds or during premature senescence [[42, 43], reviewed in ].
We have also shown that sustained activation of p38 is required for contact-inhibition in murine and human fibroblasts . Several mechanisms explaining p38-dependent cell cycle arrest have been described. For instance, p27KIP1, a well-known inhibitor of Cdk2 and Cdk4, is one important downstream target of p38 upon contact-inhibition [12, 44]. The protein p27KIP1 is also upregulated in response to genotoxic agents and here is supposed to be crucial for maintenance of cell cycle arrest . However, we did not observe accumulation of p27KIP1 in response to anisomycin or sorbitol (unpublished oberservation). It is also known that the Cdk inhibitors p21WAF1/CIP1 and p16INK4a mediate p38-dependent senescence, for instance in response to DNA-damaging agents and reactive oxygen species, which might be related to the role of p38 as a tumour suppressor [42, 45–51]. Very recently, p38-dependent induction of p21 due to sorbitol-treatment has been described in nucleus pulposus intervertebral disc cells . Whether p21WAF1/CIP1 or p16INK4a are upregulated in response to anisomycin or sorbitol in fibroblasts needs to be determined.
Moreover, we observed sustained phosphorylation of CREB. Two kinases are known to phosphorylate this transcription factor: MK2 [27, 28] and MSK1 . Since MK2 was also persistently activated in response to anisomycin, we conclude that MK2 at least partially contributes to phosphorylation of CREB. A possible involvement of MSK1 remains to be elucidated. Interestingly, cell cycle arrest due to sustained activation of the cAMP/CREB-pathway was also detected in prostate carcinoma cells, which were chronically exposed to pituitary adenylate-cyclase-activating polypeptide. In contrast, transient stimulation of cAMP/CREB induces proliferation . Constitutive activation of CREB by bacterial toxins leads to G1-arrest in a murine macrophage cell line by induction of p27 and downregulation of cyclin D1 . In accordance, cholera toxin, a potent inducer of cellular accumulation of cAMP and thereby phosphorylation of CREB, is able to cause G1 arrest by upregulation of p27, p21 and downregulation of cyclin D1 in rat and primary human glioma cells . However, we could not reverse downregulation of cyclin D1 in response to sorbitol-exposure by SB203580 arguing that p38 is not the sole entity responsible for the decrease in cyclin D1. This is in line with the observation that arsenite-induced downregulation of cyclin D1 in NIH3T3 cells cannot be restored by SB203580  and that cyclin D1 decrease can also be mediated by JNK . Hence, the precise function of the p38-MK2-CREB axis in anisomycin- or sorbitol-induced cell cycle arrest remains to be determined.
Phosphorylation of CREB in response to mitogenic stimulation could not be blocked by SB203580, which is in line with previous observations that CREB phosphorylation in response to growth factors is mediated by the ERK pathway .
Interestingly, phosphorylation of ATF-2 was not inhibited by SB203580, although it has been identified to be an excellent substrate for p38 in vitro . This observation is in perfect accordance with the work of Hazzalin and coworkers  and the work by Maher  ruling out involvement of p38 in phosphorylating endogenous ATF-2 in fibroblasts. Since phosphorylation of ATF-2 could be blocked by pharmacological inhibition of JNK, phosphorylation of ATF-2 is very likely dependent on JNK .
Persistent activation after anisomycin is consistent with the supposed mechanism of action: anisomycin inhibits protein synthesis hence blocking transcription of phosphatases. As a result, p38 dephosphorylation does not occur in the presence of anisomycin. Furthermore, dephosphorylation of the upstream acting MKK3/6 is inhibited thereby allowing prolonged activation of p38. A similar mechanism has been described for arsenite-induced JNK activation [58, 59]. To gain more insight into upstream events regulating p38 activity in response to mitogens, we identified the involvement of small GTPases and made use of selective bacterial toxins. Clostridium sordellii Lethal Toxin (LT) inhibits Ras, Rac, and Rap function, Clostridium difficile Toxin B abolishes Rac, Cdc42, and Rho function, and Botulinus C3 exoenzyme, in our study used as C2IN-C3 fusion toxin displaying high cell permeability , selectively inhibits Rho function [29, 30]. The observation, that ERK activation was selectively abolished only in the presence of Lethal Toxin indicates selectivity of the toxins. In control experiments with Botulinus C2 toxin, which ADP-ribosylates actin , we ruled out that the observed inhibitory effects of the toxins occurred unspecifically due to degradation of the actin cytoskeleton (unpublished observation). If anisomycin-induced p38 activation is due to its suppression of phosphatases (see above) it should be independent of Rho proteins. Indeed, blocking the activity of the small GTPases Rho, Rac, and Cdc42 by preincubation with Lethal Toxin or Toxin B, had no effect on anisomycin-induced p38 phosphorylation.
On the contrary, serum-induced p38 activation could be blocked by preincubation with Toxin B and Lethal Toxin. In view of the fact that C2IN-C3 had no effect on serum-induced p38 phosphorylation, the results strongly argue against an involvement of Rho and point to a potential role of Rac and/or Ras and Cdc42. Indeed, in overexpression studies, Rac and Cdc42 have been identified to mediate p38 activation [10, 61, 62]. More detailed analysis is required to identify which of the small GTPases is involved in p38-mediated control of proliferation.