Cell Density-Dependent Increase in Tyrosine-Monophosphorylated ERK2 in MDCK Cells Expressing Active Ras or Raf

The extracellular signal-regulated kinase (ERK) is one of the principal hub proteins that transmit growth signals from upstream oncogene products including Ras and BRaf to downstream effector proteins. However, there are both reports supporting and refuting the increase in ERK activity in cancer tissues expressing the active Ras and BRaf proteins. We considered that the cell density might account for this discrepancy. To examine this possibility, we prepared Madin-Darby canine kidney (MDCK) cells that expressed an active HRas, NRas, KRas, or BRaf and an ERK biosensor based on the principle of Förster resonance energy transfer (FRET). As we anticipated, expression of the active Ras or BRaf increased ERK activity at low cell densities. However, the ERK activity was markedly suppressed at high cell densities irrespective of the expression of the active Ras or BRaf. Western blotting analysis with Phos-tag gel revealed the decrease of tyrosine and threonine-diphosphorylated active ERK and the increase of tyrosine-monophosphorylated inactive ERK at high cell density. In addition, we found that calyculin A, an inhibitor for PPP-subfamily protein serine/threonine phosphatases, decreased the tyrosine-monophosphorylated ERK. Our study suggests that PPP-subfamily phosphatases may be responsible for cell density-dependent ERK dephosphorylation in cancer cells expressing active Ras or BRaf protein.


Introduction
Ras mutation is found up to 30% of human cancer patients [1]. Raf is one of the three major effectors of Ras and is also mutated frequently in human cancers [2]. The extracellular signalregulated kinases (ERK1/2; MAPK3/1) are considered the canonical terminus of the Ras-Raf branch, from which signals are dispatched to a number of proteins with different functions [3]. In agreement with these facts, an increase in phosphorylated active ERK (pERK) has been reported in a number of cancer tissues [4,5]. However, there are also reports claiming that pERK is not necessarily elevated in cancers harboring Ras and Raf mutations [6,7]. The failure to detect elevated pERK in Ras-or Raf-transformed cells may be ascribable to adaptation to the constitutively-active signals [8,9], or to technical pitfalls of immunohistochemistry [10]. It should also be recalled that many paradigms of oncogene signaling have been established by using rapidlygrowing tissue culture cells, which may be markedly different from cancer cells in patients.
One of the marked differences between in vitro and in vivo cellular milieus is cell density. In contrast to tissue culture cells, which are often seeded at low cell densities to promote cellular replication, in vivo cancer cells grow mostly in a high cell density environment. It has been well established that inhibition of cell proliferation occurs at high cell density; this phenomenon is known as contact inhibition of cellular growth or simply contact inhibition [11,12]. In non-transformed fibroblasts [13], epithelial cells [14], and vascular endothelial cells [15], cellto-cell contact causes downregulation of ERK and a subsequent decrease in cyclin D1. On the other hand, the loss of contact inhibition is a hallmark of cancer cells in vitro [16]. Cells infected by oncoretroviruses or transfected with oncogenes exhibit morphological changes and uncontrolled cell growth even at high cell density [17][18][19]. Many oncogene products exert their effect through activation of the Ras-Raf-ERK pathway; therefore, we can speculate that constitutive activation of Ras or Raf and the resulting ERK activation may contribute to the loss of contact inhibition of cancer cells. However, it has not been examined whether Ras or Raf activation is sufficient to activate ERK at high cell density.
The development of biosensors based on Förster resonance energy transfer (FRET) has opened a path to the analysis of cellular heterogeneity and temporal changes of the activities of signaling molecules in vitro and in vivo [20,21]. For the measurement of ERK activity, we generated an intramolecular (unimolecular) FRET biosensor named EKAREV, which consists of a donor fluorescent protein CFP, an ERK substrate peptide derived from Cdc25, an optimized linker, a FHA1 phosphate binding domain, and an acceptor fluorescent protein YFP ( Fig 1A) [22,23]. Activated ERK phosphorylates the substrate peptide and induces intramolecular binding of the FHA1 domain to the phosphorylated peptide, thereby bringing the two fluorescent proteins in close proximity to evoke FRET. The FRET biosensor is reversed to the pre-phosphorylation state by protein serine/threonine phosphatases (PSPs). The halflife of active ERK is approximately 30 seconds, which is slow enough to be monitored by the FRET biosensors [24]. Thus, by measuring the fluorescence intensities derived from FRET and CFP (FRET/CFP ratio for brevity), we can obtain spatiotemporal information of the activity balance between ERK and PSPs in living cells.
The EKAREV FRET biosensor has already been used in many studies. For example, we and others have found that ERK can be stochastically activated in tissue culture cells [25,26], and such stochastic ERK activation can be propagated to neighboring cells both in vitro and in vivo [25,27]. Here, by using EKAREV, we examined the effects of oncogenic Ras and BRaf proteins on ERK activity in Madin-Darby canine kidney (MDCK) cells cultured at high cell density. We found that, at high cell density, ERK activity was markedly decreased even in the Rasor BRaf-expressing MDCK cells. This suppression appears to be caused by the activation of PSPs and resulting accumulation of inactive tyrosine-monophosphorylated ERK. Our observations imply that the discrepancy between the expression of active Ras or BRaf and ERK activity in cancer tissues might reflect the difference in cell density-dependent activity of PSPs.

Image processing
Metamorph software (Molecular Devices, Sunnyvale, CA) was used for background noise subtraction and image analysis. Background intensities were determined by using an empty culture dish with the same amount of media. After background subtraction, the FRET/CFP ratio images were represented in the intensity-modulated display (IMD) mode. In the IMD mode, eight colors from red to blue are used to represent the FRET/CFP ratio, with the intensity of each color indicating the mean intensity of FRET and CFP channels. Heatmap images representing the relationship between the FRET/CFP ratio and time course were generated using MATLAB software (MathWorks, Natick, MA).

Statistics
Statistical analyses were conducted using R software (ver. 3.1.3). In the beeswarm plot, a bold line indicates the median, and fine lines, if any, denote the upper and lower quantiles, respectively. To compare two sets of data, Student's t test or Welch's t test was performed according to a result of F test. As a multiple comparison test, Tukey's honestly significant difference test or Steel-Dwass test was used after validation of homoscedasticity by Bartlett's test. In all tests, values of p < 0.05 were considered significant.

Establishment of MDCK cells expressing an active Ras or BRaf protein and a FRET biosensor for ERK activity
For the measurement of ERK activity in living cells, a FRET biosensor for ERK, EKAR-EV-NLS, was expressed in the nucleus of MDCK cells. The 4OHT-inducible Cre-dependent expression system was employed to minimize the inter-clonal variation (Fig 1A and 1B). Active mutants of four oncogene products, HRasG12V, NRasG12V, KRasG12V, and BRafV600E, were designed to be expressed in MDCK cells. Induction of each oncogene product was indeed confirmed by Western blotting analysis (Fig 1C).

Cell density-dependent suppression of ERK activity in MDCK cells expressing an active Ras or BRaf protein
To examine the effect of the cell density on ERK activity, we first plated the MDCK-EKAR-EV-NLS cells without induction of oncogene at increasing cell densities. In FRET/CFP ratio images acquired before induction of NRasG12V, as reported previously [25,26], ERK activity was heterogeneous at low cell density (Fig 2A). A timelapse video and a heatmap of other preinduction cell line, MDCK-EKAREV-NLS-floxed-tdTomato-KRasG12V cells, also showed this heterogeneity and it was generated by stochastic ERK activity pulses (Fig 2B left).
However, at high cell density, both the basal ERK activity and the stochastic ERK activity pulse were suppressed (Fig 2B right, S1 Movie). We next induced KRasG12V by 4OHT and analyzed their activities at low and high cell densities (Fig 2C, S2 Movie). At low cell density, KRasG12V cells induced robust ERK activation and abolished ERK activity pulses (Fig 2C left). However, ERK activity was markedly suppressed to a level comparable to that of the pre-induced cells at high cell density (Fig 2C right). We also analyzed other MDCK cell lines that encoded HRasG12V, NRasG12V, and BRafV600E about ERK activity at different cell densities and the expression of oncogenic mutants. Again, the cells were plated at low and high cell densities and examined for ERK activity (Fig 2D). The beeswarm plots confirmed that the induction of oncogene products raised the ERK activity at low cell density, but that the oncogene productsinduced high ERK activation could be antagonized by high cell density. These observations clearly indicate that the high ERK activity induced by active oncogene products is subject to the suppression by high cell density.

ERK activity change in the wound-healing assay
The low ERK activity at high cell density suggests the two following possibilities. First, low cell density, i.e., a large cell-to-substratum area, may be required for the induction of high ERK activity by oncogene products. Second, a free edge, i.e., a plasma membrane that does not face the other cells, may be required for ERK activation by the oncogene products. In support of the latter view, it has been proposed that Ras is activated at the free edge but not at the plasma membrane facing the neighboring cells [36]. To resolve this question, we performed a wound healing assay. MDCK cells expressing NRasG12V were plated on two chambers separated by a silicon wall, which could be removed during the experiment. No difference in ERK activity was found whether or not the cells were facing the silicon wall (Fig 3A). Upon removal of the silicon wall, the cells that had faced the silicon wall moved forward and exhibited a rapid increase in ERK activity, even though these cells were still in contact with the follower cells (Fig 3B and 3C). When the cells of both edges filled the gap, ERK activity was decreased gradually. In the leading cells that faced to the silicon insert, the length of the cell border in contact with the other cells did not change significantly during cell migration. Therefore, the high cell density, but not the cell-to-cell contact, appears to suppress ERK activity in MDCK-NRasG12V cells cultured at high cell density.

Accumulation of tyrosine-monophosphorylated inactive ERK at high cell density
How can ERK activity be suppressed in MDCK cells expressing the active Ras or BRaf? To answer this question, we examined the phosphorylation status of ERK by using Phos-tag gels as described previously [34]. ERK has four distinct forms according to the phosphorylation status of tyrosine and threonine residues within the catalytic loop (Fig 4A). Non-phosphorylated ERK (np-ERK) is first phosphorylated on tyrosine to generate tyrosine-monophosphorylated ERK (pY-ERK). pY-ERK is then further phosphorylated to generate tyrosine and threonine diphosphorylated ERK (pTpY-ERK), which is the enzymatically-active form. Dephosphorylation can follow two pathways via either pY-ERK or threonine-monophosphorylated ERK (pT-ERK). These four states of ERK could be separated by Phos-tag gels and probed with an anti-ERK antibody (Fig 4B). Because ERK2 is significantly more abundant than ERK1, we concentrated our efforts on quantification of the amount of ERK2 phospho-isoforms in MDCK cells with or without HRasG12V expression ( Fig 4C). As was anticipated from the imaging data, the amount of pTpY-ERK was increased by the induction of HRasG12V and correlated inversely with the cell density. Notably, this decrease in pTpY-ERK at high cell density was accompanied by the increase in pY-ERK. Meanwhile, np-ERK was not significantly changed by the difference of cell density. Similar results were obtained by using MDCK cells expressing NRasG12V, KRasG12V, or BRafV600E (Fig 4D). Taken together, our observations raised the possibility that serine/threonine phosphatases were activated to invoke the shift from pTpY-ERK2 to pY-ERK2 under the high cell density condition (Fig 4E).
Recently, Iwamoto et al. reported that pTpY-ERK and pT-ERK, but not pY-ERK, are increased in the Ras-transformed keratinocyte cell line HaCaT A5 [37]. In this cell line, DUSP6 has been shown to contribute to the accumulation of pT-ERK. The cause of the discrepancy may be the difference of cell type.

Decreased phosphorylation of Akt and S6 ribosomal protein at high cell density
If serine/threonine phosphatase activity was increased at high cell density, decreased phosphorylation would be observed not only in ERK2 but also other downstream effector proteins of Ras. To examine whether this is the case, the phosphorylations of Akt and its downstream effector S6 ribosomal protein were examined by Western blotting (Fig 5). Upon induction of KRasG12V, phosphorylations of Akt and the S6 protein were increased at low cell density. This increase in phosphorylation was attenuated at high cell density (A and C). Upon the induction of BRafV600E, neither phosphorylation of Akt nor that of the S6 ribosomal protein was increased, which may suggest negative regulation at the level of Ras (B and D). Nevertheless, the phosphorylations of Akt and the S6 ribosomal protein were decreased at high cell density, implying that serine/threonine phosphatase activity was increased at high cell density.

Decreased pY-ERK2 in calyculin A-treated MDCK cells at high cell density
To gain further insight into the nature of phosphatases upregulated in confluent MDCK cells, we used inhibitors of PSPs. PSPs comprise three major families: phosphoprotein phosphatases (PPPs), metal-dependent protein phosphatases (PPMs), and the aspartate-based phosphatases [38,39]. The largest PPP-family proteins include PP1 and PP2A, which are ubiquitously expressed. Calyculin A inhibits a broad range of both PP2A and PP1. Okadaic acid inhibits PP2A completely, but inhibits PP1 only partially at the concentrations used in this study [40,41]. Neither of them inhibits PPMs or the aspartate-based phosphatases [42].
MCDK-HRasG12V cells plated at low and high cell densities were treated with calyculin A and okadaic acid and examined for ERK activity by FRET imaging (Fig 6A and 6B). Calyculin A, but not okadaic acid, was found to increase the ERK activity, suggesting that PP1 is Cell Density-Dependent Increase in Tyrosine-Monophosphorylated ERK2 responsible for the confluency-dependent suppression of ERK. To further confirm the effect of calyculin A, the phosphorylation status of ERK was examined by Western blotting with Phostag gels. In agreement with the imaging data, calyculin A was found to decrease np-ERK2 and to increase pTpY-ERK2 at all cell densities (Fig 6C and 6D). Importantly, calyculin A decreased pY-ERK2 at high cell density in MDCK cells expressing HRasG12V, which supports the findings shown in Fig 3 that accumulation of pY-ERK was the primary cause of ERK suppression at high cell density.

Discussion
The loss of contact inhibition is a hallmark of cancer cells in vitro [19]. In earlier studies, expression of active oncogenes was believed to be sufficient to induce contact inhibition; however, later studies have shown that additional mutations endow robust malignant features [43,44]. In agreement, we have also found that expression of Ras or BRaf was not sufficient to override the cell density-dependent suppression of ERK2 activity. We have further shown that the suppression of ERK activity was caused primarily by the transition of the active pTpY-ERK2 to inactive pY-ERK2. The sensitivity to calyculin A, but not to okadaic acid, suggests the involvement of PP1 in the cell density-dependent suppression of ERK2.
Dephosphorylation is the reverse reaction of protein phosphorylation and should play important roles in the regulation of cancer cell growth. Nevertheless, the number of PSPs are markedly less than that of protein kinases, implying that PSPs revert broad spectrum of protein kinase reactions, each of which is dictated by a specific protein kinase(s) [39]. In contrast to the pro-oncogenic roles played by protein kinases, PSPs are generally regarded as the tumor suppressor. For example, PP2A is known to regulate cell cycle and apoptosis of cancer cells. Furthermore, recent preclinical studies have shown that PP2A-activating drugs can antagonize cancer development and progression [45]. Similarly to PP2A, PP1 regulates a number of biological phenomena including cell division, apoptosis, metabolism, protein synthesis, regulation of cytoskeleton and ion receptors [46]. Due to the pleiotropic functions of PP1, there are some discrepancies on the mechanism by which PP1 regulates ERK. PP1 dephosphorylates phospho-Ser 259 of CRaf, which binds to 14-3-3 and thereby suppresses kinase activity. Consequently, PP1 activates ERK via CRaf activation [47,48]. On the other hand, PP1 has been shown to negatively control ERK via DARPP-32 protein in a subset of medium-size spiny neurons of the dorsal striatum and nucleus accumbens [49]. We suggest that PP1 negatively regulates ERK activity via dephosphorylation of threonine (Fig 4) in a cell density dependent manner. Therefore, the action of PP1 on the Ras-Raf-ERK pathway appears to be dependent on not only the cell types but also the cell density. It should also be mentioned that the discrepancy between the expression of active Ras or BRaf and ERK activity in cancer tissues may be ascribable to the activity of PP1 in each cancer tissue.