Constitutive expression of multidrug resistance in human colorectal tumours and cell lines.

In this study we report detection of mdr1 gene expression in the liver metastases of 7/11 patients with colon carcinoma and characterise the MDR phenotype associated with a panel of 19 human colon carcinoma cell lines. Within this panel, mdr1 mRNA biosynthesis and surface localisation of Pgp were assessed with respect to MDR functionality where the cell lines are representative of different clinical stages of tumour progression, metastatic potential and differentiation. The data indicates that constitutive levels of mdr1 mRNA/Pgp expression may not necessarily result in the functional expression of the MDR phenotype. While low levels of mdr1 mRNA/Pgp were detected in 5/8 well differentiated colon cell lines, only 2/8 were functionally MDR. In contrast, 10/11 moderate and poorly differentiated lines expressed mdr1 mRNA/Pgp and of these, 9/11 were functionally MDR. The phosphorylation status of the mature 170 kD P-glycoprotein and the surface localisation of this glycoprotein showed the strongest correlation with functionality. Analysis of cell lines for cross-resistance and chemosensitivity profiles against a battery of chemotherapeutic drugs suggests multiple mechanisms, in addition to Pgp, contribute to the overall resistance of colorectal cancer. ImagesFigure 1Figure 2Figure 3Figure 4

Colorectal cancer is second only to lung cancer as the leading cause of death due to cancer in the United States. Although surgery is successful in a large percentage of these cases, the survival rate deteriorates rapidly if the tumour has invaded through the serosa or has metastasised to regional lymph nodes or liver (Silverburg & Lubera, 1986). As a rule, chemotherapy is the first line of treatment for disseminated disease, however colorectal cancer is refractory to most chemotherapeutic agents (Haller, 1988). Unfortunately, very little is known about the mechanisms responsible for the intrinsic drug resistance of this disease. Successful chemotherapy will ultimately depend on the elucidation of these mechanisms.
In vitro studies utilising tumour cells that acquired resistance by selection with anti-tumour drugs have identified a form of resistance (i.e. multidrug resistance or MDR) that is characterised by decreased sensitivity to a broad range of structurally and mechanistically dissimilar natural product anti-cancer drugs (e.g. doxorubicin, vincristine, etoposide, Riehm & Biedler, 1971;Bech-Hansen et al., 1986). These compounds represent some of the most effective anti-cancer agents currently available for treatment of a wide range of malignancies. The hallmark of MDR is decreased drug accumulation that is related to the increased expression of the mdrl gene product, a 170 kD membrane glycoprotein ternied Pgp (Kartner et al., 1983). The predicted amino acid sequence of mdrl suggests that Pgp functions as an energy dependent efflux-pump (Chen et al., 1986) and is consistent with studies showing that Pgp can bind drug and facilitate efflux by an ATP dependent process (Cornwell et al., 1986).
Increased expression of mdrl mRNA/Pgp has been observed clinically in a variety of cancers (e.g. multiple myelomas, sarcomas, neuroblastomas, breast) that relapsed following an initial response to chemotherapy (Ma et al., 1987;Gerlach et al., 1987;Dalton et al., 1989;Goldstein et al., 1990), a situation that is analogous to in vitro models for establishing resistant cell lines. Pgp has also been found localised to the lumenal surface of normal epithelial cells lining the colon, kidney, pancreas and bile ducts in addition to its localisation in the adrenal gland, placenta and vascular endothelial cells in the testes and brain (Thiebaut et al., 1987;Yang et al., 1989;Cordon-Cardo et al., 1989. This pattern of distribution is consistent with a physiological role for Pgp in the protection of normal tissues against toxicants. Expression of mdrl mRNA has been frequently detected in tumours derived from these tissues (Fojo et al., 1987a,b;Lai et al., 1989;Goldstein et al., 1989), before the patients received chemotherapy, suggesting that MDR may also be responsible for the inherent resistance of these tumours. Nevertheless, it is difficult to establish a causal relationship between intrinsic clinical resistance and total mdrl mRNA/Pgp expression levels given the possibility that tumour specimens are often contaminated with the normal mdrl expressing tissues, and that mdrl mRNA/Pgp may not be expressed uniformly throughout the tumour (Schlaifer et al., 1990;Weinstein et al., 1991). This may explain the high degree of variability in mdrl mRNA expression reported for solid tumours (Fojo et al., 1987a,b;Lai et al., 1989;Goldstein et al., 1989). It is also not known whether the level of Pgp found in untreated samples is sufficient to confer resistance, and/or whether the Pgp expressed in these tumours is fully functional. Finally, it is likely that additional drug resistance mechanisms (e.g., glutathione-S-transferase, topoisomerase, etc.) contribute to the overall level of intrinsic tumour drug resistance (Kramer et al., 1988(Kramer et al., , 1989. In this manuscript we report detection of mdrl mRNA in the liver metastasis of patients with colon carcinoma and evaluate the association of mdrl gene expression with Pgp biosynthesis and function (i.e. verapamil-inducible drug accumulation and cytotoxicity) in 19 human colon carcinoma cell lines.
M. Brattain, Baylor College, Houston, TX, USA. All the aforementioned cell lines were established from tumour tissue prior to any chemotherapy (personal communications). The remaining cell lines were obtained from the American Type Culture Collection (ATCC).
Northern blot analysis Total cellular RNA was isolated from adherent cells by scraping in guanidium isothiocyanate followed by centrifugation through cesium chloride (Chirgwin et al., 1979). Fresh tumour and normal tissue were treated to minimise RNA digestion by snap freezing in liquid nitrogen. Frozen tissues were pulverised on a metal surface placed on a bed of dry ice prior to homogenisation. Equivalent amounts of RNA (10 itg/lane) were electrophoresed through a denaturing 1% agarose/formaldehyde (7%) gel and stained with ethidium bromide, to check for RNA integrity and equal loading, prior to transfer to GeneScreen Plus membrane (NEN, Boston, MA, USA). Filters were prehybridised for 60 min at 60°C in 50% formamide, 7% SDS, 0.25 M sodium phosphate pH 7.2, 0.25 M NaCl, 1 mM EDTA, 1I00 tg ml-I denatured salmon sperm DNA, 200 pg ml-' tRNA followed by hybridisation for 18 h at 60°C in the same buffer containing 2 x 106 c.p.m. ml1of labelled riboprobe. A synthetic riboprobe was prepared by SP6 polymerase transcription of the pHDR5A pGEM human mdr probe (Ueda et al., 1987). Filters were washed twice for 60 min in PSE (0.25 M sodium phosphate, pH 7.2, 2% SDS, 1 mM EDTA) followed by final wash in 30 mM NaCl, 3 mM sodium citrate, 0.1% SDS at 65%. Filters were exposed to Kodak X-AR5 film at -70°C for 1-3 days.
Immunoprecipitation of P-glycoprotein Subconfluent dishes of colon carcinoma cell lines were washed twice in phosphate buffered saline (PBS) and incubated for 1 h in methionine or phosphate-free Dulbecco's modified Eagles medium (DMEM) followed by incubation (3 or 18 h) in medium containing 35S-methionine (150 yCi ml-') or 32p Orthophosphate (200 JACi ml-'). Metabolically labelled cells were rinsed briefly in PBS, and lysed in PBSTDS buffer (PBS pH 7.4, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 2 mM phenylmethylsulfonylfluoride, 10 U ml-' aprotinin). Lysis was carried out for 20 min at 4°C, followed by shearing of lysates through a 24 gauge needle. Lysates were clarified in a microfuge for 20 min at 40C, supernatants were removed and a 10 IL aliquot was taken for protein estimation using the bovine serum albumin protein assay system (Pierce, Rockford, IL, USA). Prior to immunoprecipitation samples were standardised for protein concentration using 400 iLg for each cell line. Lysates were incubated overnight at 4°C with mdrl polyclonal antibody (Oncogene Science, Manhasset, NY, USA), followed by the addition of protein A Sepharose beads and a further 90 min incubation. Immune complexes were washed three times in PBSTDS, three times in 0.1 PBS, followed by incubation in standard sample buffer for 20 min at room temperature. All samples were run on 7.5% polyacrylamide gels, dried and exposed for 1-3 days to X-ray film.
Functional assays Drug accumulation studies The effect of the Pgp antagonist verapamil (OiLg ml-'), on [3H]-daunorubicin (1O im; specific activity 1 mCi/10 ILM) accumulation was determined in replicate suspensions of colon cells (2 x 106 cellsml-') using a standard silicone oil technique for separating cells from extracellular medium, as previously described (Kramer et al., 1988). Initial time vs uptake studies showed that [3H]-daunorubicin accumulation reached a plateau between 60-90min (data not shown). In all subsequent studies, the 90 min time point was used to quantitate the fold increase in drug accumulation (nmoles/106 cells) caused by the verapamil treatment. Although this procedure does not take into account possible differences in cell volumes, drug metabolism or passive drug permeation coefficient (Spoelstra et al., 1992), it is adequate for our purpose of comparing multiple cell lines, with each cell line serving as its own control.
Drug sensitivity studies The sensitivity of the cell lines to doxorubicin in the presence and absence of verapamil (10 fig ml-') was determined in monolayers of cells using the sulforhodamine B (SRB) protein binding colorimetric cytotoxicity assay (SRB) described by Skehan et al. (1990). In this assay, 1 x I04 cells were plated in each well of a 96-well microtitre plastic, and the following day the cells were incubated with drugs for 3 h, washed twice in phosphate-buffered saline and fresh medium was added. Three hour drug incubations were used to minimise the toxicity of prolonged verapamil treatment. Cells were fixed in 10% trichloroacetic acid (TCA) and stained with SRB 4 days after drug treatment. Absorbance values were recorded with a microplate reader (molecular Devices) and values were reported as per cent of control (T/C) from the means of duplicate determinations. IC50 values in the absence and presence of verapamil were calculated and dose modifiying factors (DMF) were determined (IC50 control/IC50 + verapamil). Chemosensitivity profiles against a panel of chemotherapeutic agents were conducted using a slight modification of this procedure. In these studies, drug treatments were for 48 h and duplicate plates of cells were fixed with TCA at the time of drug addition to establish To values, as described by Monks et al. (1991). The effect of drug treatments over a range of concentrations were calculated using the formula, T-To/C-To, to determine the IC50 value.
Flow cytometry Surface staining of cells for P-glycoprotein expression was accomplished using the 4E3.16 anti-P-glycoprotein monoclonal antibody (Arceci et al., 1993). Adherent cells were collected in cold phosphate buffered saline (PBS) by gentle scraping with a rubber policeman. Cells were washed twice in cold PBS and 1 x 106 cells were resuspended in 100 yl of PBS containing 1:1 dilution of human serum with PBS and incubated at 4°C for 30 min to block Fc receptors. Two milliliters of PBS were then added to the cells which were collected by centrifugation at 600 g for 3 min. Pelleted cells were resuspended in 1001l of PBS containing 2% goat serum and 10 g ml-' of the anti P-glycoprotein antibody 4E3.16 or an IgG2a isotype matched control antibody. This mixture was incubated for 30 min at 4°C, cells were washed twice with cold PBS, followed by resuspension in 100 jlI of PBS containing 2% goat serum and FITC-labelled goat-antimouse Ig (Fab)2 fragment (TAGO) at a 1:30 dilution. Cells were incubated with the second antibody for 30 min at 4°C in the dark, followed by two washes in cold PBS and fixation in 2% paraformaldehyde prior to analysis. The level of P-glycoprotein expression was determined using a Becton-Dickinson FASCAN II using LYSYS software application.
Results mdrl mRNA expression in metastatic colon cancer mdrl mRNA was measured in 16 colon tumour specimens ( Figure 1) that included four primary lesions, 11 liver and one lymph node metastasis, that were obtained from patients before they received chemotherapy. Northern blot analysis revealed a single 4.5 kb mRNA (mdrl) present in most tissues ( Figure 1) with the exception of one liver metastasis in which we detected a second transcript of approximately five kilobases. Integrity of RNA, equal loading and transfer completion was confirmed by ethidium bromide staining ( Figure   1).  (N, nI), primary colon tumours (1'), lymph node metastasis (ly) and liver metastasis of colon tumours. The integrity and comparative loading of RNA is shown in the ethidium bromide stained gel in each panel. a, mdrl expression in adjacent normal mucosa (N) and primary colon tumours (1°) from three patients, b, detection of mdrl transcripts in normal (nl), tumour (1°) and lymph node metastasis (ly) from a single patient. From both panels a 4.5 kb mdrl message is detected in seven of 11 liver metastases with an additional larger message detected in patient 1 a.

MULTIDRUG RESISTANCE IN COLORECTAL
Levels of expression of mdrl mRNA varied widely between tumour samples. Nevertheless, mdrl mRNA was readily detected in 7/11 liver metastases although no mdrl message was detected in the one lymph node metastasis that was evaluated. In two additional patients, we were able to obtain both the primary tumour and the corresponding liver metastasis. In one case, the primary tumour and the corresponding liver metastasis expressed comparable levels of mdrl transcripts; in the second case mdrl mRNA levels were much higher in the metastasis of the second patient (data not shown). When possible, tumour samples were run in parallel with adjacent normal mucosa obtained at the time of surgery. The relative level of mdrl mRNA detected in primary tumours and adjacent normal tissues also varied considerably, with normal tissues expressing higher levels in two of the three pairs evaluated (Figure 1, panel a).
Characterisation of MDR in human colon carcinoma cell lines Although every attempt has been made to standardise loaded samples in Northern blot analysis, heterogeneity within tumours and variability between tumour content make interpretations of results difficult. Immunohistochemistry and in situ hybridisation have been used previously to resolve this problem, however, such approaches cannot relate Pgp expression to functionality. To clarify this issues we have used a panel of colon carcinoma cell lines representing a more homogeneous starting cell population for the analysis of mdrl mRNA/Pgp expression and Pgp function. Nineteen human colon carcinoma cell lines were classified according to differentiation state based on a variety of established criteria that include the histopathology of the original tumour and subsequent xenografts, carcinoembryonic antigen (CEA) production and, in some cases, invasive potential (Table I). In this way, the panel of human colon cell lines represents a range of differentiation phenotypes, in vitro and in vivo, with different invasive and metastatic potentials. From the data presented in Table I, histologically well differentiated cell lines exhibited many characteristics distinct from the phenotype expressed by histologically moderate or poorly differentiated cells.  Expression of mdrl mRNA and Pgp in colon cells Northern blot analysis of RNA from a representative panel of colon cell lines showed expression of a 4.5 kb mdrl transcript in the majority of cell lines evaluated (12/19) ( Figure  2). Mdrl mRNA was detected in only 3/8 well differentiated cell lines, and in 9/11 of the moderate and poorly differentiated cell lines (Table II). Higher levels of mdrl mRNA were apparent in the moderate and poorly differentiated cell lines with the highest levels found in Clone A and MIP 101, which were derived from, and are classified as, representative of poorly differentiated phenotypes (Dexter et al., 1979(Dexter et al., , 1981Niles et al., 1987). The wide range of mdrl mRNA levels detected in colon cell lines (Table II) was analogous to our observations with colon tumour tissue ( Figure 1). Pgp biosynthesis was established after overnight labelling of cells with "5S-methionine followed by immunoprecipitation with a polyclonal mdrl antibody ( Figure 3, Table II). In a number of cell lines small amounts of the 170 kd P-glycoprotein were detected where no mdrl mRNA was detected in repeated Northern blot analysis (e.g. CaCo-2, Figure 3). Pgp was detected in 5/8 well differentiated and 9/10 moderate and poorly differentiated cell lines tested. In comparison to well differentiated colon cells, the moderate and poorly differentiated cell lines displayed higher levels of Pgp, proportional to the respective cellular levels of mdrl mRNA detected (Table  II). The highest levels of Pgp were found in the poorly differentiated Clone A and MIP 101 cell lines (Figure 3, top panel). The antibody used in these experiments immunoprecipitates a second protein of >200 kD. This protein appears unrelated to Pgp, since it is precipitated from cell lines which do not express detectable P-glycoprotein (HT29, CCL238). However, it serves as a useful internal control for Repetition of experiments following overnight labelling with 32P-orthophosphate confirmed previous observations of phosphorylation of the 170 kd Pgp (Figure 3, lower panel), where the degree of phosphorylation was found to be a good indicator of Pgp pump activity (Table II) (Greenberger et al., 1988;Richert et al., 1988). Labelling of cells for shorter periods (3 h likely that the altered mature P-glycoprotein in this cell line results from underglycosylation. It is interesting to note that Moser displays significant Pgp pump activity despite aberrant processing of the immature form (Table II). Representation of the 140 kd precursor in DLD-l is at comparable levels to that of Moser and yet only minimal mature P-glycoprotein is resolved in immunoprecipitation experiments (compare lanes, Figure 4, (Table II).

Functional assessment of MDR in colon cells
Colon cells expressed a range of differences with respect to net daunorubicin accumulation and doxorubicin cytotoxicity (  Figure 4 Immunoprecipitation of P-glycoprotein from a representative panel of colon carcinoma cell lines following a short labelling period (3 h) with 35S-methionine (upper panel) or 32P-orthophosphate (lower panel). P-glycoprotein is resolved as a doublet (upper panel) representing the mature 170 kD glycoprotein and the immature unglycosylated 140 kD precursor. Note the faster migrating mature P-glycoprotein associated with Moser (unlabelled arrow) and the absence of detectable mature Pglycoprotein (170) in DLD-1 (upper panel). Parallel experiments with immunoprecipitation of Pgp from 32P-orthophosphate cell lysates results in detection of the mature 170 kD P-glycoprotein (lower panel), once again demonstrating altered migration in the Moser cell line. rubicin/106 cells, with the lowest levels being found in mdrl mRNA/Pgp positive colon cells lines. The doxorubicin concentration that inhibited colon cell growth by 50% (ICm; 3 h drug exposure) varied as much as 50-fold among the various colon cell lines, ranging from 0.05p1M to 2.5 1M, with mdrl mRNA/Pgp-positive cells having the highest IC50 values.
Pgp function was estimated by determining the fold increase in net [3H]-daunorubicin accumulation and doxorubicin cytotoxicity (decreases in IC50 value) that was caused by treating cells with an antagonist of Pgp (i.e. verapamil, 25 LM) (Table II). Verapamil treatment increased drug accumulation in colon cell lines by 20 to 510%, and resulted in a corresponding increase in doxorubicin cytotoxicity (i.e. fold decrease in ICo value). The percentage increase in drug accumulation and cytotoxicity was used to estimate Pgp function, and was found to correlate with mdrl mRNA/Pgp expression levels. For example, colon cell lines that had no measurable mdrl mRNA or Pgp (e.g. CCL238, CX-1, HT29, CCL227), exhibited no functional Pgp activity as determined by the verapamil-inducible drug uptake and cytotoxicity assays (Table II). Cells expressing the lowest detectable levels of mdrl mRNA/Pgp (e.g. LoVo, CL187, CaCo-2, CCL220.1) also had no measurable increases in drug accumulation or cytotoxicity in response to verapamil. However, cell lines expressing low to moderate levels of mdrl mRNA/Pgp (e.g. CL188, CCL231, DLD-1), did exhibit a range (10-100%) of verapamil-induced increases in drug accumulation and cytotoxicity. The highest levels of Pgp activity (>250%) were found in those cell lines expressing the highest levels of mdrl mRNA/Pgp (i.e. Moser, MIP 101, Clone A). Only 2/8 of the well differentiated cell lines expressed the MDR phenotype as defined by these criteria, and these well differentiated MDR positive cell lines (i.e. CLI88 and CCL233) exhibited the lowest measurable levels of functional activity (20-30%). In contrast, 9/11 moderate and poorly differentiated cell lines displayed functional MDR phenotypes exhibiting Pgp activities ranging from 20->500%.
Non-Pgp mechanisms of MDR in colon carcinoma cells Representative colon cell lines were analysed for crossresistance and chemosensitivity profiles against a battery of chemotherapeutic drugs (Table III). In these studies, the cells were exposed to drugs for 48 h. Cells expressing high constitutive levels of mdrl mRNA/Pgp i.e. MIP 101, Clone A, Moser, were cross-resistant to drugs normally associated with the MDR phenotype e.g. vincristine, etoposide and doxorubicin. The relative degree of drug resistance was determined by comparing the IC50 values in these cells to a representative Pgp-negative cell line e.g. CCL238. The drug sensitivity profile of CCL 238 was not appreciably different to the sensitive human leukaemia cell line HL60. The relative resistance (i.e. IC50 of MDR+ cells/IC50 CCL238) of these high expressing cell lines was proportional to the cellular levels of mdrl mRNA/Pgp and ranged from 2.5 to 50-fold for doxorubicin 11 to 50-fold for vincristine and 5 to 10-fold for etoposide. While, vincristine resistance correlated with the level of mdrl mRNA/Pgp expression, a direct relationship between etoposide resistance and mdrl mRNA/Pgp expression could not be established. For example, CCL228 expressed lower levels of mdrl mRNA/Pgp compared to MIP 101, Clone A or Moser (Table III) and were proportionally less resistant to vincristine, and yet of the MDR positive cells, CCL228 was the most resistant to etoposide. Moreover, the cell lines demonstrating the highest levels of etoposide resitance i.e. CaCo-2 and CX-1 were not MDR-positive as shown by the lack of mdrl expression as well as by functional assays. As might be expected the MDR cells were not cross-resistant to 5-fluorouracil, cisplatin or chlorambucil (Table III). Cells exhibiting a modest degree of functional Pgp activity (Table II) expressed detectable but correspondingly lower levels of mdrl mRNA/Pgp. These cells, CL188, DLD-1, CCL231, were not cross-resistant to the standard MDR drugs. For example, both mdrl mRNA and Pgp were detected in CCL231 cells at levels that were proportional to the verapamil mediated increase in 3H-daunorubicin accumulation of 80% (Table II). However, CCL231 was equally sensitive to doxorobucin, vincristine and etoposide compared to our non-Pgp reference CCL238 colon carcinoma cell line. These observations suggest that in colon cancer, other mechanisms of resistance e.g., altered topoisomerase, glutathione may also contribute to the pattern of resistance particularly when mdrl mRNA/Pgp when expressed at low levels. However at higher levels of expression, mdrl mRNA/Pgp becomes the major determinant of resistance.

Discussion
Multidrug resistance has been implicated as a contributing factor in the intrinsic resistance of several solid tumours, including colon, ever since the original observation by Thiebaut et al. (1987) that the mdrl gene product was constitutively expressed in the normal tissues from which these tumours were derived. Subsequent studies by Fojo et al. (1987b), Goldstein et al. (1989) and others, provided additional support for this by demonstrating that mdrl mRNA was often present in the primary colon tumour specimens before patients received chemotherapy. A recent report by Weinstein et al. (1991)  cells and in the lymph node metastasis of patients with colon cancer. The present study extends these observations by demonstrating that mdrl mRNA was also expressed in liver metastasis. These findings demonstrate that colon carcinoma cells not only retain the capacity to express the mdrl gene, but that this characteristic of the normal colonic epithelium can be maintained throughout all stages of colon tumour progression. This observation is consistent with a recent study in neuroblastoma patients in which Pgp was detected in a high percentage of the advanced lesions (Chan et al., 1991). The findings in neuroblastoma and colon cancer patients provide a compelling basis for understanding why metastatic disease is refractory to certain chemotherapeutic drugs, particularly if the metastasis developed from primary tumours that were derived from Pgp-expressing normal tissues.
In this study, mdrl mRNA levels varied considerably in colon tumours and in normal tissue, and is consistent with all previous studies evaluating surgical material (Fojo et al., 1987a,b;Lai et al., 1989;Goldstein et al., 1989). A relationship between mdrl mRNA/Pgp expression and clinical drug resistance is particularly hard to establish in colon cancer because these patients rarely receive chemotherapy with MDR-associated drugs. Therefore, one cannot relate expression levels with patient outcome in response to chemotherapy as has been done in clinical studies involving patients with soft tissue sarcomas (Gerlach et al., 1987), myelomas (Dalton et al., 1989), or neuroblastomas (Chan et al., 1991). Previous studies using surgical material from untreated colon cancer patients have attempted to place significance on expression levels in tumours that were higher than the levels expressed in adjacent normal tissues. However, normal tissue mdrl  mRNA levels were frequently higher than tumour levels (Fojo et al., 1987b). It has never actually been established if normal colon cells are also drug resistant, and at what level of constitutive Pgp expression does resistance actually occur. This may be important given that most of what we know about Pgp and resistance has come from studies with cell lines that were selected on the basis of functional criteria (i.e. they survived treatment with escalating doses of chemotherapy). Thus we felt that human colon tumour cells offered the best available model to study the relationship between constitutive mdrl mRNA/Pgp expression and functional resistance, and the possible relationship between mdrl mRNA/Pgp expression and colon tumour progression. Cell lines DLD-1, Clone A, Clone D, MIP 101, and Moser, all of which express P-glycoprotein, were established from tumour material prior to exposure from chemotherapeutic agents. The colon carcinoma cell lines used in this study were selected primarily on the basis of differentiation using established criteria e.g. histology, carcinoembryonic antigen secretion, these parameters are summarised in Table I. Previous studies have shown that poorly differentiated colon carcinoma cells were the most aggressive as assessed using in vitro adhesion and invasion assays (Daneker et al., 1989). These in vitro studies support clinical observations which have related poorly differentiated colon tumour histologies with a poor prognosis and a greater likelihood of metastatic involvement. It is clear from the biochemical data presented that the most aggressive, poorly differentiated colon tumour cell lines within the panel expressed the highest constitutive levels of P-glycoprotein correlating with their relative functionality recorded in drug uptake assays. In contrast to previous reports (Mickley et al., 1989;Mizoguchi et al., 1990) we find no correlation between differentiation status and the MDR phenotype in the cell panel studied, but observe that P-glycoprotein expression can be maintained in both well and poorly differentiated colon cell lines. This observation is consistent with that reported by Park et al. (1990). Although it can be argued that cell lines are not representative of the in vivo lesion, it should be considered that, since poorly differentiated tumours represent approximately 5% of the colorectal tumours resected, it is difficult to generate sufficient numbers to definitively evaluate the differentiation/MDR phenotype association using human tumour material. Interestingly, Pgp was immunoprecipitated from some cell lines where no MDR functionality was detected. This may reflect the limitations of sensitivity of the assays used or the requirement for a threshold level of Pgp expression to acquire functional drug resistance. However, moderate/well differentiated colon cell lines HT29 and CCL238 consistently displayed an absence of detectable mdrl message or immunoprecipitable Pgp and demonstrated a lack of MDR functionality in repeated assays consistent with results reported by Spoelstra et al. (1991). The amount of mature 170 kD P-glycoprotein resolved in protein standardised immunoprecipitation was directly reflected by the phosphorylation status of Pgp, revealing an absolute correlation with MDR functionality within the colon carcinoma cell panel studied.
Previous characterisation of the biosynthesis of Pgp has identified a 140 kD precursor molecule which is processed, via N-linked glycosylation, to a 170 kD species identified as the mature P-glycoprotein (Greenberger et al., 1988;Richert et al., 1988). Despite the altered maturation of Pgp in the Moser cell line the mature form is found to be phosphorylated and cell surface associated establishing the MDR phenotype displayed by Moser in functional assays. In contrast, DLD-1 displayed only minimal detectable functional Pgp activity despite relatively high expression levels of mdrl mRNA in Northern blot analysis. Consistent with this observation is the lack of mature Pgp (170 kD) found in this cell line despite detection of comparable levels of the precursor molecule (140 kD) to that found in other cell lines within the panel e.g. Moser. This apparent lack of processing of the DLD-1 Pgp is further reflected in the absence of phosphorylated Pgp and the lack of P-glycoprotein at the cell surface. The correlation observed between the phosphorylation status and membrane association of Pgp with the MDR phenotype implicates both of these factors as important in establishing cellular drug resistance. In all mdrl immunoprecipitation protocols involving 32P-orthophosphate labelling, Pgp was resolved as a single band comigrating with the mature 170kD P-glycoprotein. From these results we conclude that only the mature form of P-glycoprotein is phosphorylated in colon cells.
The functional significance of Pgp processing may be particularly important in non-selected, constitutively expressing cells and tumours. This possibility further complicates attempts to attribute the clinical resistance of colon cancer solely to changes in the expression of mdrl mRNA and/or detection of Pgp in immunohistochemistry or immunoprecipitation protocols. Although our results, showing a high frequency of mdrl mRNA/Pgp expression in colon tumour specimens and colon carcinoma cell lines supports a role for MDR in the clinical resistance of colon carcinomas, we also report that the low constitutive levels of Pgp expressed in many colon carcinoma cells may not be sufficient to confer resistance. Moreover, mdrl mRNA/Pgp expression levels correlated poorly with etoposide resistance, and several colon carcinoma cell lines with appreciable levels of expression and demonstrated Pgp activity (e.g. CCL 231) were no more resistant to MDR drugs (e.g. vincristine) than were Pgpnegative colon cell lines. These observations are consistent with the view that multiple mechanisms in addition to Pgp (e.g. topoisomerase, 6-alkylguanine-DNA-alkyltransferase, glutathione peroxidase) contribute to the overall resistance of colorectal cancer (Kramer et al., 1988;Redmond et al., 1991). This work was supported by NIH grants CA50473 (R.K.); CA42944 and CA44704 (I.C.S.). We are grateful to Carol Ann Hannan for typing the manuscript.