Inhibition of fibroblast growth factor receptor 3 induces differentiation and apoptosis in t(4;14) myeloma

The inhibition of the primary genetic lesion (BCR-ABL, PML-RARα) can induce differentiation and apoptosis in human myeloid tumors.^1,2  Transient inactivation of MYC in an MYC transgenic mouse tumor model results in tumor cell differentiation, followed by apoptosis on subsequent MYC reactivation.³  We wondered to what extent these observations can be generalized to other human cancers and, in particular, to lymphoid malignancies. There are 3 main reasons why this has been difficult: (1) the challenge in identifying genes that participate in the primary oncogenic process, (2) the lack of specific molecular inhibitors, and (3) the need for a well-ordered model of cell differentiation.

B-cell differentiation to plasma cells (PCs) provides one such model. Following isotype switch recombination centrocytic B cells differentiate into low-rate immunoglobulin (Ig)-secreting plasmablastic cells, migrate to the bone marrow (BM), and proliferate as immature plasmablasts.^4,5  Differentiation of BM plasmablastic cells into high-rate Ig-secreting PCs is the final step to long-lived PCs, the end-stage of B-cell differentiation. Multiple myeloma (MM) has typically been defined as a neoplasm of terminally differentiated PCs; however, in contrast to fully differentiated PCs that do not proliferate, myeloma cells proliferate slowly and secrete significantly lower amounts of Ig.^6,7 

The t(4;14) translocation, which occurs in approximately 15% of patients with MM, appears to be mediated by errors in isotype switch recombination and results in the dysregulated expression of 2 genes, FGFR3 and MMSET.^8,9  In particular, wild-type (WT) FGFR3, which is not normally expressed by B cells or PCs, becomes ectopically expressed at very high levels (M.C., unpublished data, November 1998). WT FGFR3 induces proliferative signals in myeloma cells and appears to be weakly transforming in a hematopoietic mouse model.^10,11  The subsequent acquisition of FGFR3-activating mutations in some MM is clearly associated with disease progression and is strongly transforming in several experimental models.^11,12  The clinical impact of t(4;14) translocations has been demonstrated in 3 large studies each reporting a marked reduction in overall survival^13-15  with no apparent therapeutic benefit from high-dose chemotherapy. The clinical significance of FGFR3 expression, however, remains somewhat ambiguous, because Keats and colleagues found that t(4;14) conferred a poor prognosis irrespective of FGFR3 expression. Indeed, FGFR3 is not up-regulated in all cases of t(4;14) myeloma, and it has been found that the der14 chromosome can be lost in primary samples and cell lines.^14,16,17  These results suggest that if ectopic expression of WT FGFR3 is important as an initiating event in tumorigenesis, it does not always provide a strong selective advantage during tumor progression. They also raise questions regarding the status of FGFR3 in maintenance of t(4;14) myeloma and its potential value as a therapeutic target. We have used pharmacologic inactivation of FGFR3 to directly assess its role in established t(4;14) MM and determine the potential therapeutic effect of inhibiting a single oncogenic lesion in a nonmyeloid cancer.

PD173074 (Pfizer, Ann Arbor, MI) was dissolved in dimethyl sulfoxide (DMSO) at a stock concentration of 20 mM. For in vivo mice experiments PD173074 was formulated in 0.05 M lactic acid buffer (pH 4). Acidic fibroblast growth factor (aFGF) and heparin were purchased from R&D Systems (Minneapolis, MN) and Sigma (St Louis, MO), respectively. FGFR3 antibodies (clones C15 and B9) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and anti-phosphotyrosine (4G10) from Upstate Biotechnology (Lake Placid, NY).

All human MM cell lines were maintained in RPMI 1640 supplemented with 5% fetal calf serum (FCS), 1% l-glutamine, 100 μg/mL penicillin, and 100 μg/mL streptomycin (BioWhittaker, Walkersville, MD). NIH 3T3 cells were grown in Dulbecco modified Eagle (DMEM) medium (BioWhittaker) supplemented with 10% calf serum, 1% l-glutamine, 100 μg/mL penicillin, and 100 μg/mL streptomycin.

The NIH 3T3 cell line expressing WT FGFR3 has been described.¹¹  Cells (1 × 10⁶) were seeded in 10-cm dishes and allowed to grow for 48 hours. The cells were then serum starved overnight. Thirty minutes prior to aFGF stimulation, PD173074 was added to the dishes at the indicated concentrations. The cells were then stimulated with 30 ng/mL aFGF and 10 μg/mL heparin for 10 minutes at 37°C. Similarly, 10 × 10⁶ MM cells were starved overnight and stimulated as above. Cell lysates were prepared as described.¹⁸  Clarified cell extracts were immunoprecipitated with the C15 anti-FGFR3 and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting with an anti-phosphotyrosine antibody (4G10). After stripping, the filter was reincubated with the B9 anti-FGFR3.

NIH 3T3 cells were transfected in triplicate by calcium phosphate precipitation as described previously.¹¹  For each transfection 500 ng WT FGFR3, Y373C FGFR3, or 100 ng Ras V12 in the presence of 5 × molar excess of pCEFL empty vector were used. After 24 hours, cells were washed with phosphate-buffered saline (PBS), selected with 750 μg/mL Geneticin (Gibco, Rockville, MD), and incubated in the presence or absence of 25 nM PD173074. The culture medium, Geneticin, and PD173074, were replaced every 2 to 3 days. Serial dilutions of neomycin-resistant cells were performed to measure the efficiency of transfection. Fourteen to 16 days after transfection the plates were fixed in methanol, stained with Wright staining (Sigma), and scored for the presence of cellular foci. Transfections were conducted on triplicate plates, and 3 independent experiments were performed.

Six-week-old athymic nude mice (Harlan, Indianapolis, IN) were inoculated subcutaneously with 3 × 10⁵ NIH 3T3 cells expressing Y373C FGFR3 (right flank) and Ras V12 (left flank). Intraperitoneal injections of either 20 mg/kg PD173074 or 0.05 M lactic acid buffer were initiated on the day of tumor injection and continued for 9 days. Ten mice for each experiment were included.

Cell viability was assessed by 3-(4,5-dimethylthiazol)-2,5-diphenyl tetrazolium (MTT) dye absorbance. Cells were seeded in 96-well plates at a density of 20 000 cells/well in RPMI with or without 5% FCS. Cells were incubated with 30 ng/mL aFGF and 100 μg/mL heparin and increasing concentrations of PD173074. For each concentration of PD173074, 10-μL aliquots of drug or DMSO diluted in culture medium were added. Plates were incubated for 48 hours at 37°C, 5% CO₂. The MTT assay was performed according to the manufacturer's instructions (Roche, Indianapolis, IN). Each experimental condition was performed in triplicate.

Cell cycle distribution of cells cultured with 50 nM PD173074 or DMSO for 24 and 48 hours was evaluated by ethidium bromide (EtBr) staining and flow cytometry as described previously.¹⁹  The nuclear preparations were acquired on a Cytomics FC-500 flow cytometer (Beckman Coulter, Miami, FL) and analyzed using CellQuest software (Becton Dickinson, San Jose, CA).

For studies of apoptosis, cells were seeded at an initial density of 2 × 10⁵/mL in 5% FCS/RPMI medium supplemented with DMSO, 12.5 nM or 50 nM PD173074, and cultured for up to 14 days. The medium and drug were replenished every 3 days, and the cell density was adjusted to 2 × 10⁵/mL. Apoptosis was determined by annexin V staining (Boehringer Mannheim, Indianapolis, IN) and analyzed by flow cytometry.

Subconfluent KMS11 and KMS18 cells were cultured in the presence of DMSO or 50 nM PD173074. The medium and drug were replenished every 3 days and cell density adjusted. After 4 (KMS11) and 10 (KMS18) days of culture, cells were sampled and analyzed for evidence of differentiation. Morphologic maturation was assessed by cytospin, followed by Wright-Giemsa staining and examination under × 400 microscopy. A direct immunofluorescence staining technique was applied to analyze PC markers. Single-cell suspensions were incubated for 30 minutes on ice with mouse phycoerythrin (PE)-labeled isotype-matched control antibody, anti-CD19, -CD20, -CD27, -CD31, -CD38, -CD45RA, -CD138, or anti-HLA DR (PharMingen, San Diego, CA). Stained cells were acquired with a Cytomics FC-500 flow cytometer and analyzed using CellQuest software. Dead cells were excluded from the analysis by gating on live cells with forward and side scatter. Light-chain secretion was determined by enzyme-linked immunosorbent assay (ELISA; Bethyl Laboratories, Montgomery, TX). After 4 (KMS11) and 10 (KMS18) days of culture with DMSO or 50 nM PD173074, cells were washed twice in PBS, counted, and resuspended in culture media at a concentration of 5 × 10⁵ viable cells/mL as determined by trypan blue exclusion. Following 5 hours of culture at 37°C cells were spun down, viable cells determined, and supernatant collected for analysis. Samples from 5 to 8 experiments were run in triplicate.

Tissue samples were fixed in 10% formalin and embedded in paraffin, from which 5-μm histologic sections were cut and stained with hematoxylin and eosin. Clot sections were prepared by pelleting 2 × 10⁶ cells, resuspending in 50 μL fibrinogen and 5 μL thrombin, pelletting again, and allowing a clot to form by incubation at 22°C for 5 minutes. The clot was fixed in 10% formalin and embedded in paraffin.

Immunohistochemistry (IHC) studies were performed by indirect immunoperoxidase staining of paraffin tissue sections using a TechMate500 BioTek automated immunostainer (Ventana Medical Systems, Tucson, AZ) and antibodies recognizing Ki-67 (Zymed, San Francisco, CA), CD31 (Dako, Glostrup, Denmark), and cleaved caspase 3 (Cell Technology, Beverly, MA) as previously described.²⁰ 

The xenograft mouse model was prepared as previously described.^21,22  Briefly, 6- to 8-week-old female BNX mice obtained from Frederick Cancer Research and Development Centre (Frederick, MD) were inoculated subcutaneously into the right flank with 3 × 10⁷ KMS11 or 8226 cells in 150 μL RPMI, together with 150 μL Matrigel basement membrane matrix (Becton Dickinson, Bedford, MA). Treatment was initiated when tumors reached 0.5 to 1.0 cm at which time mice were assigned to receive either 25 mg/kg PD173074 orally twice daily or 0.05 M lactic acid buffer at the same schedule. Dosing was continued for 9 days. Twenty-five mice were included in each treatment group. Caliper measurements of the longest perpendicular tumor diameters were performed twice weekly to estimate tumor volume, using the formula: 4π/3 × (width/2)² × (length/2). Animals were killed when tumors reached 2 cm. Survival was evaluated from the day of inoculation until death or 60 days from the start of the study. Cox regression analysis was used to compare differences between groups treated with vehicle and PD173074.

Previous studies¹⁸  identified PD173074 as inhibitor of FGFR1. It is both potent, inhibiting FGFR1 at nanomolar concentrations, and in contrast to many other protein tyrosine kinase (PTK) inhibitors, it is also highly specific, inhibiting Src, the receptors for insulin, epidermal growth factor (EGF), vascular endothelial growth factor (VEGF), and platelet-derived growth factor (PDGF), as well as several serine/threonine kinases with 100-fold or greater inhibitory concentration of 50% (IC₅₀) values. We confirmed the activity of PD173074 against FGFR3 by demonstrating inhibition of autophosphorylation in cultured cells. NIH 3T3 cells transfected with WT FGFR3 were treated with increasing concentrations of PD173074 and stimulated with aFGF and heparin. PD173074 inhibited autophosphorylation of FGFR3 in a dose-dependent manner with an IC₅₀ of approximately 5 nM (Figure 1A). Parental NIH 3T3 cells included, as a negative control, did not have detectable FGFR3 (data not shown).

In a previous study we reported that the same constitutively active mutants of FGFR3 found in MM when expressed in NIH 3T3 cells induce focus formation suggesting an oncogenic role for activated FGFR3 in MM.¹¹  The efficacy of PD173074 in blocking FGFR3-mediated transformation was assayed in this model system. We transfected NIH 3T3 cells with WT FGFR3, activated Y373C FGFR3, or constitutively active Ras V12 as a control for specificity. Consistent with our previous experiments, WT FGFR3 failed to transform NIH 3T3 cells, whereas numerous foci were evident with both Y373C and Ras V12 (Figure 1B). Treatment with 25 nM PD173074, on the other hand, completely abolished NIH 3T3 transformation mediated by Y373C FGFR3 but not by Ras V12, demonstrating that PD173074 specifically targets FGFR3-mediated cell transformation and lacks nonspecific cytotoxic effect.

Similarly, PD173074 inhibited in vivo growth of mutant FGFR3-transfected NIH 3T3 cells in nude mice. Each of 10 mice received subcutaneous injections of Y373C FGFR3 and Ras V12-transfected cells on opposite flanks. Mice were treated with either vehicle or PD173074 for 9 days. Twelve to 14 days after the injection, progressively growing tumors appeared on both flanks in all placebo-treated mice but only at site of Ras V12 inoculation in the PD173074-treated mice (Figure 1C). The mice were killed when tumor volumes reached 2 cm, limiting follow-up to a median of 22.5 days. No tumor was histologically evident on the FGFR3-injected flank by this time point. Taken together, the results indicate that PD173074 suppresses the transforming ability of activated FGFR3 and suggests that suppression of transformation by PD173074 is mediated through its ability to inhibit FGFR3 kinase activity.

To extend these studies to a more clinically relevant model system, the effect of PD173074 inhibition of FGFR3 on the phenotype of FGFR3-expressing myeloma cells was examined. We first confirmed that PD173074 could directly inhibit the tyrosine kinase activity of mutant and WT FGFR3 receptors expressed as a result of a t(4;14) translocation in MM cell lines. We have reported previously that 3 of 10 FGFR3-expressing t(4;14) MM cell lines have activating mutations of FGFR3 identical or similar to those reported in thanatophoric dysplasia.^11,23  KMS11 cells express the Y373C mutation that activates FGFR3 by promoting constitutive dimerization in the absence of the ligand. OPM-2 cells express the K650E mutation that occurs within the kinase domain of the receptor and is thought to relieve the inhibitory conformation of the kinase enabling its constitutive activation. KMS18 cells have a G384D mutation of FGFR3; transfection studies suggest that this mutation is weakly transforming. In contrast, H929 cells express WT FGFR3 but in addition have acquired an activating mutation of N-ras (N13). The level of FGFR3 phosphorylation in each of these serum-deprived cell lines was measured in the absence or presence of aFGF. As expected, autophosphorylation of WT FGFR3 was only observed on ligand stimulation, whereas mutant receptors (Y373C, K650E, and G384D) exhibited varying degrees of ligand-independent phosphorylation with significant augmentation following addition of aFGF (Figure 2A). Potent inhibition of FGFR3 tyrosine phosphorylation was seen in all cell lines pretreated with 5 and 25 nM PD173074. Thus PD173074 appears to be active against the WT receptor and FGFR3 mutations identified in MM.

We have previously shown that aFGF stimulation of serum-deprived FGFR3-expressing cell lines activates the mitogen-activated protein kinase (MAPK) pathway.¹¹  On the basis of this observation we hypothesized that aFGF would stimulate proliferation of these cells and that PD173074 would block this effect. A panel of FGFR3⁺ cell lines, expressing WT (H929, LP1, UTMC2) or mutant (OPM-2, KMS18, KMS11) FGFR3 and cells lacking FGFR3 expression (EJM, MM1, 8226) were stimulated with aFGF/heparin and incubated with increasing (0-100 nM) doses of PD173074 (Figure 2B). To restrict growth responses to those induced by aFGF, this experiment was performed in serum-free media. Cell growth induced at 48 hours was determined by MTT assay. As predicted, FGFR3-expressing cell lines responded to aFGF with a 1.8- to 2-fold increase in optical density (OD) reading over nonstimulated cells. This growth was inhibited by PD173074 in a dose-dependent manner. In contrast, MM cell lines lacking FGFR3 expression did not demonstrate enhanced growth in response to aFGF, and the inhibitor had no effect on cellular proliferation. These results confirm that FGFR3 can activate signaling pathways in MM cells that play a role in cell proliferation including MAPK phosphorylation, which was inhibited by exposure to PD173074 (data not shown). Furthermore, they demonstrate that inhibition of aFGF-stimulated growth by PD173074 is highly correlated with the expression of FGFR3.

To establish that FGFR3 activation provides critical and nonredundant proliferative and antiapoptotic signals in MM the following experiments were performed in the presence of full serum. We first tested the effects of FGFR3 inhibition on a panel of human MM cell lines in short-term viability assays (Figure 3). Cells were incubated with increasing concentrations of PD173074 (0-100 nM) and percentage of viable cells was determined by MTT. PD173074 reduced viability of FGFR3-expressing KMS11 and KMS18 cells with 50% inhibition achieved at 20 nM or less. H929 cells, on the other hand, demonstrated inactivation of FGFR3 autophosphorylation in response to PD173074, but showed only modest reduction in viability. Thus, H929 viability is not dependent on FGFR3 signaling. This is most likely the result of having acquired an IC (N-ras) codon 13-activating mutation that seems to fulfill an analogous role to mutant FGFR3 in MM.¹⁰  Similarly, only a modest reduction in viability was observed for the FGFR3-expressing cell lines UTMC2, LP1, and OPM-2. We cannot account for the lack of response of LP1 and UTMC2 to PD173074; however, it is likely that these cells have acquired additional genetic mutations that have rendered them independent of FGFR3 as is evident by their ability to grow in culture in the absence of FGF. On the other hand, OPM2 cells have both a Ras-activating mutation and biallelic PTEN deletions.²⁴  Loss of PTEN has recently been shown to counteract the antitumor activity of EGFR inhibitors in various cell types.²⁵  The 8226, MM1, and EJM cells, which lack FGFR3 expression, displayed no decrease in viability, demonstrating that at effective concentrations, PD173074 exhibits minimal nonspecific cytotoxicity on myeloma cells (Figure 3B).

To further characterize whether the lesser number of viable KMS11 and KMS18 cells observed after 48 hours of treatment with PD173074 was the result of decreased proliferation or induction of apoptosis, we first analyzed the potential of PD173074 to mediate cell cycle arrest. The DNA content of cells exposed to 50 nM PD173074 was analyzed by EtBr staining. After 48 hours, PD173074-treated KMS11 and KMS18 cells exhibited an increase in the percentage of G₀/G₁ cells (74.33% ± 1.96% and 71.33% ± 3.36%, respectively) in comparison with controls (48.96% ± 2.86% and 45.66% ± 3.04%, respectively). A reciprocal reduction in the percentage of cells in S phase was observed (Figure 4A). Consistent with the data from the MTT assay, FGFR3 inhibition in H929, UTMC2, LP1, and OPM-2 resulted in only a marginal increase in the percentage of G₀/G₁ cells and we saw no effect of PD173074 on the DNA content of 8226, MM1, and EJM cells (data not shown). These data demonstrate that inhibition of FGFR3 in KMS11 and KMS18 cells induces inhibition of cell growth related to an arrest at the G₀/G₁ phase of the cell cycle.

Subsequently, we evaluated whether FGFR3 activation further provides antiapoptotic signals in MM cells. To assess for apoptosis, cells exposed to PD173074 for up to 14 days were sampled at intervals and annexin V binding was monitored by flow cytometry. Interestingly, both cell lines displayed markedly delayed cytotoxicity in response to PD173074. An increase in the number of apoptotic cells over that of controls was apparent 4 (KMS11) and 10 days (KMS18) after exposure to PD173074 and the percentage of apoptotic cells increased with dose and days of treatment (Figure 4B). This progressive increase was still observed when the cells were washed free of PD173074 on day 3 (but not day 1 or 2) and placed back into culture for 4 days (data not shown). On the other hand, H929 and 8226 cells similarly treated with PD173074 for up to 10 days did not demonstrate an increase in percentage of annexin V⁺ cells over vehicle-treated controls (data not shown). To confirm that KMS11 cells were dying of apoptosis, we performed immunohistochemical analysis for cleaved caspase 3 on clot sections of cultured cells. Cells treated with vehicle alone exhibited a low frequency of background apoptosis consistent with the annexin V data. In contrast, at 6 days after FGFR3 inactivation KMS11 cells exhibited a marked increase in apoptosis (Figure 4C). Comparable results were obtained when KMS18 cells were similarly analyzed after 14 days of treatment.

The marked delay in apoptosis observed following PD173074 treatment of MM cells is unusual when compared to dexamethasone- and chemotherapy-induced cell death, which usually occurs within 2 days.²⁶  We hypothesized that the inhibition of FGFR3 allowed for differentiation of these cells to occur. Indeed, induction of differentiation is generally associated with a loss of proliferative capacity and eventually cell death.²⁷  On morphologic examination KMS11 and KMS18 cells have an immature blastlike appearance with fine chromatin, multiple nucleoli, irregular nuclei, and high nuclear-cytoplasmic ratio (Figure 5A). In contrast, cells treated with PD173074 displayed increased cytoplasm, more condensed chromatin, eccentric nuclei, and in some cells, a perinuclear hauf, classic features of PCs. In parallel, we assessed for changes in cell surface expression of markers associated with increasing PC maturity. Immunophenotyping was performed using a panel of markers that are either downregulated (CD45, CD19, CD20, and HLA-DR) or up-regulated (CD27, CD31, CD38, CD79b, and CD138) on terminally differentiated PCs. Of these antigens, CD27, CD31, and CD79b were aberrantly absent on KMS11 and KMS18 cells. Exposure of these cells to PD173074 up-regulated surface expression of CD31 in the majority of treated cells (Figure 5B). Although the appearance of CD31 may represent a mechanism other than differentiation, recently published microarray data, comparing CD19-enriched B cells, tonsil PCs, and BM PCs, identified CD31 (platelet endothelial cell adhesion molecule [PECAM]) as one of the genes most tightly linked with early and late PC differentiation.^28,29  We also observed a 2-fold increase in CD27 mRNA but failed to detect a significant increase in its surface expression by flow cytometry (data not shown).

We next assessed for evidence of functional maturation. B-cell terminal differentiation is characteristically associated with the onset of high-level antibody secretion. We found that KMS11 and KMS18 secreted measurable amounts of κ and λ, respectively, and that the levels of light-chain mRNA on a Northern blot (data not shown) and titer of secreted light chain increased significantly after exposure to PD173074 (Figure 5C). These data indicate that PD173074 induces functional maturation of KMS11 and KMS18 cells. No effects on cell cycle distribution, apoptosis, morphologic, or phenotypic measures of differentiation were seen on treatment of an FGFR3^- MM cell line, 8226 (data not shown).

To determine whether FGFR3 inactivation had similar effects in vivo, we examined the effect of PD173074 on the growth of KMS11 cells in the subcutaneous plasmacytoma xenograft mouse model²¹  that has been used in preclinical studies of PS-341 and immunomodulatory analogs (IMiDs) in MM.^21,22  Consistent with our in vitro data, FGFR3 inhibition induced differentiation of the tumor cells (Figure 6). Tumors harvested from placebo-treated animals contained predominantly CD31^- cells with a marked blastic appearance as well as numerous mitotic figures. These findings contrast with the areas of necrosis and scattered, admixed, mature-appearing, CD31⁺ PCs observed in PD173074-treated mice. Similarly, the inactivation of FGFR3 was associated with a decreased proliferative capacity and increased apoptosis as assessed by immunohistochemical analysis for the nuclear proliferation antigen Ki-67, and cleaved caspase 3, respectively.

This was associated with delayed tumor progression and enhanced overall survival. The average tumor volume was 83.7 mm³ at the time of treatment initiation when mice were assigned to receive either placebo or 25 mg/kg PD173074 by oral gavage twice daily for 9 days. Mice treated with drug showed statistically significant (P < .002) inhibition of tumor growth as compared with controls (Figure 7A). Latency was prolonged in excess of that of the duration of PD173074 therapy and complete tumor regression was observed in 3 mice. Regrowth of tumors in these mice, however, occurred prior to the study end point suggesting that PD173074 failed to eradicate FGFR3-expressing cells at the concentration and dosing used in this study. These results were not unexpected because experience with imatinib mesylate in animal models suggests that continuous exposure to drug is crucial to achieve complete remissions.³⁰  Given the short half-life of PD173074 in mice this would have required 4 times daily dosing. The duration of treatment was limited to 9 days because in earlier studies this was found to be the maximum tolerable duration in mice (C. Omer, Pfizer, verbal communication, September 2002). Figure 7B shows the survival curves of mice treated with either placebo or PD173074. Despite suboptimal dosing, Kaplan-Meier analyses for survival of placebo- and drug-treated animals demonstrates significantly prolonged survival of the treatment group (P < .0003). As a control for specificity a total of 20 mice were subcutaneously injected with 8226 cells and similarly treated with placebo or PD173074. The antitumor effect of PD173074 was specific for the cell line with activated FGFR3 because we saw no examples of tumor regression and no evidence of tumor delay or enhanced survival in 8226 xenograft mice treated with PD173074 (data not shown).

The recent clinical success of the BCR-ABL inhibitor imatinib in chronic myelogenous leukemia (CML) suggests that therapies targeting activated kinases may be an effective treatment approach in certain malignancies.¹  Inhibition of gain-of-function mutations of c-Kit tyrosine kinase in gastrointestinal stromal tumors has been particularly effective against this chemoresistant tumor.³¹  A similar kinase inhibitor strategy targeting activated FLT3, present in approximately 30% of patients with acute myelogenous leukemia (AML), is also showing promising results in preclinical studies.^32-34  We have previously shown that FGFR3 is overexpressed in approximately 15% of patients with MM and that approximately 10% of these acquire activating mutations of this receptor.³⁵  Myeloma cell lines with activated FGFR3 appear to be dependent on activated FGFR3, displaying exquisite sensitivity to inhibition of FGFR3 both in vitro and in vivo. The data presented here support the therapeutic strategy of targeting activated FGFR3 in t(4;14) MM. Although potentially providing a therapeutic benefit to only a minority of patients with MM, the results also lend support to the idea of targeting other dysregulated tyrosine kinases in lymphoid malignancies (eg, the fusion protein NPM-ALK^36,37 ), and other activated FGFRs involved in human cancers (eg, FGFR3 in bladder cancer,³⁸  and BCR-FGFR1 or ZNF198-FGFR1 in myeloproliferative disease³⁹ ).

It is formally possible that the responses observed could be mediated through the inhibition of additional cellular kinases that have been implicated in myeloma including VEGF, insulin-like growth factor 1 (IGF-1) receptor, and MAPKs.⁴⁰  However, the concentration of PD173074 required to inhibit these kinases is 100-fold higher than that which induced apoptosis of FGFR3-expressing cells. In addition, the specificity of PD173074 for FGFRs was demonstrated by the lack of effect on FGFR3^- cell lines and Ras-transformed NIH 3T3 cells. Finally, SU5402, which has a different selectivity profile for PTK inhibition than PD173074, also inhibits FGFRs and similarly induces cytotoxic responses in FGFR3-expressing cell lines.⁴¹  Taken together, these findings suggest that FGFR3 inhibition is responsible for the cytostatic and cytotoxic effects observed in KMS11 and KMS18 cells.

It is of interest to note that responses were achieved only in cells expressing activated mutant forms of FGFR3. This is not altogether surprising because the cell lines that express WT FGFR3 have already established themselves as ligand independent by virtue of their ability to grow in vitro. Although it is possible that extremely high levels of FGFR3 expression or an autocrine loop could render WT cells independent of exogenous ligand, none of the WT FGFR3-expressing cell lines demonstrated ligand-independent phosphorylation (data not shown). Rather, we have previously shown that many of the cell lines expressing WT FGFR3 have acquired secondary mutations of genes downstream of the FGFR3 pathway, most notably Ras.¹¹  In addition, both a Ras-activating mutation and biallelic PTEN deletion have recently been described in OPM-2, the one cell line with activated FGFR3 that is insensitive to PD173074.²⁴  These data suggest that FGFR3 inhibitors may not be of clinical benefit in the patient population with advanced disease who have acquired Ras mutations rather than activating FGFR3 mutations. On the other hand, fewer than 5% of premalignant monoclonal gammopathy of undetermined significance (MGUS) tumors have Ras mutations (T. Rasmussen, unpublished data, 2003). We cannot therefore draw firm conclusions regarding the therapeutic value of targeting WT FGFR3 in patients with MGUS or early stage t(4;14) MM based on our analysis in human MM cell lines, nor can we say to what extent primary MM cell growth in a patient's BM is dependent on ectopic FGFR3. Furthermore, studies of FLT3 inhibitors in AML cell lines have similarly demonstrated activity only in cells with activated FLT3; however, when primary patient AML blasts were tested some samples with WT FLT3 responded, reflecting the inherent limitations of using cell lines that have acquired multiple genetic mutations.³²  Studies using primary samples to directly address this issue are limited by the inability of primary MM cells to proliferate in vitro or to be maintained for the 5 to 10 days required to see effects. A clearly more meaningful model for the study of primary MM cells is the elegant SCID-hu model whereby primary MM cells may be grown in the context of a human fetal BM microenvironment.⁴²  This model, on one occasion, has even been used to demonstrate the anti-MM activity of a drug (thalidomide).⁴³  Unfortunately it is not a suitable model to validate PD173074, which can only be given for 9 days and will not be developed for clinical use.

Our results suggest that FGFR3 in the postgerminal center cell prevents it from exiting the cell cycle and completing a normal differentiation program. The modest markers of differentiation induced in FGFR3-expressing MM cell lines in response to PD173074 may simply be an obligatory response to growth arrest. Serum starvation, however, failed to induce morphologic differentiation despite arresting the cells in G₁ (data not shown). Although morphologic evidence of differentiation is seen, other evidence of differentiation is weak. The only modest increase in light-chain secretion and lack of induction of other markers of PC differentiation other than CD31 implies that inhibition of FGFR3 in itself may not be sufficient for induction of complete functional differentiation. This is not entirely surprising given the complexity of genetic alterations that characterize these cell lines. These observations, however, implicate the existence of a malignant self-renewing precursor cell (immortalized as a cell line). These pre-PCs may serve as a reservoir of MM precursors, capable of expanding and differentiating into malignant PC tumors and contributing to patient relapses.

When KMS11 cells were washed free of inhibitor after 3 days of exposure to PD173074 and placed back into culture, we found that the cells demonstrated continued commitment to maturation and apoptosis. FGFR3 therefore may not directly induce apoptosis. We hypothesize that even a modest degree of differentiation induced by PD173074 is sufficient to reverse the malignant phenotype rendering the cell sensitive to apoptotic signals and to a limited life span as is seen in terminally differentiated PCs. This observation has been noted in HL60 cells exposed to retinoic acid²⁷  and most recently in a MYC transgenic mouse tumor model whereby myc inactivation resulted in tumor cell differentiation, followed by apoptosis on subsequent MYC reactivation.³  Although speculative, this would have profound implications for disease treatment, much akin to that of the combination of all-trans-retinoic acid (ATRA) and chemotherapy for acute promyelocytic leukemia (APL) where durable remissions have been achieved.⁴⁴ 

In summary, we have shown that inhibition of activated FGFR3 causes morphologic differentiation followed by apoptosis of FGFR3-expressing MM cell lines. This demonstrates that despite the presence of multiple genetic mutations, and the acquired ability to grow independently of the BM microenvironment, these cells remain dependent on the primary oncogenic event. The results presented here validate activated FGFR3 as a therapeutic target in t(4;14) MM and encourage the clinical development of FGFR3 inhibitors for the treatment of patients who fall within a poor prognosis group and who are in need of novel treatment options.

We thank Charles Omer for useful discussions and for supplying PD173074 and Mike Kuehl for many useful and stimulating discussions, and we are grateful to Richard LeBlanc for sharing details of the xenograft model.