Phenotypic and molecular analysis of six human cell lines derived from patients with plasma cell dyscrasia


Dr Roger Gooding, The Myeloma Unit, Department of Haematology, Imperial College School of Medicine, Hammersmith Hospital, 150 DuCane Road, London W12 ONN. e-mail:


Cell lines RPMI 8226, JJN3, U266 B1, NCI-H929 (all EBV) and ARH77 and HS-Sultan (both EBV+) have been extensively characterized in this study. EBV lines expressed the phenotype (CD138, CD19+, CD20+) whereas EBV+ were (CD138+, CD19, CD20). CD56 expression was restricted to EBV cell lines, with the exception of U266 B1, whereas PCA-1 was strongly expressed on five of the six cell lines. Only EBV+ cell lines bound peanut-agglutinin (PNA). However, all cell lines bound the lectin Jacalin that binds the same receptor as PNA, irrespective of the receptors sialylation status. By RT-PCR and direct sequencing of their IgH V/D/J domains, ARH77 was demonstrated to use the germline sequence VH4-34/dm1/JH6b, whereas no arrangement was demonstrated for RPMI 8226, suggesting IgH gene deletion or mutation. HLA class I and II antigens were detected using HLA typing on all cell lines warranting their use as suitable targets for HLA-restricted cytotoxic T cells. By sensitive RT-PCR, mRNA for IL-6, IL-6R and TNFβ was found expressed in all cell lines. IL-1 mRNA expression was predominantly associated with the EBV+ phenotype. Although mRNA for IL-3 and GM-CSF was never detected, transcripts for c-kit ligand and, more commonly, its receptor were. Likewise GM-CSF, M-CSF and erythropoietin mRNA transcripts were detected in the majority of cell lines.

The phenotypic profile of malignant plasma cells is variable both between and within individual samples from patients with multiple myeloma (MM) (Kawano et al, 1995; San Miguel et al, 1995). Consequently, myeloma plasma cell isolation for in vitro study remains problematic. Because human plasma cell-like cell lines are assumed to reproduce at least some of the characteristics of primary myeloma cells, they represent a readily available alternative to the potentially heterogenous primary myeloma cell population.

Some doubt persists however, as to the eventual usefulness of using cell lines in decoding the signalling events between myeloma cells and the non-tumour cells of the bone marrow. Additional molecular events must occur before primary myeloma cells can be cultured indefinitely in vitro (Zhang et al, 1994; Shimizu et al, 1992). Cell lines are de-differentiated outgrowths of terminally differentiated mature myeloma plasma cells, with a subsequent morphological resemblance to lymphoblastoid cells and are potentially unrepresentative of myeloma cells in any stage of the disease. In addition, some myeloma cells are known to contain EBV nucleotide sequences. Further characterization of the original tumour is not possible, but without this it is uncertain as to whether the EBV transformed cells actually originate from the malignant clone (Ralph, 1985; Drewinko et al, 1985; Pellat-Deceunynk et al, 1995).

Myeloma cell lines have been a more valuable and reliable resource where direct conclusions relating to MM per se have not been sought. For example, investigations into the regulation of immunoglobulin (Ig) gene, the signalling requirements of differentiating germinal centre B cells and the mechanisms of post-transitional modifications of proteins. Furthermore, myeloma cell lines have been used as targets for anti-tumour chemotherapy, immunotoxins, oligonucleotide anti-sense strategies, and for investigating drug resistance. Finally, myeloma cell lines have a use in the development of human hybridoma systems.

Although myeloma cell lines are not representative of primary myeloma cells, their sharing of fundamental characteristics, i.e. Ig gene re-arrangement, cytoplasmic idiotype expression and the production of, or requirement for, a number of peptide regulators, e.g. IL-6, make them ideal for use in the development and optimization of therapeutic strategies such as graft purging and the semi-quantitative molecular monitoring of clonal tumour cells in patients after high-dose therapy. Furthermore, myeloma cell lines may have a use as model HLA-matched targets for cytotoxic T cells generated in vitro. This study therefore describes the phenotypic and molecular characterization of a number of readily available cell lines obtained from patients with plasma cell dyscrasia.


Cell lines

Details of the source of cell lines used in this study are shown in 1Table I. Cells were maintained in tissue culture medium (TCM) consisting of RPMI 1640, 50 μm 2-mercaptoethanol, 10 mm glutamine and 10% fetal calf serum (FCS). All cell lines tested negative for mycoplasma by PCR using the ATCC Mycoplasma Detection Kit (ATCC, Rockville, Md., U.S.A.). Cell line culture supernatants (5 × 105 cells/ml cultured for 4 d) were screened for the presence of secreted immunoglobulin. Briefly, 50 ml of conditioned TCM was freeze-dried and the product reconstituted in 1 ml water prior to assay for Ig by standard immunofixation and immunoturbidometry methods. Diagnostic reagents and equipment, for Ig detection were supplied by Bayer Diagnostic (Berkshire) and Helena Labs Ltd (Newcastle) respectively and used in accordance with the manufacturer's instructions.

Table 1. Table I. Human myeloma cell lines: source and origin (Matsuoka et al, 1967, Nilsson et al, 1970, Burk et al, 1978, Gazdar et al, 1986, Kozbor et al, 1983, Jackson et al, 1989). ECACC: European Collection of Animal Cell Cultures, Salisbury; ATCC: American Type Culture Collection, Rockville, Md., U.S.A.; PB: peripheral blood; LP: localized plasmacytoma; BM: bone marrow; PE: pleural effusion; PCL: plasma cell leukaemia; M: myeloma.Thumbnail image of

Immunostaining of phenotypic cell surface antigens

Antibody details are shown in 2Table II. Cells were stained using direct and indirect staining methods and fixed prior to cytometric analysis in 0.5 ml formyl saline (0.2% para-formaldehyde, 0.9% (w/v) NaCl) for a minimum period of 15 min. Isotype-matched antibodies were used for determination of non-specific staining.

Table 2. Table II. Antibodies used for the phenotyping of myeloma cell lines. PNA: peanut agglutinin; mAb: monoclonal antibody; pcAb: polyclonal antibody; RPE: rhodomine phycoerythrin; FITC: fluoroscence isothiocyanate.* Host animal mouse unless otherwise indicated.Thumbnail image of

Immunostaining of lectin receptors

Cells were washed in Hanks calcium and magnesium-free buffer containing 10 mm hepes and incubated at 106 cells/400 μl in the same buffer containing peanut-agglutinin (PNA) at 10 mg/ml (Sigma, St Louis, Mo.). Incubation was for 1 h at 4°C with occasional mixing after which unbound PNA was removed by washing the cells twice in hepes-buffered Hanks containing 1% (w/v) bovine albumin (BA). PNA bound to cell surface receptors was detected by incubating the cells for a further 30 min at 4°C in 200 μl rabbit anti-PNA polyclonal antibody diluted 1 in 50. Cells were washed twice, suspended in 100 μl swine anti-rabbit polyclonal antibody conjugated to fluoroscein isothiocyanate (FITC), previously diluted 1 in 250, and incubated for a further 30 min at 4°C. Cells were washed and fixed as previously described. For determination of non-specific staining, normal rabbit sera, comparably diluted, were substituted for the anti-PNA rabbit polyclonal antibody. Rhodamine-conjugated Jacalin (Vector Laboratories, Burlingame, Calif.) at 3 μg/200 μl in hepes-buffered saline containing 0.1 mm Ca2+ was used to stain 5 × 105 cells. Specific staining was demonstrated by blocking reactions using 800 mm galactose in replicate staining reactions.

Cytometric analysis was performed using a Becton Dickinson FACScan running Cell Quest data acquisition and analysis software. Immunostained cells which demonstrated >10% of their numbers with a fluorescence intensity greater than the brightest 2% of a non-specifically stained identical cell population were considered positive for the stained-for antigen.

Detection of immunoglobulin gene rearrangement by Southern blot analysis

DNA was prepared from each cell line by cell lysis, proteinase K digestion, phenol extraction, and ethanol precipitation. 5–10 μg of DNA were digested with the restriction enzymes HidIII and BglII, electrophoresed in a 0.8% agarose gel, denatured, neutralized, transferred to nitrocellulose filters, and hybridized. The probe, a 32P-labelled fragment of the Ig loci representative of J region of the Ig heavy chain, was prepared as previously described (Vulliamy & Kaeda, 1995).

Isolation of total RNA and reverse transcription

5 × 106 cells were lysed in 500 μl guanidium thiocynate solution (GITC) and stored at −20°C. Total RNA was isolated using phenol/chloroform/isoamyl alcohol extraction (Chomczynski & Sacchi, 1987). The precipitated RNA was washed in 80% ethanol, dissolved in 20 μl diethylpyrocarbonate-treated sterile water, and reverse transcribed into cDNA in a 50 μl reaction primed with random hexamers (pdN6, Pharmacia) or oligo- (dT)12–18 primer (GibcoBRL) using standard methodology. cDNA was stored at −20°C.

Immunoglobulin heavy chain sequence amplification and sequencing

Target sequences were amplified by PCR using primers specific for immunoglobulin heavy chain sequences as described elsewhere (Deane & Norton, 1991; Aubin et al, 1995). The 3′ primer 5′acgggatccgctcagcgggaagacctt3′, specific to the IgA constant region, was substituted for JH primers in the amplification of NCI-H929 IgH due to the absence of J sequence in this line. Extensive precautions were taken to avoid contamination of PCR reactions. Reaction products were electrophoresed on a 1.2% agarose gel containing ethidium bromide and bands were visualized using UV illumination. Amplification products were directly sequenced using the Dye Terminator Cycler Sequencing Ready Reaction Kit (Perkin Elmer) on an automated sequencer (ABI Prism). Sequence data were confirmed by comparing sequences obtained from both the 5′ and 3′ sequencing directions. The identical sequences obtained from a minimum of three sequencing runs was then compared with the most homologous germline genes in a data base of germline sequences available in the University of Wisconsin GenBank (Madison, Wis.) and supplied by the Human Genome Mapping Project Resource Centre (Hilton, Cambridge).

Cytokine and receptor sequence amplification

2.5 μl cDNA as template was mixed with 47.5 μl PCR mix (cytokine PCR mix = 20 mm Tris-HCl (pH 8.4), 50 mm KCl, 5 mm MgCl2, 0.25 mm each of dATP, dCTP, dTTP and dGTP, 0.25 mm each of the cytokine primer pairs (Table III), 25 units/μl Taq polymerase). PCR was performed on a MJ Research Inc. programmable heating block by 41 or 30 cycles of 90°C 1 min, 65°C 1 min, 72°C 1 min followed by a 10 min extension at 72°C. To increase the specificity of the reaction, tubes were placed in the thermocycler only when it had reached the denaturing temperature of 95°C and were held at this temperature for 3 min before cycling started. PCR products were analysed as described previously and gels photographed. A PCR reaction was considered positive if a band of the appropriate size was visible in the appropriate control lane (phytohaemagglutinin (PHA) stimulated peripheral blood mononuclear cells (PBMC) and/or 5637 bladder carcinoma cell line cDNA) and the negative controls did not give a signal. PCR reactions were considered negative only if the positive control gave an appropriately-sized band in an ethidium bromide-stained agarose gel, whereas the negative controls did not.

Table 3. Table III. Primer pairs. Sequences are in the 5′-3′ orientation.Thumbnail image of

Primer pairs for amplification of, interleukin (IL)-1β, IL-3, IL-6, c-kit, kit ligand (KL), granulocyte-colony stimulating factor (G-CSF) macrophage-colony stimulating factor (M-CSF), granulocyte-macrophage-colony stimulating factor (GM-CSF) and transforming growth factor (TGF)-β gene sequences were as previously described (Thalmeier et al, 1996). Primer pairs for amplification of IL-α, IL-11, erythropoietin (Epo), hepatocyte growth factor (HGF), tumour necrosis factor (TNF)-α, TNF-β/lymphotoxin, (LT) and IL-6 receptor (R) gene sequences were designed to amplify across intron/exon boundaries and have an annealing temperature of 65°C. Products of these amplifications were further analysed by direct sequencing, confirming their identity. Further primer details are shown in 3Table III.

Histocompatability locus antigen (HLA) typing of cell lines

Serological HLA typing by microcytoxicity was as follows: 300 μl of a growing cell suspension (106/ml) was stained with 6 μl 1% 5–6-carboxyfluorescein diacetate (Molecular Probes) in acetone for 30 min at 37°C. The cells were pelleted for 2 s in a Beckman microfuge and resuspended at approximately 106/ml in Hepes-buffered RPMI 1640 plus 20% heat inactivated FCS. 1  μl of the cell suspension was added to 1 μl of mAb in a Robbins 72-well typing tray, using a Lambda Jet (One Lambda) and incubated for 30 min at 25°C. Then 5 μl rabbit complement (Pel Freeze) was added, also using the Lambda Jet, and incubated for a further 60 min before addition of 1 μl ethidium bromide (450 μg/ml). Plates were read by eye using a Leitz Diavert fluorescence microscope (Marsh et al, 1990). In addition to locally available monoclonal antibodies, cells were also typed using One Lambda Special Monoclonal Typing trays (One Lambda).

For molecular typing, HLA-A, B and C typing was performed by direct sequencing of cell line DNA as described previously (Prokupek et al, 1998; Scheltinga et al, 1997; unpublished observations). In some cases, where allele drop out was suspected as the cause of homozygosity of a particular HLA type, DNA from the cell line was typed for HLA-A and B by SSO typing using the Lifecodes PCR SSO typing kit (Lifecodes Corporation, U.S.A.) and HLA-C typing was performed by SSP using Dynal SSP typing kit (Dynal). HLA-DRB1 and DQB1 typing was performed using AMPLICOR reverse line blot assay (Roche Diagnostic Systems, Inc.). HLA-DP typing was performed by direct sequencing of PCR products as described previously (Verluis et al, 1995). HLA typing was confined only to those cell lines in which a productive rearrangement of the IgH gene was detected. HLA peptide motif search was performed using on-line software available at the web site http://www-bimas.dcrt. (Parker et al, 1994).


Culture morphology of human myeloma derived cell lines

Myeloma cell lines were maintained using standard culture conditions and passaged according to cell density. Most of the cell lines exhibited some degree of pleomorphism with variation in cell size and degree of hetero- and homotypic adhesion. Morphologically, the cells resembled lymphoblastoid cells.

Immunophenotype of myeloma cell lines

Cytometric data are summarized in 4Table IV. Surface immunoglobulin (sIg) expression was demonstrated on a number of the cell lines. In particular, ARH77 and JJN3 demonstrated 87% and 49% positivity for κ-type light chain, respectively, in agreement with previous reports (Drewinko et al, 1985; Pellat-Deceunynk et al, 1995; Jackson et al, 1989). RPMI 8226 stained weakly positive for λ-type light chain expression, with an increase in mean fluorescence intensity (MFI) from 2.05 to 5.71 between control and βl chain-positive cells. U266 B1 and HS-Sultan did not stain for surface light chain expression as previously published (Pellat-Deceunynk et al, 1995). Surface light chain expression was also absent from NCI-H929 previously reported to weakly express surface κ-type light (Gazdar et al, 1986; Pellat-Deceunyk et al, 1995). Attempts to demonstrate Ig present in the supernatant of myeloma cell line cultures using immunochemical methods were unsuccessful (data not included).

Table 4. Table IV. Phenotypic analysis of myeloma cell lines. Results were obtained using immunostaining and flow cytometric analysis. Cells were scored negative (−) for expression of surface antigen if less than 10% of cells stained positive compared to control antibody stained cells. Cells were scored positive if greater than 10–50% (+), 50–90% (++) or 90–100% (+++) of cells stained positive compared to control antibody-stained cells.Thumbnail image of

Strong expression of B lymphocyte differentiation-associated antigens CD10, CD19 and CD20 were confined to HS-Sultan and ARH77, both of which failed to express CD56. These data confirm and extend the phenotype data previously reported (Barut et al, 1992; Pellat-Deceunyk et al, 1995; Anderson et al, 1983). Syndecan-1 expression, detected using the mAb B-B4, was however absent on HS-Sultan and ARH77 although others have previously found both cell lines to stain weakly positive using the same antibody (Pellat-Deceunyk et al, 1995). U266 BI, RPMI 8226, JJN3 and NCI-H929 expressed different degrees of plasma cell-associated antigens (PCA-1, CD38, CD56, syndecan-1) expression in the absence of B-lymphocyte antigens (Table IV).

Myeloma cell lines were also investigated for their lectin binding ability (Table V). Unexpectedly, the binding of PNA in a very high proportion of cells (> 90%) was a finding common to those lines expressing lymphocyte differentiation antigens. Only one of the four cell lines expressing plasma cell surface antigens, RPMI 8226, bound PNA at a comparable proportion (93%). The matrix attachment molecule, and common receptor for PNA binding, CD44 (Slupsky et al, 1992), was detectable on all of the cell lines not expressing CD10, albeit very weakly on RPMI 8226. HS Sultan has previously been reported to express CD44 (Hiroshi et al, 1992). Jacalin, a lectin isolated from Artocarpus integriflora (Jackfriut) seeds, was used to test for the presence of the glycosidically linked oligosaccharide galactosyl (β-1,3) N-acetylgalactosamine. This structure, the so-called ‘T-antigen’, is the oligosaccharide to which PNA binds. However, unlike PNA, Jacalin will bind the T-antigen even in a mono- or disialylated form. All cell lines were > 90% positive for Jacalin binding and this binding could be inhibited by 800 mm galactose.

Table 5. Table V. Lectin binding. PNA: peanut agglutinin.Thumbnail image of

DNA analysis of myeloma cell lines

IgH gene rearrangements were identified in five of the six cell lines using Southern blot analysis (Fig 1). Probing of RPMI 8226 DNA failed to demonstrate any evidence of IgH gene rearrangement suggesting that all or part of the IgH gene sequences have been deleted in this cell line.

Figure 1.

. Southern analysis of IgH gene in myeloma cell lines. DNA was digested with Hind III or Bgl II, electrophoresed on a 1% agarose TBE gel overnight and transferred to a nylon membrane. IgH sequences were located by hybridization to a 32P-labelled probe specific to the joining sequence of the IgH gene and visualized by autoradiography. Arrows indicate the position of germline bands obtained using DNA from control PBMCs.

Sequence analysis of myeloma cell line

Using cell line cDNA, the VH family usage by each cell line was determined by VH family specific PCR amplification of the Ig H sequence. cDNA from RPMI 8266 failed to yield an amplified product with any of the VH family specific primers in combination with a JH consensus primer, providing further evidence for the existence of an aberrant or deleted IgH gene sequence in this cell line. NCI-H929 failed to produce a PCR product when using the primer combination VH3/JH consensus. IgH PCR product for NCI-H929 was eventually obtained using the primer combination VH3/Cα. In all cases where a PCR product was found, it was directly sequenced from both the 5′ and the 3′ directions. Cell line IgH sequences were analysed for homology with known germline VDJH sequences (Fig 2). U266 B1 showed a rearrangement using a germline sequence belonging to VH1 (Shin et al, 1991), both NCI-H929 and HS-Sultan to VH3, and ARH77 and JJN3 to VH4a (Tomlinson et al, 1992). Although limited homology with the DH sequences dxp4, dm1 and dk1 (Rabbits, 1983) could be identified in U266 Bl, ARH77 and JJN3 sequences respectively, no previously identified DH sequences matched HS-Sultan and NCI-H929 sequences. Four of the five IgH sequences could be defined as using a recognizable JH germline sequence: U266 Bl and JJN3 using JH4a (Ravetch et al, 1981), and ARH-77 and HS-Sultan using JH6b and JH6c respectively (Matilla et al, 1995). Although no recognizable D and J sequences could be located within the NCI-H929 IgH sequence, the post framework 3 sequence of NCI-H929 was found to significantly match that of a previously reported IgH sequence (HGMP accession number: hsigh354p) (Matsuda et al, 1993).

Figure 2.

. IgH nucleotide sequences from human myeloma cell lines. Differences from previously published related VDJ sequences, where available, are shown in letters beneath the sequences. Homology is shown as dots. For U266 Bl, differences from the sequence reported by Kenten et al (1982) are also shown. Continuous lines show the position of primers used for PCR and sequencing reactions. Dashes within the CDR2 region are to facilitate comparison between the sequences and do not represent missing sequence.

Cell line HLA

HLA typing showed maintenance of both class I and II antigens on the myeloma cell lines, indicating that these cells could be used as targets in cytotoxicity assays employing HLA-restricted cytotoxic T cells (Table VI). Where cells were typed by both serological and molecular methods, in general the results correlated. However, some discrepancies between the results obtained by the two methods were seen in determinations of the HLA-B type of both JJN3 and U266 Bl. The serological method is regarded as more prone to false results, particularly when used in the typing of cell lines. Some cell lines, here NCI-H929, are directly sensitive to the complement intended to lyse cells that are specifically bound by anti-HLA mAbs.

Table 6. Table VI. HLA identity of cell lines with a detectable rearranged Ig H gene.Thumbnail image of

Cytokine profile of myeloma cell lines

RT-PCR studies demonstrated that the cell lines expressed mRNA transcripts for a number of interleukins and haemopoietic growth factors (Table VII). Amplification of cytokine message was attempted at low (31) and high (41) cycles, giving some qualitative information of message expression by individual cell lines. Both the IL-6 and IL-6R gene were transcribed by all cell lines (Fig 3a). Amplification of the M-CSF mRNA transcript produced detectable PCR product of the correct size (672 bp) and identity by sequence analysis from four of the six cell lines. A smaller 400–500 bp product was amplified from U266 Bl, and once sequenced, corroborated the observation of Nakamura et al (1989) that U266 Bl expresses a shortened M-CSF transcript. No message for M-CSF was detected from HS-Sultan (Fig 3b).

Table 7. Table VII. Cytokine mRNA expression by myeloma cell lines. Results were from 41 and, in parentheses, 31 cycles of amplification (nd: not done).Thumbnail image of
Figure 3.

. Forty-one cycle RT-PCR amplification of (A) IL-6 receptor (390 bp) and (B) M-CSF (672 bp) transcripts using cDNA from 1: HS-Sultan; 2: ARH77; 3: RPMI-8226; 4: JJN3; 5: U266 Bl; and 6: NCI-H929. PCR-positive control reactions used cDNA from 7: PMA stimulated blood MNCs; and 8: the human bladder carcinoma cell line 5637. Negative controls, 9 and 10, were a GITC blank and water respectively. The same cDNAs used for the IL-6R PCR were used for the M-CSF. A smaller 400–500 bp product was always found when amplifying M-CSF using cDNA from U266 Bl.

Forty-one cycle RT-PCR for the erythropoietin mRNA transcript amplified a predicted 447 bp product. Direct cloning and sequence analysis confirmed that the amplified product contained the identical sequence to erythropoietin as reported by Jacobs et al (1985) (data not shown). Occasionally, using cDNA from JJN3, HS-Sultan and RPMI 8226, a smaller-sized amplification product was detected on ethidium bromide stained gels. This smaller band was excised, re-amplified and directly sequenced (Fig 4). The 390 bp product was confirmed as having the erythropoietin sequence but lacking 57 bases. The missing sequence coded for 19 amino-acids contributing to the third α-helix, and a portion of the linking sequence to the fourth α-helix, of the erythropoietin molecule (Elliot et al, 1996).

Figure 4.

. Comparison of erythropoietin sequence from Jacobs et al (1985) with sequences obtained from PCR products amplified from three myeloma cell lines using erythropoietin-specific primers. Homology is shown as dots and differences as letters. Continuous lines donate primer binding sites. Differences at codons 53 and 152 do not result in amino acid changes. Sequence differences at codons 40 and 53 between the published erythropoietin sequence and those obtained from the myeloma cell lines result in threonine to arginine and methionine to valine substitutions respectively. Missing sequence coding for 19 amino acids occurs from codon 61–79 inclusive in all sequences generated from myeloma cell lines.


Human cell lines sharing characteristics with B-lineage cells obtained from patients with plasma cell dyscrasia may act as useful in vitro models for the study of multiple myeloma. To this end, other investigators have stressed the importance of cell surface phenotype in differentiating plasma cell-like cell lines from EBV-positive cell lines (Pellat-Deceunynk et al, 1995). In this study a clear distinction could be made between the EBV-negative lines RPMI 8226, JJN3, U266 Bl, NCI-H929 and the EBV-positive HS-Sultan and ARH77 simply on the basis of syndecan-1 (CD138) expression on the former and CD19 expression on the latter. Investigation of other B-cell-specific antigens (CD10 and CD20), surface light chain expression, and plasma cell antigens (PCA-1, CD38 and CD56) confirmed two phenotypically distinct cell types (Table IV).

The N-CAM CD56 was previously found on HS-Sultan (37%) and ARH77 (13%) (Hiroshi et al, 1992), but in our study was not detected on either cell line. PCA-1 has been found expressed on primary plasma cells from pleural effusions or peripheral blood from patients with myeloma, suggesting its expression is related to malignant cell escape from the bone marrow into the circulation (Hata et al, 1994).

These phenotypic data generally corroborate previous investigations, although differences in the level of specific staining may have resulted from the use of different antibodies and changes in cell lines with time as previously found (Kozbor et al, 1983; Barut et al, 1992). A failure to detect Ig in any cell line conditioned medium may reflect compromised Ig production, assembly and secretion. IgH gene deletion has probably occurred in RPMI 8226.

PNA binds to primary bone marrow plasma cells (Rhodes & Flynn, 1988; Slupsky et al, 1992), but three of the four EBV-negative lines did not bind PNA. Both EBV-positive lines bound PNA, a not unexpected finding as PNA binds to EBV transformed Burkitt's lymphoma cell lines (Galili et al, 1981). CD44 is a major cell surface ligand for PNA and its glycosylation may modulate this interaction (Slupsky et al, 1992). The incompletely sialylated form of CD44 binds PNA, whereas completely sialylated CD44 does not. Culture conditions may effect the sialylation profile of the cell lines thus altering PNA binding. HS Sultan, ARH77 and RPMI 8226 have been reported to express CD44 (43%, 97% and 98% positive cells respectively) (Hiroshi et al, 1992). We failed to detect CD44 expression on HS Sultan, plus decreased expression by RPMI 8226, suggesting either loss of CD44 expression or masking of the antibody binding epitope. The latter suggestion is more likely as all cell lines bound Jacalin, a lectin binding to the same structure as PNA but which is not influenced by adjacent sialyic acid residues.

We have successfully identified the VH family and VH gene used by five of the six cell lines. ARH77 has a rearranged VH 4–34 gene. The VH4–34 gene has rarely if ever been found to be associated with MM (Rettig et al, 1996), but is mandatory for the production of pathological Ig responsible for anti-erythrocyte autoimmune conditions such as cold haemagglutinin disease (Chapman et al, 1996).

In myeloma and MGUS naturally occurring idiotypic immunity has been described (Yi et al, 1995). Therapies using idiotype vaccination in the presence of growth factors to encourage HLA restricted T-cell cytotoxicity have been proposed (Österborg et al, 1998). Data presented here allow a rational approach to the investigation of idiotypic antigen presentation in the context of class I MHC molecules. HLA peptide motif search results (data not shown) demonstrated that the Ig coded for by the cell lines JJN3 and NCI-H929 did not contain peptide motifs which would associate with the HLA class I molecules expressed by these cell lines. Although not unexpected, these findings should be treated with caution as the software used to perform the HLA peptide motif search is largely based upon empirical data from conventional peptide elution studies and the theoretical likelihood of particular sequences binding to certain HLA molecules. Thus, findings obtained using this software should be corroborated using conventional methods, a task outside the scope of the present work. Regardless of these limitations, peptide motif search results identified the following peptide ALGQGLEWV from U266 Bl to have a high likelihood of association with HLA-A2, but this sequence was within the framework 2 region. This cell line sequence differs from that observed in the germline sequence 1-02 by a proline instead of leucine, at position 2 of the peptide. A comparable analysis of the entire germline Ig sequence of 1-02 failed to show any likely association of peptide motif with HLA-A2.

IL-6 is regarded as the prototype myeloma cell growth factor primarily because IL-6 knockout mice are incapable of developing plasma cell tumours (Hilbert et al, 1995). Barut et al (1992) previously reported the lack of IL-6 mRNA and protein expression by RPMI 8226 and HS Sultan, and no evidence of IL-6R mRNA or protein expression by ARH77 and HS Sultan. JJN3 was reported to be IL6 dependent, secreting low levels of IL-6 protein (Hamilton et al, 1990). Both U266 Bl and RPMI 8226 express the IL6-R (Lemoli et al, 1994) although controversy exists as to whether or not they produce IL-6 and are IL-6 dependent (Taetle et al, 1994; Barut et al, 1992; Schwab et al, 1991; Diamant et al, 1996). In this study IL-6 and IL-6R mRNA transcripts were detected in all cell lines. These discrepancies may be due to differences in sensitivity between the different detection methods or changes in cytokine production in culture with time.

Production of IL-1 may be of pathological significance in the biology of multiple myeloma mainly because of its ability to act as an osteoclast-activating factor and influence IL-6 production via PGE2. Healthy individuals express the gene for IL-1β (but not IL-1α) constitutively in their peripheral blood and bone (Cluitmans et al, 1995), but in patients with MM, the source of IL-1 has been controversial (Lichtenstein et al, 1989; Cozzolino et al, 1989; Borset et al, 1993). Recently Costes et al (1998) demonstrated IL-1β protein production in tumoural samples of the majority of patients with MM. In situ hybridization and immunohistochemistry defined myeloid and megakaryocytic cells as the major IL-1β producing cells. Production by MM plasma cells was at a significantly lower level. Furthermore, medullary but not circulating tumoural cells expressed the IL-1β gene and no IL-1β transcript was found in a panel of nine myeloma cell lines which included JJN3, RPMI 8226 and U266. Here, all EBV-negative cell lines were negative for IL-1 transcripts using 31 cycles of PCR. Increased cycle number enabled the detection of IL-1α and IL-1β transcript in RPMI 8226 and JJN3 respectively.

GM-CSF, M-CSF and erythropoietin, but not IL-3 or G-CSF mRNA transcripts, were expressed by the cell lines. Stem cell factor (c-kit ligand) transcript was detected in ARH77 and JJN3 alone, whereas the transcript for its receptor (c-kit) was detected in the majority of the cell lines. Recent studies have reported absence of c-kit on normal plasma cells and its presence on approximately one third of fresh myeloma plasma cells where binding to c-kit ligand is able to stimulate plasma cell growth (Lemoli et al, 1994; Ocqueteau et al, 1996). Absence of c-kit ligand mRNA in U266 Bl, and presence in RPMI 8226 is confirmed here (Lemoli et al, 1994).

We report for the first time that the majority of the putative myeloma cell lines derived from patients with plasma cell dyscrasia express erythropoietin mRNA. Expression of erythropoietin receptors on a human myeloma cell line has been reported previously (Okuno et al, 1990), and their expression on myeloma cells in vivo has been linked to disease progression to plasma cell leukaemia in a patient who received erythropoietin therapy (Olujohngbe et al, 1997). Our additional finding that some of the cell lines also express an erythropoietin mRNA transcript variant at low level is also novel, and further work is warranted to understand the significance of this finding.


The authors are grateful for the excellent technical services provided by Steven T. Cox, Frieda Jordan and Anila Shah, Tissue-typing Laboratoy, Anthony Nolan Bone Marrow Trust. This work was supported by the Leukaemia Research Fund.