ATIC-ALK: A Novel Variant ALK Gene Fusion in Anaplastic Large Cell Lymphoma Resulting from the Recurrent Cryptic Chromosomal Inversion, inv(2)(p23q35)

Abstract

The subset of CD30-positive anaplastic large cell lymphomas (ALCL) with the NPM-ALK gene fusion arising from the t(2;5)(p23;q35) forms a distinct clinical and prognostic entity. Recently, various cytogenetic, molecular, and protein studies have provided evidence for the existence of several types of variant ALK fusions in up to 20% of ALK+ ALCL, of which only one, a TPM3-ALK fusion resulting from a t(1;2)(q25;p23), has so far been cloned. A cryptic inv(2)(p23q35) has been described as another recurrent cytogenetic alteration involving ALK and an unidentified fusion partner in some ALCL. In a screen for variant ALK gene fusions, we identified two ALCL that were negative for NPM-ALK by reverse transcriptase-polymerase chain reaction, but were positive for cytoplasmic ALK with both polyclonal and monoclonal antibodies to the ALK tyrosine kinase domain, consistent with ALK deregulation by an alteration other than the t(2;5) Case 1 was a T-lineage nodal and cutaneous ALCL in a 52-year-old woman, and Case 2 was a T-lineage nodal ALCL in a 12-year-old girl. FISH analysis confirmed ALK rearrangement in both cases. An inverse polymerase chain reaction approach was then used to identify the ALK translocation partner in Case 1. We found an in-frame fusion of ALK to ATIC, a gene previously mapped to 2q34-q35. We then confirmed by DNA polymerase chain reaction the localization of ATIC to yeast artificial chromosome (YAC) 914E7 previously reported to span the 2q35 break in the inv(2)(p23q35). FISH analysis in Case 1 confirmed rearrangement of YAC 914E7 and fusion to ALK. The ATIC-ALK fusion was confirmed in Case 1 and also identified in Case 2 by conventional reverse transcriptase-polymerase chain reaction using ATIC forward and ALK reverse primers. ATIC encodes an enzyme involved in purine biosynthesis which, like other fusion partners of ALK, is constitutively expressed and appears to contain a dimerization domain. ATIC-ALK fusion resulting from the inv(2)(p23q35) thus provides a third mechanism of ALK activation in ALK+ ALCL.

Anaplastic large cell lymphoma (ALCL) is recognized as a distinct subtype of non-Hodgkin’s lymphoma (NHL) in recent lymphoma classifications. This disease constitutes approximately 5% of all NHL but accounts for 30 to 40% of pediatric large cell lymphomas. 1 ALCL expresses Ki-1 (CD30), an antigen originally detected on Reed-Sternberg cells of Hodgkin’s disease, later shown to be a member of tumor necrosis factor receptor family. 2,3 In its classical or common form, ALCL shows an often bizarre, anaplastic morphology with sinusoidal infiltration of lymph nodes and a pseudocohesive appearance, and T or null phenotype. 4 The status of ALCL as a discrete entity had long been controversial, 5,6 and its recent separation into at least two subsets (see Discussion) stems from cytogenetic and molecular studies of the translocation seen in about 40 to 60% of cases, t(2;5)(p23;q35). 7-9

In 1994, the t(2;5) was found to involve a novel gene at 2p23 encoding a tyrosine kinase, ALK (anaplastic lymphoma kinase), and the NPM (nucleophosmin) gene at 5q35, which encodes a nucleolar phosphoprotein. 10 The resulting fusion gene encodes a chimeric protein, NPM-ALK, with a molecular weight of 80 kd, consisting of the N-terminal portion of NPM fused to the catalytic domain of ALK. 10 ALK is a tyrosine kinase receptor belonging to the insulin growth factor receptor superfamily, highly related to the leukocyte tyrosine kinase (LTK) gene but normally expressed only in the nervous system. 11,12 The fusion with NPM contributes the NPM promoter and the NPM oligomerization domain to NPM-ALK, and removes the ALK extracellular and transmembrane domains. As a result, the ALK kinase domain within NPM-ALK is constitutively activated through autophosphorylation, and its expression is deregulated and ectopic, both in terms of cell type and cellular compartment (cytoplasm; reviewed in Ref. 13 ). Downstream targets of the ALK kinase domain that may be relevant in mediating the oncogenicity of NPM-ALK are being identified. 14

Because of the highly restricted expression of native ALK in the nervous system and its absence in normal lymphoid tissues, immunohistochemical detection of aberrantly expressed ALK protein using monoclonal 15,16 or polyclonal 17,18 antibodies to the ALK kinase domain (retained in NPM-ALK) was found to be a sensitive and specific method for detecting NPM-ALK-positive ALCL. Interestingly, ALK immunostaining was observed in both cytoplasm and nucleus in most cases, but only in cytoplasm in some cases. The nuclear localization of NPM-ALK is due to the formation of heteromeric complexes with native NPM, which contains a nuclear localization signal. 14 Initially, the occasional variability in subcellular localization of ALK immunostaining was thought to reflect unknown factors affecting either the heteromerization of NPM-ALK with NPM, or the entry of the resulting heteromeric complexes into the nucleus. However, it soon became apparent that ALCL with exclusively cytoplasmic ALK immunoreactivity usually lacked NPM-ALK by reverse transcriptase-polymerase chain reaction (RT-PCR). 15,16,19 At the same time, using an artificial TPR-ALK construct, it was shown that only cytoplasmic localization is required for transformation by the ALK portion of NPM-ALK. 20 Taken together, these results suggested that in some ALCL, ALK may become oncogenically activated through fusion with other translocation partners unassociated with nuclear transport.

Studies of large series of Ki-1 ALCL by ALK immunostaining now indicate that up to 20% of cases show cytoplasmic staining only. 16,21,22 Furthermore, Western blot analysis has identified at least four different types of aberrant ALK-positive proteins in different cases of this subset of Ki-1 ALCL that show ALK immunostaining confined to the cytoplasm. 23 Thus, up to 20% of ALK+ Ki-1 ALCL may contain variant ALK translocations, and these may be of at least four types. By cytogenetic analysis, several variant translocations involving 2p23 have been reported in Ki-1 ALCL. These include t(2;13)(p23;q34), 24 t(1;2)(q25;p23), 25 a cryptic inv(2)(p23q35), 26 , t(1;2)(q21;p23), and t(2;3)p23;q21). 27 Of these, only the t(1;2)(q25;p23) has so far been cloned. Using a PCR-based genomic walking technique, Lamant et al 28 demonstrated that the gene involved at 1q25 is TPM3, which encodes nonmuscular tropomyosin and was previously known to similarly rearrange with and activate the NTRK1 receptor tyrosine kinase in some papillary thyroid carcinomas (reviewed in Ref. 29 ).

In the present report, we describe the cloning of a novel variant ALK gene fusion, ATIC-ALK, which is associated with the previously reported recurrent cryptic inversion, inv(2)(p23q35).

Materials and Methods

Patient Samples

Among 26 cases of ALCL diagnosed at Memorial Sloan-Kettering Cancer Center that had material available for molecular studies, we identified 13 cases (50%) negative for NPM-ALK by reverse transcriptase-polymerase chain reaction (RT-PCR), performed as reported previously, 9 using the primers NPM-5′ and ALK-3′ listed in Table 1 ▶ . Molecular data on 10 of 13 NPM-ALK+ and 8 of 13 NPM-ALK− cases have been described in part in previous studies. 9,30,31 All cases were immunohistochemically positive for Ki-1 antigen (CD30) using monoclonal antibody Ber-H2 (Dako, Santa Barbara, CA). Lineage phenotype and genotype were determined according to standard immunophenotypic and molecular genetic methods, as described in more detail elsewhere. 4 Case histories of the two patients that were studied in more detail (see Results) are summarized below.

Table 1.

Primers

Primer	Sequence
NPM-5′	TCCCTTGGGGGCTTTGAAATAACACC
ALK-3′	CGAGGTGCGGAGCTTGCTCAGC
ALKREV	GCAGTAGTTGGGGTTGTAGTC
ALKREV3	TGCAGCTCCTGGTGCTTCC
ALKREV4	TGGTGCTTCCGGCGGTAC
ALKFWD4	ACAAGCTGAGCAAGCTCCGC
ALKFWD5	ACCTCGACCATCATGACC
ATIC-FWD	CACGCTCGAGTGACAGTG
ATIC-REV	TCGGAGCTCTGCATCTCCG
PGK-FWD	CAGTTTGGAGCTCCTGGAAG
PGK-REV	TGCAAATCCAGGGTGCAGTG

All sequences are 5′ to 3′.

Case Summaries

Case 1

This 52-year-old woman was diagnosed overseas with malignant lymphoma in a left axillary mass and was treated with four cycles of chemotherapy without response. She came to MSKCC 4 months later for a second opinion. Clinical restaging showed left axillary adenopathy, focal infiltration of fat and adjacent muscle, and retraction and thickening of overlying skin. Bone marrow biopsy did not show infiltration by lymphoma. Biopsies of axillary and skin tumors showed large, polymorphic cells, with amphophilic cytoplasm, round and lobulated nucleus with two or three nucleoli. Frequent mitotic figures, necrosis, and phagocytosis were also seen (Figure 1) ▶ . The tumor showed the following staining features: CD30+, epithelial membrane antigen (EMA) +, CD43+, CD3−, CD45RO−, CD20−. No clonal rearrangement involving the immunoglobulin heavy chain gene (IGH) was detected by Southern blotting, but the TCRβ gene did show clonal rearrangement. This pattern was consistent with a Ki-1-positive T cell ALCL. Cytogenetic analysis of this biopsy showed the following clonal karyotype: 46, XX, del(1)(p32),der(3)dic(1;3)(q13;q27), der(6)t(1;?;6)(q25;?;p25), hsr(7)(p22), I(8)(q10), der(12)t(9;12)(q13;p13), add (13)(p11), der(15;15)(q10;q10), add(16)(q22), add(19) (q13), add(21)(p11), +mar. The patient returned overseas and was lost to follow-up.

Figure 1.

Representative histological appearance and ALK immunohistochemistry in Cases 1 and 2. Both cases showed strong cytoplasmic, but not nuclear, immunoreactivity with the ALK-11 polyclonal antibody and the ALK-1 monoclonal antibody, both generated to the same peptide from the C terminal portion of ALK retained in ALK fusion proteins.

Case 2

At the age of 12, this girl was first diagnosed with a nodal diffuse large cell lymphoma. The clinical stage was IIIA (negative bone marrow, no extranodal sites). She was treated with chemotherapy and external radiation, and achieved complete remission. Three years later, she developed a relapse and was treated with a similar chemoradiotherapy combination, and achieved a prolonged second complete remission. Twelve years later, she developed a second nodal relapse and started a new chemotherapy protocol. She died 3 months later due to sepsis and granulocytopenia. Biopsy of the second nodal recurrence showed rounded, monomorphic tumor cells with round nuclei and one or two nucleoli (Figure 1) ▶ . Numerous mitotic figures were seen. The tumor showed the following immunostaining: CD30+, EMA+, CD45+, CD43+, CD20−, CD15−. No clonal rearrangement involving IGH was detected by Southern blot analysis, but the TCRβ gene was clonally rearranged. This pattern was consistent with a Ki-1-positive T cell ALCL. Cytogenetic analysis showed the following clonal abnormalities: 47, XX, +2, del(6)(q21), t(8;13;20)(p11.2;p11.2;p11.2).

ALK Immunostaining

ALCL were subjected to immunostaining with a polyclonal antibody generated to amino acid residues 419–520 of NPM ALK, designated ALK−11 (gift of S. W. Morris), 18 after heat-induced epitope retrieval in citrate buffer for 10 minutes. Equivalent results were obtained at dilutions of 1:1000 and 1:2000. This antibody produced strong nuclear and cytoplasmic staining in all 5 ALCL tested that were positive for NPM-ALK+ by RT-PCR (results not shown). Cases positive with ALK-11 were further tested with the ALK-1 monoclonal antibody, generated to the same amino acid residues of NPM-ALK as the ALK-11 antibody (Dako), 15 at a dilution of 1:50, after heat-induced epitope retrieval in citrate buffer for 20 minutes. Immunoperoxidase staining was performed on paraffin sections, using a standard avidin-biotin peroxidase procedure.

Fluorescence in Situ Hybridization (FISH) with ALK and 2q35 Probes

Bicolor FISH studies were performed on cytologic touch preparations of Case 1 and on extracted nuclei from paraffin-embedded tissue blocks from Case 2 and the two ALK-11+ but ALK-1− cases (see Results) using the Vysis LSI ALK probe assay (Vysis, Inc., Downers Grove, IL) according to the manufacturer’s instructions. In addition, FISH studies with a 2p23 (ALK) breakpoint spanning probe (600 kb, courtesy of Dr. J. Proffitt, Vysis, Inc.) and yeast artificial chromosome (YAC) 914E7 (a 1750-kb CEPH YAC probe mapped to 2q35, Research Genetics, Huntsville, AL) were also performed on Case 1 and FISH studies with an ALK P1 clone (designated ALK-DMPC-HFF#1–1111H1, courtesy of Dr. S. W. Morris) and 914E7 were performed on Case 2. With respect to the latter hybridizations, probe mixtures containing 200 ng biotin-labeled YAC 914E7 (Nick Translation Kit, Vysis) and Spectrum Orange-labeled 2p23 (ALK) breakpoint spanning probe or digoxygenin-labeled ALKP1 was applied to a slide and sealed under a coverslip. The cells and probes were codenatured at 85°C for 5 minutes and incubated overnight at 37°C in a humidity chamber. Detection of signals was performed as described in detail elsewhere. 32 As negative controls, metaphase cells obtained from a cytogenetically normal lymph node and cytologic touch preparations of normal skeletal muscle were simultaneously hybridized with these probes.

Images were prepared using the Cytovision Image Analysis System (Applied Imaging, Santa Clara, CA). One hundred interphase nuclei with strong and well-delineated signals were examined by two different individuals. A separation of the Spectrum Orange- and Spectrum Green-labeled 2p23 breakpoint flanking probes (3′ telomeric and 5′ centromeric respectively to the 2p23 breakpoint of the t(2;5)(p23;q35) associated with non-Hodgkin’s lymphoma; Vysis LSI ALK probe assay) was interpreted as a rearrangement of the ALK gene.

Inverse PCR

For both inverse PCR and conventional RT-PCR, total RNA was extracted using Trizol method (Tel-Test, Friendswood, TX), and the adequacy of the extracted RNA was confirmed by amplification of a 247-bp fragment of the ubiquitous phosphoglycerate kinase (PGK) transcript, using primers PGK-FWD and PGK-REV. For inverse PCR, double-stranded cDNA was synthesized as follows using a cDNA Synthesis Kit (Boehringer Mannheim, Indianapolis, IN). Reverse transcription was performed on 1 μg of RNA, and primed with 2 pMol of ALKREV primer using AMV reverse transcriptase. The ALKREV primer binds 98 bp from the ALK fusion point in NPM-ALK and TPM3-ALK. Second-strand cDNA synthesis was performed using Escherichia coli DNA polymerase I and RNase H. The resulting double-stranded cDNA was then blunt-ended with T4 DNA polymerase and subsequently purified using the QIAquick PCR Purification Kit (Qiagen, Valencia, CA). The cDNA was then circularized by overnight incubation at room temperature in the presence of 1 U/μl T4 DNA ligase (Boehringer Mannheim) in a final volume of 30 μl. The ligation reaction was stopped by 65°C incubation for 10 minutes. The circularized cDNA was then relinearized by digestion with PstI, which cuts the ALK cDNA between the ALKREV3 and ALKFWD4 primer binding sites (oriented 5′ end to 5′ end). After a manual hot start, the cDNA was then amplified by PCR (35 cycles: 95°C for 45 seconds, 58°C for 1 minute, 72°C for 3 minutes, final extension 72°C for 10 minutes) with primers ALKREV3 and ALKFWD4 (15 pMol of each) using 2U/μl rTth DNA Polymerase (Perkin Elmer, Branchburg, NJ). Nested PCR was performed on 1 μl of the first PCR product using primers ALKREV4 and ALKFWD5, Taq polymerase (DNA PCR Beads, Pharmacia Biotech, Piscataway, NJ), for 35 cycles (95°C for 1 minute, 62°C for 45 seconds, 72°C for 1 minute, final extension 72°C for 10 minutes).

RT-PCR for ATIC-ALK

RT-PCR was performed using ATIC-FWD and ALKREV primers. First, reverse transcription was performed for 30 minutes at 42°C on 1 μg of RNA using 10× buffer II, 25 mmol/L MgCl₂, 50 mmol/L random hexamers, 10 mmol/L dNTP, 40U/μl RNase inhibitor, 200U/μl reverse transcriptase, and DEPC-treated H₂O for a final volume of 20 μl. A negative control (10 μl of DEPC-water) was included at this stage. The reverse transcriptase was inactivated at 99°C for 5 minutes. Eighty μl of master mix (water, 10× buffer II, MgCl₂ for a final concentration of 1.5 mmol/L, 15 pMol of each primer, 5U/μl of AmpliTaq DNA Polymerase, Perkin Elmer, Norwalk, CT) were added to the tube. PCR consisted of 35 cycles of 95°C for 1 minute, 60°C for 1 minute, 72°C for 1 minute, final extension of 72°C for 10 minutes. The PCR products were electrophoresed in 2% NuSieve agarose gel (FMC Bioproducts, Rockland, ME) and visualized by ethidium bromide staining.

YAC DNA Extraction

YACs 914E7 and 777D5 were obtained from Research Genetics. One colony was inoculated in 5 ml YPD medium (yeast extract, peptone, dextrose; Bio 101, Vista, CA) and incubated in an orbital shaker for 24 hours at 30°C. Eight hundred μl of this product were combined with 200 μl of glycerol and stored at −70°C. The remaining product was inoculated in 100 ml YPD medium and incubated in an orbital shaker for more 24 hours at 30°C. The YAC was centrifuged for 10 minutes at 3000 rpm and the pellet was resuspended in a lysis buffer (20% sodium dodecyl sulfate, 10% Triton X-100, 5 mol/L NaCl, 1 mol/L Tris, pH 8.0, 0.5 mol/L EDTA, pH 8.0). Seventy-five microliters of glass beads and 200 μl of 1:1 phenol:chloroform were added to the lysate and it was mixed in a vortex for 5 minutes. Two hundred microliters of TE buffer (10 mmol/L Tris, 1 mmol/L EDTA) was added to the lysate and it was mixed again. After 5 minutes centrifugation at room temperature, the clear supernatant was transferred to a new tube. Then, 750 μl of 100% isopropanol was added to it, mixed gently by inversion, and left for 5 minutes at room temperature. After centrifugation, a pink pellet was seen. The dried pellet was then resuspended in 300 μl TE buffer (pH 7.4). Fifteen microliters of 1 mg/ml RNase A was added and the product was incubated for 30 minutes at 37°C. The pellet was again precipitated with 100% isopropanol and 3 Mol/L NaAc. After centrifugation, it was washed in 70% ethanol and dissolved in TE buffer. The DNA was electrophoresed in a 1% agarose gel to evaluate its quality.

Results

Identification of Two Cases with Evidence of Variant ALK Rearrangements

The study group consisted of 13 ALCL of non-B cell lineage that lacked NPM-ALK by RT-PCR (see Materials and Methods). There were 7 female and 6 male patients, with median age of 47.3 years (range, 13–70 years). These 13 cases were subjected to immunostaining with polyclonal ALK-11 antibody to the ALK kinase domain. Four T cell ALCL cases were positive. These four cases were further tested by immunostaining with the ALK-1 monoclonal antibody, and by interphase FISH analysis for ALK rearrangement. Two cases, Cases 1 and 2, were also positive with ALK-1 (Figure 1) ▶ . Case 1 also showed ALK rearrangement by FISH using 2p23 (ALK) breakpoint flanking probes. Specifically, a separation of these breakpoint flanking probes was detected in 97% of the interphase nuclei analyzed in Case 1 (Figure 2A) ▶ using the two-color Vysis ALK probe FISH assay, indicative of an ALK rearrangement. Moreover, a third copy of the Spectrum Orange signal of this probe set, which is located telomeric to the 2p23 breakpoint, was observed in all abnormal cells of Case 1 (Figure 2A) ▶ . FISH studies with the two-color Vysis ALK probe FISH assay were unsuccessful in Case 2, where only extracted nuclei from paraffin-embedded tissue blocks were available. Brief case histories for these two patients are presented above in Materials and Methods. The remaining two cases that were negative by ALK-1 immunohistochemistry were also negative by ALK FISH.

Figure 2.

A: FISH studies performed on Case 1 with the LSI ALK probe assay revealed the presence of a normal chromosome 2 homologue (arrow), a rearrangement of ALK (separate orange and green signals) and an extra copy of the probe telomeric to the 2p23 breakpoint (extra orange signal). B: FISH studies performed on Case 1 with the 2p23 (ALK) breakpoint spanning probe and 914E7 (2q35 breakpoint spanning probe) revealed the presence of a normal chromosome 2 homologue (separate orange and green signals) and rearrangement with fusion of 2p23 and 2q35 breakpoints (juxtaposed orange and green signals). In conjunction with the findings illustrated in A, these results are compatible with the presence of an inv(2)(p23q35) with an extra copy of the ATIC-ALK fusion gene.

Cloning of ATIC-ALK Fusion by Inverse PCR in Case 1

As schematized in Figure 3 ▶ and described in more detail in Materials and Methods, we performed inverse PCR with nested amplification to isolate the ALK translocation partner in this case. There were two inverse PCR product bands: a broad 200- to 300-bp band (A), and a fainter band of approximately 120 bp (B; Figure 4 ▶ ). Both products were analyzed by direct automated sequencing. Sequence analysis of the 120-bp B band showed an in-frame fusion between ATIC and ALK, occurring at codon 162 of the former and codon 1058 of ALK, the same codon involved in the NPM-ALK fusion. The broad 200- to 300-bp A band was a nonspecific PCR product.

Figure 3.

Schema of inverse PCR methodology. Technical details are provided in Materials and Methods. The primers used were as follows: P1, ALKREV; P2, ALKFWD4; P3, ALKREV3; P4, ALKFWD5; P5, ALKREV4. Dimensions are not to scale, and primer positions are relative.

Figure 4.

Results of inverse PCR, ATIC-ALK RT-PCR, and sequencing of RT-PCR products in Case 1. Inverse PCR performed in Case 1, as illustrated in Figure 2 ▶ and described in Materials and Methods, yielded two predominant bands. Band A was a nonspecific amplification product. Sequence analysis of band B demonstrated an in-frame ATIC-ALK fusion transcript (see Results for details). Based on this fusion transcript, the ATIC-FWD primer was designed and used in combination with the ALKREV primer in an ATIC-ALK RT-PCR assay. This yielded a 370-bp band in Case 1, but not in liver RNA. Sequencing of this band revealed an in-frame ATIC-ALK fusion transcript. The portion of the electrophoretogram containing the fusion point is illustrated. M, φX174/HaeIII size marker bands (in bp).

Confirmation of ATIC-ALK Fusion by RT-PCR in Cases 1 and 2

Based on the ATIC-ALK chimeric transcript identified by inverse PCR, we designed primer ATIC-FWD to produce a 169-bp RT-PCR product in conjunction with the ALKREV primer. RT-PCR with these primers yielded only a single strong 370-bp band in both cases (Figures 3 and 4) ▶ ▶ , instead of the expected 169-bp product. Sequence analysis of this 370-bp band also showed an in-frame fusion between ATIC and ALK, occurring again at codon 1058 of ALK, but at a different point in ATIC, codon 229 instead of 162 (Figure 3) ▶ . In light of this result, we suspect that this major fusion transcript may have been either obscured in the inverse PCR by the nonspecific 200- to 300-bp product (see above) or that the shorter fusion transcript may have been more efficiently isolated for technical reasons. This shorter fusion transcript, which was detected only in Case 1 by the nested amplification of the inverse PCR procedure, likely arose by alternative splicing of the major fusion product. The intervening portion of ATIC may therefore correspond to one or more exons (the exon structure of ATIC is largely unknown 33 ). This shorter minor splice form is unlikely to be biologically significant because of its low expression level and because it lacks the ATIC dimerization domain (see Discussion). As an incidental observation, our sequencing data confirmed that ATIC codon 164 reads GAC (for aspartic acid), as in reference 34, instead of GGC (for glycine) reported in reference 35. Furthermore, a search of the expressed sequence tag database (dbEST) identified five perfect matches for GAC (D83877, H09885, H06543, R20197, AA112359) and none for GGC at this codon.

To assess Case 2 for the presence of the ATIC-ALK fusion, we performed RT-PCR using the same primers as above, namely ATIC-FWD and ALKREV. This yielded the same 370-bp RT-PCR product (Figure 5) ▶ , confirmed by sequencing to be the ATIC-ALK fusion transcript.

Figure 5.

Detection of ATIC-ALK fusion transcript in Case 2. RT-PCR, using the same primers as in Case 1 (see Figure 4 ▶ ), demonstrated the same product in total RNA from Case 2. Controls lacking either reverse transcriptase or RNA were appropriately negative. M, 100-bp ladder size marker bands (in bp).

Mapping of ATIC to a YAC Spanning the 2q35 Breakpoint in the inv(2)(p23q35)

ATIC was previously mapped to 2q34-q35 by FISH. 33 YAC 914E7 at 2q35 was reported by Wlodarska et al to be rearranged by the cryptic inv(2)(p23q35). 26 To confirm that this YAC contains the ATIC gene, we performed DNA PCR on purified YAC DNA using primers ATIC-FWD and ATIC-REV. The expected 71-bp product was amplified from YAC 914E7 DNA, but not from an unrelated YAC (777D5 from 1q21; result not shown), confirming that ATIC maps to YAC 914E7.

Confirmation of ATIC-ALK Fusion by FISH in Cases 1 and 2

FISH studies performed on Case 1 with the Spectrum Orange-labeled 2p23 (ALK) breakpoint spanning probe and the biotin-labeled YAC 914E7 revealed a distinct or separate orange and green signal consistent with the presence of a normal chromosome 2 homologue and three orange and green signals lying directly adjacent or juxtaposed to each other indicative of 2p23 and 2q35 rearrangements in 96% of the interphase nuclei analyzed (Figure 2B) ▶ . In Case 2, FISH analysis of nuclei from paraffin-embedded tissue blocks with ALK P1 and YAC 914E7 probes revealed a distinct or separate red (rhodamine) and green (fluorescein isothiocyanate) signal consistent with the presence of a normal chromosome 2 homologue and two red and green signals lying directly adjacent or juxtaposed to each other, indicative of two copies of ATIC-ALK fusion in 88% of the interphase nuclei analyzed (not illustrated). The limited tissue quantity available permitted analysis of 50 interphase nuclei in this case. Thus, the findings in both cases were compatible with the presence of an inv(2)(p23q35) and an additional copy of ATIC-ALK.

Discussion

ATIC (also known as hPurH) encodes 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase/IMP cyclohydrolase (AICARFT/IMPCH), a bifunctional enzyme that catalyzes the penultimate and the final steps of the purine nucleotide synthesis pathway, AICARFT and IMPCH. As expected for an enzyme required for DNA synthesis, ATIC is ubiquitously expressed, 34 and this should provide a strong active promoter to the ATIC-ALK fusion gene. The promoters of the two other fusion partners of ALK, NPM and TPM3, are both constitutively active in lymphoid cells. 28 Although ATIC is known to be highly expressed in the CCRF-CEM leukemia cell line, 34 the identification of ATIC-ALK in ALCL may warrant a more detailed analysis of ATIC expression levels in lymphoid lineages.

Studies of ATIC deletion mutants have confirmed the existence of two non-overlapping functional domains, separated by a linker region. 33 Based on these deletion studies and on crystallography data, a working model of ATIC has been proposed in which residues 1 to 169 encode the IMPCH function, residues 170 to 199 encode the linker region, and residues 200 to 592 encode the AICARFT activity. 33,35 Furthermore, crystallography and equilibrium sedimentation studies indicate that ATIC exists mainly as a homodimer. 33 Gel filtration and ultracentrifugation studies of additional ATIC deletion mutants suggest that the linker region contains a dimerization domain (G. P. Beardsley, personal communication). The first 229 amino acid residues of the predicted ATIC-ALK protein are identical to those of ATIC (the valine codon at the fusion point is conserved). Thus, in addition to an active promoter, ATIC appears to contribute a dimerization domain to ATIC-ALK, which should lead to constitutive autophosphorylation and activation of the ALK kinase domain. These properties are shared by NPM (see Introduction) and TPM3. In ALCL with the t(1;2)(q25;p23), TPM3 contributes to TPM3-ALK an active promoter, and activation of the ALK catalytic domain probably results from homodimerization through the TPM3 protein-protein interaction domain. 28 Like ATIC-ALK, TPM3-ALK accumulates only within the cytoplasm. 20,28

As discussed in the Introduction, various lines of evidence suggest that up to 20% of ALK+ ALCL contain variant ALK translocations. Furthermore, these may be of at least four types, according to the Western blot studies of Pulford et al. 23 They identified four different types of putative variant ALK fusion proteins with molecular weights of 85, 97, 104, and 113 kd (compared to 80 kd for NPM-ALK and 200 kd for native ALK), in three of which tyrosine kinase activity was also confirmed. 23 Whether any of these correspond to ATIC-ALK or TPM3-ALK is presently unclear. None of their cases had a documented inv(2)(p23q35), whereas one case with a aberrant ALK protein of 104 kd had a t(1;2), but with a 1q21 break instead of the 1q25 break associated with TPM3-ALK. 28 NPM, ATIC, and TPM3 contribute 116, 229, and 255 amino acid residues to their respective ALK fusion proteins. Based on an average of 120 d per amino acid residue, the predicted molecular weight of the ATIC-ALK protein is about 94 kd, and that of TPM3-ALK is 97 kd. Aside from the approximate nature of these estimates, posttranslational modifications may alter protein mobility, making it difficult to directly assign bands on Western blots to specific predicted fusion proteins.

The inv(2)(p23q35) was first reported in 1997 19 and was then described in detail in three cases by Wlodarska et al. 26 This inversion is cryptic, ie, it is not apparent on conventional Giemsa-banded cytogenetic preparations, because terminal bands of similar size and staining pattern are exchanged. This may explain why no 2p23 or 2q35 breaks were apparent in the conventional karyotypes of cases 1 and 2 (see Case Summaries). Furthermore, Wlodarska et al 26 found a consistent association of the inv(2)(p23q35) with a secondary chromosomal aberration, ider(2)(q10)inv(2)(p23q35), which results in extra copies of the rearranged ALK gene. FISH analysis demonstrated at least one additional copy of the fusion gene in both of our cases. Consistent amplification of ATIC-ALK suggests that it may be less oncogenic than NPM-ALK, and therefore requires extra copies to exert an equivalent cellular effect. The phenomenon may be analogous to the consistent amplification of the PAX7-FKHR variant fusion gene in alveolar rhabdomyosarcoma. 36,37

ATIC-ALK may make up a significant proportion of variant ALK fusions. In the present series, ATIC-ALK accounted for 2 of 15 (13%) cases positive for ALK by immunostaining or for NPM-ALK by RT-PCR. Considering the cytogenetic reports of the inv(2)(p23q35), three cases of ATIC-ALK have been previously reported. 26 Aside from TPM3-ALK, reported in three cases, 28 no other known ALK variant translocations have been recurrent.

As the third reported translocation partner of ALK, the finding of ATIC highlights the promiscuous nature of many genes involved in oncogenic translocations. 38 NPM also rearranges with other genes, resulting in the NPM-RARα fusion in rare cases of acute promyelocytic leukemia and the NPM-MLF1 fusion seen in some cases of myelodysplastic syndrome and acute myeloid leukemia. 39,40 In ALCL, no variant fusions involving NPM but not ALK have so far been identified, although there are cytogenetic case reports of a t(3;5)(q12;q35) 7 and a t(1;5)(q32;q35). 41 In the absence of molecular studies, however, the latter two cases may simply contain a cryptic complex t(2;5). Interestingly, a recent report has also implicated ALK activation by rearrangement as a recurrent alteration in a mesenchymal tumor, the inflammatory myofibroblastic tumor. 42

The potential clinical significance of these variant ALK fusions is that ALK+ ALCL, defined either by immunohistochemistry, or by molecular or cytogenetic detection of NPM-ALK, is a prognostically favorable subset of ALCL. Before the development and application of ALK immunodetection, this important observation was statistically hampered by the limited numbers of cases with cytogenetic or molecular data. 1,43,44 Recent retrospective analyses of large series of ALCL by ALK immunostaining have established that ALK+ ALCL occurs in significantly younger patients, is more often extranodal, and has a markedly better clinical outcome. 22,45,46 Furthermore, a recent multivariate analysis indicates that the survival advantage of patients with ALK+ ALCL is not merely secondary to their younger age. 46 Although this accounts for the bimodal age distribution of Ki-1 ALCL, the age ranges for ALK+ and ALK− cases still overlap considerably.

Pathologically, ALK+ ALCL are of non-B cell lineage and almost always coexpress CD30 and EMA. 21,46 However, ALK+ ALCL may be morphologically indistinguishable from ALK− cases. ALK expression crosses all morphological forms of ALCL, including appearances which are neither anaplastic nor large cell. 21,22,47 Hence, the simpler term ALK+ NHL has been proposed by some. 21,22 In these clinicopathological studies of ALK+ ALCL, cases with variant ALK fusions have been lumped with the more common NPM-ALK cases. Although it is reasonable to expect that their clinical behavior may be closer to that of NPM-ALK+ ALCL than ALCL lacking any ALK alterations, it is only with the cloning of these variant ALK fusions that a systematic clinical comparison becomes possible.

Finally, it is tempting to speculate about possible therapeutic implications of the ATIC-ALK fusion. The AICARFT reaction mediated by ATIC is a folate-dependent reaction, and as such is thought to account in part for the anti-purine effects of antifolates such as methotrexate whose primary target is dihydrofolate reductase. 34 This raises the intriguing possibility that, as the ATIC-ALK fusion represents a reciprocal rearrangement (leaving only one normal ATIC allele/cell), Ki-1 ALCL bearing this genetic alteration could be more sensitive to antifolates because of lower cellular ATIC activity, due to haploinsufficiency and/or potential dominant negative effects of heterodimerization of ATIC-ALK with residual native ATIC.

Note added in proof: Since our study was completed, Ma et al 48 have reported in abstract form the independent cloning of the identical ATIC-ALK fusion and its detection in five cases with the inv(2)(p23q35).

Figure 6.

Schematic diagram of predicted ATIC-ALK fusion protein. The fusion protein retains the tyrosine kinase (TK) domain of ALK as well as the IMPCH and the linker-dimerization domains of ATIC. The rearrangement interrupts the AICARFT domain of ATIC and excludes the extracellular and transmembrane (TM) domains of ALK. The proportions are roughly to scale, but domain boundaries within ATIC are still somewhat approximate.