Immunohistochemical Approach to Genetic Subtyping of Anaplastic Large Cell Lymphoma

Abstract

Anaplastic large cell lymphoma (ALCL) can be classified genetically based on rearrangements (R) of the ALK, TP63, and/or DUSP22 genes. ALK-R defines a specific entity, ALK-positive ALCL, while DUSP22-R and TP63-R define subgroups of ALK-negative ALCLs with distinct clinicopathologic features. ALK-R and TP63-R produce oncogenic fusion proteins that can be detected by immunohistochemistry. ALK immunohistochemistry is an excellent surrogate for ALK-R and screening with p63 IHC excludes TP63-R in two-thirds of ALCLs. In contrast, DUSP22-R does not produce a fusion protein and its identification requires fluorescence in situ hybridization (FISH). However, DUSP22-R ALCL has a characteristic phenotype including negativity for cytotoxic markers and phospho-STAT3^Y705. Recently, we also identified overexpression of the LEF1 transcription factor in DUSP22-R ALCL. Here, we sought to validate this finding and examine models for predicting DUSP22-R using immunohistochemistry for LEF1 and TIA1 or phospho-STAT3^Y705. We evaluated these 3 markers in our original discovery cohort (n=45) and in an independent validation cohort (n=46) of ALCLs. The correlation between DUSP22-R and LEF1 expression replicated strongly in the validation cohort (P<0.0001). In addition, we identified and validated a strategy using LEF1 and TIA1 immunohistochemistry that predicted DUSP22-R with positive and negative predictive values of 100% after exclusion of indeterminate cases, and would eliminate the need for FISH in 65% of ALK-negative ALCLs. This approach had similar results in identifying DUSP22-R in the related condition, lymphomatoid papulosis. Together with previous data, these findings support a 4-marker immunohistochemistry algorithm using ALK, LEF1, TIA1, and p63 for genetic subtyping of ALCL.

Introduction

Anaplastic large cell lymphomas (ALCLs) comprise a group of CD30-positive T-cell lymphomas with varying clinical and genetic features (1, 2). Clinically, ALCLs present as systemic, primary cutaneous (PC), or breast implant-associated (BIA) disease. Genetically, ALCLs can be subtyped based on rearrangements (R) of the ALK, DUSP22, and/or TP63 genes (3, 4). ALK-R is typically seen in systemic disease and defines a well-characterized entity, ALK-positive ALCL, which lacks DUSP22-R and TP63-R (3, 5-7). ALK-positive ALCL occurs most commonly in children and young adults and has a generally favorable prognosis, though high-risk cases occur (8). DUSP22-R occurs in 20-30% of both systemic and PC ALK-negative ALCLs, typically affects adults, and also has a generally favorable prognosis with a subset of high-risk cases (3, 4, 7, 9-13). TP63-R occurs in 2-8% of ALK-negative ALCLs and is associated with aggressive disease (3, 6, 9). Rare ALK-negative ALCLs with dual DUSP22-R/TP63-R have been reported (6, 14). The remaining ALCLs lack ALK-R, DUSP22-R, and TP63-R (triple-negative [TN]); systemic TN-ALCLs have an intermediate prognosis (3, 9).

As with other cancers, there is increasing emphasis on genetic subtyping of ALCL in clinical practice (4, 15). Evaluation for ALK-R is required to classify ALCL into ALK-positive and ALK-negative entities and in some clinical settings stratifies patients for treatment with ALK tyrosine kinase inhibitors (16). There is evolving evidence that DUSP22-R also identifies a distinct genetic subtype of ALCL. In addition to its prognostic significance, DUSP22-R has been associated with unique morphologic, immunophenotypic, mutational, transcriptomic, and epigenomic features (3, 17-19). Molecular and biological features of TP63-R ALCL remain incompletely understood, but recognition of these cases is important due to their aggressive clinical course.

A variety of methods are currently used for genetic subtyping of ALCL. Since ALK-Rs encode ALK fusion proteins and wild-type ALK is not expressed in normal lymphocytes, immunohistochemistry (IHC) is typically used as a surrogate to identify the presence of ALK-R (5, 20). In equivocal cases fluorescence in situ hybridization (FISH) or next-generation sequencing (NGS) may be used. Similarly, TP63-Rs encode p63 fusion proteins that can be detected by IHC with high sensitivity; p63 IHC is useful as a screening test; confirmatory FISH is required in positive cases since 5-35% of ALCLs without TP63-R express wild-type p63 (6, 21). Unlike ALK-R and TP63-R, DUSP22-R does not encode a fusion protein and detection of DUSP22-R is currently performed clinically using a breakapart FISH assay (3, 7, 11).

Developing an IHC strategy to facilitate identification of DUSP22-R has been challenging. DUSP22-Rs disrupt the DUSP22 gene locus and are associated with diminished expression of the DUSP22 dual-specificity phosphatase, a putative tumor suppressor (7, 22); however, DUSP22 IHC has not been useful due to variability in DUSP22 expression levels and cross-reactivity with other DUSP family proteins (unpublished data). Nevertheless, DUSP22-R ALCL has a characteristic phenotype. IHC for cytotoxic markers (especially TIA1) and phosphorylated (p) STAT3^Y705 correlate strongly with genetic subtype in ALCL and are typically negative in DUSP22-R ALCLs (13, 18, 23). Most strikingly, we recently reported an association in ALCL between DUSP22-R and expression of the lymphoid enhancer-binding factor LEF1 (24), which is currently used in pathology practice as an IHC marker in the differential diagnosis of chronic lymphocytic leukemia (25). In this study, we sought to validate this finding in an independent ALCL cohort and examine models for predicting DUSP22-R by combining LEF1 IHC with TIA1 or pSTAT3 IHC.

Materials and methods

Patients

The discovery cohort comprised the 45 ALCL patients in which we previously reported the correlation between LEF1 expression and DUSP22-R (24). The validation cohort comprised 46 additional ALCL patients. The characteristics of both cohorts are shown in Table 1. Age and sex distributions were similar between the 2 cohorts. Overall mean age was 58 years (range, 13-97 years) and the M:F ratio was 1.7:1. Both cohorts included all 4 World Health Organization (WHO) ALCL entities (ALK-positive ALCL, ALK-negative ALCL, PC-ALCL, and BIA-ALCL) and all 4 genetic subtypes (ALK, DUSP22, TP63, and TN). The anatomic sites of the diagnostic biopsy for the combined cohort and their relationship to genetic subtype are shown in Supplemental Table 1. A separate, exploratory analysis was conducted on samples from 17 patients with lymphomatoid papulosis (LyP), a related CD30-positive T-cell lymphoproliferative disorder. LyP was diagnosed according to the criteria of the World Health Organization-European Organization for Research and Treatment of Cancer (WHO-EORTC) classification (26) based on both pathological features and the clinical presentation and course. The mean age was 60 years (range, 16-89 years) and the M:F ratio was 1.4:1. All patients had a typical regressing and relapsing clinical picture (Supplemental Table 2). The study was conducted under a research protocol approved by the Mayo Clinic Institutional Review Board.

Table 1.

Demographic Features and ALCL Subtypes in Discovery and Validation Cohorts

Feature	Discovery Cohort	Validation Cohort	Total
n	45	46	91
Age (years)
Mean	60	56	58
Range	13-97	23-89	13-97
Sex
Male	29	28	57
Female	16	18	34
WHO Subtype
ALK-positive ALCL	4	7	11
ALK-negative ALCL	29	24	53
PC-ALCL	10	4	14
BIA-ALCL	2	11	13
Genetic Subtype
ALK	4	7	11
DUSP22	16*	13	29
TP63	7	3	10
TN	18	23	41

^*Includes 1 case with dual DUSP22-R/TP63-R

ALCL, anaplastic large cell lymphoma; PC-ALCL, primary cutaneous ALCL; BIA-ALCL, breast implant-associated ALCL; TN, triple negative.

IHC and FISH

IHC for ALK and LEF1 was performed previously on cases in the discovery cohort (24) and an identical method was used for cases in the validation cohort. Briefly, formalin-fixed paraffin-embedded tissue sections were stained using antibodies against ALK (D5F3, Cell Signaling) and LEF1 (EPR2029Y, Abcam) on a Ventana Benchmark XT automated stainer. Cases in both the discovery cohort and validation cohort were stained using similar methods for TIA1 (TIA-1, Immunotech) (27) and pSTAT3^Y705 (D3A7, Cell Signaling; hereafter referred to as pSTAT3) (18). ALK was scored as positive or negative per ALCL diagnostic criteria (1). Other stains were scored blinded to genetic subtype as the percentage of positive tumor cells by decile as previously published (3, 18). The percentage of positive cells in LyP was based on scoring the tumor cells rather than scoring all infiltrating cells. Thresholds for positivity were based on previously established cutoffs and were: LEF1, >75% (24); TIA1, ≥20% (3); and pSTAT3, ≥20% (18, 28). To ensure phosphosite preservation in the interpretation of pSTAT3 expression, sections without positive endothelial cell staining within the tumor were considered uninterpretable (3).

Genetic subtyping was assessed previously in the discovery cohort (24) and was assessed in the validation cohort using identical methods. FISH was performed using breakapart probes for DUSP22 and TP63 in all ALK-negative cases (3, 21) and using a dual-fusion probe for TBL1XR1/TP63 in cases where TP63 breakapart FISH was equivocal (29). One case with dual DUSP22-R/TP63-R was present in the discovery cohort and was grouped with the DUSP22 genetic subtype for analysis.

IHC and FISH studies were performed similarly for the LyP cohort.

Statistics

Data were analyzed and visualized using GraphPad Prism 9.2.0. Two-group comparisons were analyzed using the Mann-Whitney test. Comparisons among multiple groups were analyzed using the Kruskal-Wallis test. Positive predictive values (PPV) and negative predictive values (NPV) were calculated using standard formulas after excluding cases with indeterminate testing results. Overall survival was assessed using the log-rank test.

Results

LEF1 expression stratified by WHO subtype and genetic subtype

We first examined the distribution of LEF1 expression based on WHO subtype of ALCL (Figure 1A-C). In the discovery cohort, there was a non-significant trend toward differential expression of LEF1 among WHO subtypes, with higher expression in ALK-negative ALCL and PC-ALCL. This differential expression was more pronounced in the validation cohort (P=0.008) and was also reflected in the combined cohort (discovery and validation cohorts together), with mean values of 37% and 41% positive cells for ALK-negative ALCL and PC-ALCL, respectively, and 11% and 3% for ALK-positive ALCL and BIA-ALCL, respectively (P=0.008). Notably, DUSP22-R does not occur in ALK-positive ALCL and BIA-ALCL (3, 7, 30). The differential expression of LEF1 across genetic subtypes of ALCL was more pronounced than across WHO subtypes (P<0.0001 in discovery and validation cohorts); mean values for LEF1 expression by genetic subtype in the combined cohort were: ALK, 11%; DUSP22, 80%; TP63, 0%; and TN, 6% (P<0.0001; Figure 1D-F). Therefore, LEF1 expression differs by WHO subtype of ALCL, and this difference is mostly attributable to differential expression among genetic subtypes, with the highest expression seen in DUSP22-R ALCL.

Figure 1. Expression of LEF1 in ALCL.

A-C. Distribution of LEF1 expression across WHO subtypes of ALCL in discovery, validation, and combined cohorts, respectively. D-F. Distribution of LEF1 expression across genetic subtypes of ALCL in discovery, validation, and combined cohorts, respectively. ALK+, ALK-positive ALCL; ALK−, ALK-negative ALCL; PC, primary cutaneous ALCL; BIA, breast implant-associated ALCL; TN, triple negative (negative for ALK, DUSP22, and TP63 rearrangements).

LEF1, TIA1, and pSTAT3 expression stratified by DUSP22-R

We then examined differences in LEF1, TIA1, and pSTAT3 expression based on the presence or absence DUSP22-R in the discovery, validation, and combined cohorts. Representative ALCLs with and without DUSP22-R are shown in Figure 2. As previously reported (24), LEF1 expression correlated strongly with the presence of DUSP22-R in the discovery cohort (mean ± standard deviation, 83±20% versus 10±21% in cases without DUSP22-R; P<0.0001); this result replicated in the validation cohort (78±25% versus 3±10%, P<0.0001) and was also reflected in the combined cohort (81±22% versus 6±16%, P<0.0001; Figure 3A-C). We observed relative lack of TIA1 expression in DUSP22-R ALCL compared to ALCL without DUSP22-R in all analyses (discovery cohort: 2±4% versus 31±30%, P=0.0004; validation cohort: 2±6% versus 26±29%, P=0.0004; combined cohort: 2±5% versus 29±30%, P<0.0001; Figure 3D-F). Similarly, we observed relative lack of pSTAT3 expression in DUSP22-R ALCL compared to ALCL without DUSP22-R (discovery cohort: 6±15% versus 48±38%, P=0.0001; validation cohort: 2±8% versus 59±35%, P<0.0001; combined cohort: 4±13% versus 54±37%, P<0.0001; Figure 3G-I). A total of 5/91 cases (5%) had indeterminate pSTAT3 due to poor phosphosite preservation and were excluded from these analyses. Taken together, these findings demonstrate and independently validate that DUSP22-R ALCL has a distinct phenotype typically expressing LEF1 and lacking expression of TIA1 and pSTAT3.

Figure 2. LEF1, TIA1, and pSTAT3 in ALCLs with and without DUSP22-R.

A. ALK-negative ALCL with DUSP22-R. The tumor cells are positive for CD30 and LEF1, and negative for TIA1 and pSTAT3. B. ALK-negative ALCL without DUSP22-R (triple negative). The tumor cells are positive for CD30, TIA1, and pSTAT3, and negative for LEF1. IHC stains are as indicated; original magnification, 40×. FISH images are from DUSP22 breakapart probe set with 5’ and 3’ probes labeled red and green, respectively; original magnification, 60×.

Figure 3. Expression of LEF1, TIA1, and pSTAT3 in ALCL based on DUSP22-R status.

A-C. Distribution of LEF1 expression in discovery, validation, and combined cohorts, respectively. D-F. Distribution of TIA1 expression in discovery, validation, and combined cohorts, respectively. G-I. Distribution of pSTAT3 expression in discovery, validation, and combined cohorts, respectively. Cases with indeterminate pSTAT3 due to poor phosphosite preservation are excluded (2/45 in discovery cohort, 3/46 in validation cohort, and 5/91 in combined cohort).

Algorithmic approach to the use of IHC in genetic subtyping of ALCL

Next, we examined the performance of LEF1 IHC alone in predicting DUSP22-R in ALCL. We excluded ALK-positive ALCL from this analysis since in pathology practice determination of DUSP22-R status is not performed in this subset due to mutual exclusivity. Results are shown in Table 2. As previously reported (24), the PPV for LEF1 in predicting DUSP22-R in the discovery cohort was 93.8% and the NPV was 96.0%. In the validation cohort, the PPV and NPV were 100.0% and 96.3%, respectively, and in the combined cohort they were 96.4% and 96.2%, respectively. We then examined the predictive power of combining LEF1 IHC with either TIA1 or pSTAT3 IHC. For these analyses, a positive result was defined as positive for LEF1 and negative for the other marker; a negative result was defined as negative for LEF1 and positive for the other marker; and an indeterminate result was defined as any other combination or, in the case of pSTAT3, an uninterpretable stain due to lack of phosphosite preservation. For the combination of LEF1 and TIA1 IHC, the PPV and NPV were each 100.0% in the discovery, validation, and combined cohorts, and the proportions of cases with indeterminate results were 29.3%, 41.0%, and 35.0%, respectively. For the combination of LEF1 and pSTAT3 IHC, the PPV and NPV were each 100.0% in the discovery, validation, and combined cohorts, and the proportions of cases with indeterminate results were 29.3%, 28.2%, and 28.8%, respectively.

Table 2.

Performance of Models to Predict DUSP22-R in ALCL

	PPV (%)	NPV (%)	% indeterminate
Discovery Cohort
LEF1 only	93.8	96.0	0.0
LEF1/TIA1	100.0	100.0	29.3
LEF1/pSTAT3	100.0	100.0	29.3
Validation Cohort
LEF1 only	100.0	96.3	0.0
LEF1/TIA1	100.0	100.0	41.0
LEF1/pSTAT3	100.0	100.0	28.2
Combined Cohort
LEF1 only	96.4	96.2	0.0
LEF1/TIA1	100.0	100.0	35.0
LEF1/pSTAT3	100.0	100.0	28.8

PPV, positive predictive value; NPV, negative predictive value.

Outcomes in the combined cohort

Outcome data were available for 26 patients with systemic ALCL in the combined cohort (median follow-up time, 52 months). Overall survival rates were most favorable for ALCL with DUSP22-R, followed by ALK-positive ALCL, TN-ALCL, and ALCL with TP63-R (P=0.002; Figure 4A). Among patients with systemic ALK-negative ALCL with available outcome data (n=21), the median overall survival time was not reached for cases with DUSP22-R versus 19 months for cases without DUSP22-R (P=0.02; Figure 4B).

Figure 4. Overall survival rates in systemic ALCL.

A. Overall survival stratified by genetic subtype among systemic ALCL patients with available outcome data (n=26; median follow-up time, 52 months). Median survival: ALK-positive ALCL (ALK), 88 months; ALCL with DUSP22-R (DUSP22), not reached; ALCL with TP63-R (TP63), 18 months; triple-negative ALCL (TN), 55 months. B. Overall survival stratified by presence or absence of DUSP22-R among systemic ALK-negative ALCL patients with available outcome data (n=21). Median survival: ALK-negative ALCL with DUSP22-R, not reached; ALK-negative ALCL without DUSP22-R, 19 months.

Prediction of DUSP22-R in lymphomatoid papulosis

Because DUSP22-R defines a WHO subtype of LyP (26, 31), we performed an exploratory analysis to determine whether IHC might facilitate identification of these cases. There were 17 cases of LyP, including 8 type A, 2 type C, 1 type D, and 6 with DUSP22-R (Supplemental Table 2). ALK IHC and TP63 FISH were negative in all cases. Cases classified as type A, C, or D lacked DUSP22-R. LyP with DUSP22-R showed a higher percentage of LEF1-positive cells than LyP without DUSP22-R expression (38±43% versus 4±7%, respectively; P=0.048), and lower percentages of TIA1-positive cells (0±0% versus 40±23%; P<0.0001) and pSTAT3-positive cells (10±15% versus 45±29%; P=0.017; Figure 5A-C). LEF1 IHC alone yielded a PPV and NPV for DUSP22-R of 100.0% and 73.3%, respectively. For the combination of LEF1 and TIA1 IHC, the PPV and NPV were each 100.0% with 35.3% of cases being indeterminate. For the combination of LEF1 and pSTAT3 IHC, the PPV and NPV were 100.0% and 90.9%, respectively, with 29.4% of cases being indeterminate. A representative case of LyP with DUSP22-R is shown in Figure 5D. As previously described (31), lesions showed a biphasic growth pattern consisting of an epidermotropic infiltrate of small, atypical lymphocytes and a dense infiltrate of larger lymphocytes in the dermis with minimal background inflammatory cells.

Figure 5. LEF1, TIA1, and pSTAT3 expression in lymphomatoid papulosis (LyP).

A. Distribution of LEF1 expression in LyP with and without DUSP22-R, respectively. B. Distribution of TIA1 expression in LyP with and without DUSP22-R, respectively. C. Distribution of pSTAT3 expression in LyP with and without DUSP22-R, respectively. D. LyP with DUSP22-R. H&E and CD30 show the typical biphasic pattern with a tumor mass in the dermis and an overlying epidermotropic infiltrate (original magnification, 20×). The tumor cells are positive for LEF1 and negative for TIA1 (original magnification, 40×).

Discussion

In this study, we developed and validated for the first time an IHC-based strategy for identifying DUSP22-R in ALCL. First, we showed that our previous finding identifying a strong correlation between LEF1 expression and DUSP22-R was strongly reproducible in an independent cohort spanning all WHO and genetic subtypes of ALCL. To develop an algorithmic approach to identifying DUSP22-R, we next combined LEF1 IHC with either TIA1 or pSTAT3 IHC, which are typically negative in DUSP22-R ALCL. Importantly, none of these markers alone is entirely specific for DUSP22-R. Using positive cutoffs established in previous studies, both combined approaches (LEF1/TIA1 and LEF1/pSTAT3) yielded PPV and NPV values of 100% in discovery and validation cohorts, with subsets of indeterminate cases that would obviate the need for FISH in 65% or 71% of ALK-negative ALCLs, respectively. Although the fraction of indeterminate results was marginally lower for pSTAT3, the LEF1/TIA1 model is more practicable since TIA1 IHC is clinically available in most laboratories and already is commonly used in the diagnostic evaluation of ALCL (3). Similarly, ALK is used routinely as a surrogate for ALK-R. We also previously showed that p63 IHC can be used in the evaluation of ALCL for TP63-R; using a positive cutoff value of ≥30%, p63 IHC showed 100% sensitivity for TP63-R and was positive in 35% of ALCLs, eliminating the need for FISH in the remaining cases (21). Taken together, these findings suggest an immunohistochemistry algorithm to facilitate comprehensive genetic subtyping of ALCL based on a 4-marker panel of ALK, LEF1, TIA1, and p63 (Figure 6).

Figure 6. Immunohistochemistry algorithm for genetic subtyping of ALCL.

The algorithm is based on a panel of 4 markers: ALK, LEF1, TIA1, and p63. Positivity for LEF1, TIA1, and p63 IHC are defined as: LEF1, >75% of tumor cells positive (24); TIA1, ≥20% of tumor cells positive (3); and p63, ≥30% (21). Green boxes indicate current WHO entities of systemic ALCL. Assessment of both DUSP22-R and TP63-R is recommended in ALK-negative ALCL. This algorithm can also be applied to localized forms of ALCL (primary cutaneous and breast implant-associated).

The use of LEF1 and TIA1 IHC to predict DUSP22-R had similar utility in an exploratory cohort of LyP, a CD30-positive T-cell lymphoproliferative disorder with some pathologic similarities to ALCL but with distinct clinical features including a waxing and waning clinical course. Similar to ALK-negative ALCLs, LEF1 and TIA1 provided 100% PPV and NPV and was indeterminate in 35% of cases, which would still require FISH to evaluate DUSP22-R. In practice, this percentage might be even lower since LyP with DUSP22-R also has characteristic clinical and morphologic features (31). Interestingly, TIA1 alone completely segregated LyP with and without DUSP22-R, whereas the difference in LEF1 expression was somewhat less than that seen among ALCLs. These observations merit further study in larger LyP cohorts.

CD30 expression can be seen in some T-cell lymphomas and lymphoproliferative disorders other than ALCL and LyP, including peripheral T-cell lymphoma, not otherwise specified (32). Entities other than ALCL and LyP were not included in the present study and the IHC markers tested were examined only for their ability to facilitate genomic subtyping of ALCL or LyP. Thus, our findings do not address whether these markers have any utility in the differential diagnosis of these entities.

DUSP22-R has been associated with favorable prognosis in ALK-negative ALCL in most studies (3, 9, 10, 12, 23). Likewise, most patients with systemic ALK-negative ALCL with DUSP22-R in the current series had favorable outcomes, and median overall survival time was not reached. High-risk cases with DUSP22-R also have been reported (13); the clinicopathologic and molecular features of high-risk ALK-negative ALCL with DUSP22-R merit additional study in larger series.

The choice of markers for use in our proposed IHC algorithm was based in part on clinical availability. ALK and TIA1 IHC is used routinely in the work-up of ALCL, while LEF1 and p63 are used in the diagnosis of B-cell lymphomas and solid tumors, respectively. Among cytotoxic markers we selected TIA1 as in a previous study it was most consistently expressed in ALCLs without DUSP22-R and also most commonly used in the diagnostic work-up (3). Biologically, cytotoxic marker expression is in part transcriptionally regulated by STAT3 and lack of cytotoxic marker expression in DUSP22-R ALCL likely reflects growing evidence of subgroup of ALCLs that lack JAK-STAT3 signaling and require further study (18). The biologic implications of LEF1 expression in DUSP22-R ALCL also should be explored further (24).

Our data raise the question of what level of accuracy is required when constructing algorithms for genetic subtyping in clinical practice. In general, IHC markers may lead to more uniform use of subtyping algorithms than FISH because of their broader availability and faster turnaround time. The use of ALK IHC in detecting ALK-R in ALCL represents a near-ideal case because of the robust detection of ALK fusion proteins and absence of native ALK protein in most normal tissues, although rare ALK protein-positive cases may lack ALK fusions (33). In other cases, such as the use of p63 IHC in identifying TP63-R, IHC facilitates a diagnostic approach to genetic subtyping but FISH is still required in a subset of cases (21). Importantly, false negative FISH results due to cryptic structural alterations also can occur, as in the case of cryptic MYC-R in B-cell lymphomas (34, 35); we also recently identified a FISH-cryptic TP63-R in a case of ALK-negative ALCL (unpublished observation). In the case of DUSP22-R, indeterminate FISH results occasionally occur in which the breakapart probe identifies loss of the 5'-DUSP22 signal and cannot distinguish between telomeric 6p deletion and an unbalanced DUSP22-R (36). In the future, platforms other than FISH might be employed in genetic subtyping. A gene expression signature successfully identified high-grade B-cell lymphomas with FISH-cryptic MYC-R or BCL2-R (35) and we previously developed and validated a gene expression signature for DUSP22-R in ALCL (18). As gene expression platforms become more widespread in diagnostic practice, a signature-based approach to genetic subtyping of ALCL may be clinically useful.