ENKTL has a significant higher incidence in Asian and Latin American countries than the Western [1, 2, 24]. From the published series of 1983 cases of malignant lymphoma from Thailand, T-cell lymphoma accounted for 23% among overall new lymphoma cases , but the frequency of ENKTL was not reported due to the lack of markers for making a definite diagnosis. However, there were 69 cases of the so-called "angiocentric lymphoma" in the REAL classification, a predecessor of ENKTL, among the 381 cases of mature T-cell lymphoma in the series, thus the frequency of angiocentric lymphoma was approximately 18% of mature T-cell lymphoma, the second most common mature T- and NK-cell lymphoma after PTCL, NOS, according to the previous version of WHO classification (2001) . A recently published article from Thailand showed ENKTL as the most common among the mature T- and NK-cell lymphomas , accounting for 31%, which is comparable to Taiwan (26%), Korea (37%) and Hong Kong (39%) [26–28], but it is more common when compared to China (up to 4.7%) , Japan (6%)  and India (8%) . The difference of ENKTL proportion between Thailand and India may be due to the difference in the ethnic origin. In most parts of Japan, proportion of ENKTL is not high, but it is the second most common mature T- and NK-cell lymphoma in Okinawa; and if exclude cases of adult T-cell lymphoma/leukemia that is ultimately higher than other countries, ENKTL proportion in this region will be 34% and seems to be comparable to Thailand, Taiwan, Korea and Hongkong . The reasons for a lower ENKTL proportion in most parts of Japan and mainland China are not known.
In the study period of the present study, besides the 31 ENKTL cases, there was only one extranodal EBV-negative T-cell lymphoma identified from head and neck region. It was a PTCL, NOS, involving nasopharynx, and was positive for CD3, CD4, CD5, TIA-1, βF1, but negative for CD8, CD20, CD56, CD30 and ALK. Unfortunately, TCR gene rearrangement could not be testified in this case due to the inadequate extracted DNA.
ENKTL in the present study showed a strong male predilection with male to female ratio of 2.9, similar to the other previous reports [5, 24]. In the present study, the most common presenting symptom was nasal obstruction, similar to the other previous reports [5, 24]. The most common site of involvement was nasal cavity while the second most common site was nasopharynx. Nearly 84% of the cases (26/31) involved nasal cavity and/or nasopharynx. In Thailand, a previous study showed that ENKTL was the most common lymphoma in nasal and paranasal sinuses while it might be less common than diffuse large B-cell lymphoma in nasopharynx . In the present study, 2 of 31 cases (7%) had periorbital soft tissue involvement; both cases showed periorbital soft tissue involvement together with mild haziness of paranasal sinuses demonstrated by CT scan, while no lesion was observed from endoscopic examination. Unfortunately, no pathologic sample was sent for proving the occurrence of nasal disease. Thus, it is not possible to conclude whether these 2 cases had concomitant paranasal sinus involvement or they were a genuine primary orbital lesion which was rarely reported .
ENKTL with bone marrow involvement at presentation is rare. There were 2 out of 29 cases (7%) in the present study, similar to other series [16, 24, 34, 35]. In particular cases with obvious marrow involvement, distinction between an aggressive NK-cell leukemia and a leukemic phase of ENKTL should be problematic . A presence of nasal or nasopharygeal lymphoma would be favored for ENKTL.
The present study demonstrated a high proportion of cases with necrosis, similar to most other studies, while evidence of angioinvasion varied [16, 34, 36]. The varied proportion of necrosis among these studies presumably depends on tissue sampling whether necrotic area was included in a small biopsy or not. For the varied proportion of angioinvasion, it could be resulted from either a subjective determination by pathologists or from the amount of tissue obtained for evaluation.
Cytological aspects of ENKTL varied from series to series. In the present study, most cases (77%, 17/22 cases) had predominantly medium-sized cells or mixed medium-sized and large cells. The lymphoma with medium-sized cells commonly showed oval irregular nuclei with occasional nuclear elongation, inconspicuous nucleoli. Cases with large number of small cells might be difficult to distinguish from reactive processes, as a case found in the present study had been misdiagnosed as fungal infection in the first biopsy. In this situation, even presence of necrosis and fungi in the nasal cavity, making diagnosis of fungal infection without careful evaluation of the lymphoid component was dangerous. Thus, careful evaluation of lymphoid cells along with immunohistochemistry and EBER in situ hybridization might be helpful for making a diagnosis of lymphoma. In our experience, a few cases showing monotonous small lymphoma cells or anaplastic morphology were also observed, but they were not in the study period; such cases are uncommon as described in the WHO classification blue book .
An advantage of TMA technique, other than cost and time effectiveness, is that the tissue samples can act as multiple controls within a single slide. The relatively small tissue did not affect the interpretation much, since most samples were from small cup biopsies. According to the TMA technique, most of the cases in the present study had typical ENKTL immunophenotype but with minor differences to other series. The details were discussed below.
For the T-cell markers, CD3 is a highly sensitive marker and helpful for diagnosing ENKTL. However, CD3-negative ENKTL were account for 9% (2/22 cases) in the present study, which were also previously documented from other series [6, 15, 37]. Thus, negative for CD3 does not rule out ENKTL. For other T-cell markers, CD5 is negative in almost all ENKTL, similar to previous study . The present study had only 2 partially CD5-positive cases. This could be resulted from CD5 on reactive T-cells which sometime difficult to distinguished from neoplastic cells. Since there was no any case with expression of β TCR protein as well as the absence of demonstrable TCR gene rearrangement, these results probably reflect that at least almost all cases were genuine NK-cell in origin. However, we did not have antibodies react with γ or δ TCR to determine whether they were γδ T-cell derived ENKTL which they might have similar basic immunophenotype to that of the NK-cell derived ENKTL. In contrast to CD5, CD7 is normally expressed by normal NK-cells, but the present study demonstrated that most of ENKTL (68%) lack CD7 expression, which is similar to the previous studies [29, 38]. For CD43, even though it is a highly sensitive marker for ENKTL, as demonstrated in the present study as well as in other series [7, 18], it is less useful for subclassifying of T- and NK-cell neoplasms. For CD4, we found a few cases (14%) with some staining as designated as partially positive. However, in these cases, distinction from CD4 expression by the admixed reactive histiocytic and dendritic cells might be difficult. For CD8, there was no positive case in the present series. However, a previous study demonstrated up to 22% (9/41 cases) of CD8-positive ENKTL which those cases were also positive for CD56, but only 2 of 8 cases had clonally rearranged PCR-based TCRγ gene .
CD56 is a very helpful marker for diagnosing of ENKTL, especially for distinguishing from reactive processes which usually have only rare or a few scattered CD56-positive small cells. The present study demonstrated a high sensitivity of CD56 immunohistochemical staining as 82% (18/22 cases), comparable to some previous studies such as 74% (31/42 cases) by Ng et al and 82% (37/45 cases) by Ko et al [34, 36]. Higher CD56 sensitivity has also been reported such as 97% (36/37 cases) by Barrionuevo et al  and 100% (22/22 cases) by Kuo et al . Interestingly, a study using flow cytometric immunophenotyping demonstrated all ENKTL cases expressed CD56 . Furthermore, a study on frozen section immunohistochemistry also demonstrated all cases of ENKTL had CD56 expression . Hypothetically, the variation in detection of CD56 expression possibly reflects false negative results from immunohistochemical technique performed on FFPE tissue, since this technique might not be able to detect the cases with weak CD56 expression. In addition, durability of CD56 antigen in tissue, sensitivity of antibody clones, and staining techniques should be considered to improve the immunophenotypic results.
For the cytotoxic granules, as shown in Table 2 and 4, TIA-1 is the most sensitive cytotoxic protein in ENKTL, which some other studies also demonstrated 100% sensitivity of TIA-1 expression [18, 35, 36, 41]. On the other hand, some studies showed the lacking of TIA-1 expression in a few cases [34, 37]. In some series, granzyme B was expressed in all cases tested [18, 42]. In addition to the highest sensitivity of TIA-1, it also was the most intense staining when compared to granzyme B and especially to perforin, based on the present study. Perforin is the least sensitive cytotoxic marker has also been reported . In term of lymphoid biology, since both granzyme B and CD30 were considered as activated-cell markers [43, 44], and they were found to be positive in most cases. These results support an activated NK-cell phenotype of ENKTL, as it is believed. Moreover, a recent study on gene expression profiling also demonstrated that ENKTL appeared to have a similar expression profile to activated NK-cell rather than a normal non-activated one .
CD30 was positive in 75% (15/20 cases) of cases in the present study, higher than other previous studies such as 48% (13/27) by Ko et al and 41% (9/22) by Kuo et al [34, 35]. When compared to the CD30-negative cases, there was no significant correlation of CD30 with other immunostaining, patient's age or site of involvement. Cautiously, it is a potential difficulty for distinguishing between CD56-, CD30+ ENKTL and CD56+ anaplastic large cell lymphoma (ALCL), since 15% of ALCLs can be positive for CD56 . In this situation, presence or absence of EBV or ALK protein may be helpful for making a definite diagnosis.
CD45 (leukocyte common antigen) was not expressed by some ENKTL in the present series. And this should not be confused with a non-hematologic malignancy. To the best of our knowledge, loss of this marker on paraffin-section immunohistochemistry has never been reported.
For B-cell markers, CD20, CD79a, PAX5 and BOB.1 were generally negative. Interestingly, some cases were positive for Oct2, most of them showed focal and weak positivity, however, a case with diffusely- and strongly-positive lymphoma cells was documented. The expression of Oct2 in 1/2 ENKTL as well as some other types of T-cell lymphomas was previously reported by Saez et al . But its significance is not known. Therefore, study in more cases as well as in biologic detail might be of interest.
IRF4/MUM1 is normally expressed in melonocytes, plasma cells, B-cells and activated T-cells . Besides its expression in B-cell neoplasms, a variable IRF4/MUM1 expression in systemic and cutaneous T-cell lymphoproliferative disorders was reported . The IRF4/MUM1 expression has also been reported in 1 of 3 ENKTL by Natkunam et al . While more recent studies from other regions of the world did not evaluate IRF4/MUM1 expression in their case series [3, 15]. The significance of IRF4/MUM1 expression is not known, and might be considered for further study in the future.
Of the 18 cases with PCR-based TCR gene rearrangement study, which using BIOMED-2 primer design, none of them showed positive result from both gel and capillary fluorescence electrophoresis, which similar to some previous studies [37, 51, 52]. On the other hand, some studies using PCR-based techniques, demonstrated TCR-gene rearranged ENKTL in a variable proportion, such as 8% (1/12 cases) by Gaal et al , 9% (7/74) by Gualco et al , 10% (3/31) by Ko et al , 27% (11/41) by Ng et al , 30% (3/10) by Lin et al , and up to 71% (10/14) by Mitarnun et al . The varied results of PCR-based TCR gene rearrangement among these studies could be caused by the differences in techniques and designed primers. While using Southern blot analysis, the gold standard for TCR gene rearrangement analysis, Suzumiya et al demonstrated rearrangement only in 8% (1/13) of cases, but with the same technique, Nakamura et al failed to demonstrate rearrangement in all their 6 cases [6, 7].
For the specificity of PCR-based TCR gene rearrangement, it is noteworthy to emphasize that positive results can also be found in a significant proportion of B-cell lymphomas [55, 56], acute non-lymphoid leukemia , and even in reactive lymphoid tissues . Thus, cases with positive PCR-based TCR gene rearrangement should not be always indicated as a T-cell lineage.
Interestingly, a study revealed genuine γδ T-cell lines of nasal ENKTL, verified by demonstration of surface γδ TCR by flow cytometry and immunophenotypic studies, and was also positive for TCR gene rearrangement . Such cases may be responsible for some ENKTL with rearranged TCR gene, together with a few cases of commonly known αβ-TCR-positive ENKTL. However, methods for γ- or δ- TCR staining for demonstration of γδ T-cell origin were not available in the present study.
EBV subtype A was noted from all tested cases (15 cases). As far as we know, the present study is the first to demonstrate EBV subtype of ENKTL in Thailand. The high prevalence of this subtype is similar to those reports from East and Southeast Asia [2, 5, 6, 10–12]. In the non-Asian countries, while subtype B was found in the majority of ENKTL in Peru, USA, and Germany [13, 16, 18], but subtype A was predominant in Mexico, Chile, Brazil, and Spain [14, 15, 17, 19]. Furthermore, Gualco et al recently described the striking differences of EBV subtype among geographic regions in Brazil which may reflect the heterogeneity of the ancestral population .
In the present study, 87% of cases were found to be the deletion type of EBV LMP-1 genes, similar to other studies [5, 6, 12, 35]. Since LMP-1 protein on EBV-positive lymphoid cells is an antigenic target for cytotoxic T-cells, high frequency of the deletion-type in ENKTL compared to reactive conditions which wild-type are more common, might suggest a clonal selection of immunologically escapable EBV-infected cells in neoplastic processes .
The present study also demonstrated a highly aggressive clinical course of ENKTL which was widely known. Upfront radiotherapy was proved by recently published studies to improve the survival outcome [5, 8]. However, most of the patients in this series received only chemotherapy, and there were too few patients received combined chemo-radiation therapy to do a statistic comparison with.
The present study demonstrated a significant poor survival outcome in cases without perforin expression. The cause of aggressive behavior is not known, but loss of perforin expression possibly reflects the more complex genetic abnormality in the tumor cells.