OUP user menu

T-Cell Large Granular Lymphocyte Leukemia Associated With Myelodysplastic Syndrome
A Clinicopathologic Study of Nine Cases

Yang O. Huh MD, L. Jeffrey Medeiros MD, Farhad Ravandi MD, Sergej Konoplev MD, Jeffrey L. Jorgensen MD, PhD, Roberto N. Miranda MD
DOI: http://dx.doi.org/10.1309/AJCP6YHI1JEXAWAP 347-356 First published online: 1 March 2009


We describe 9 patients with T-cell large granular lymphocyte leukemia (T-LGL) who also had a myelodysplastic syndrome (MDS). There were 6 men and 3 women with a median age of 60 years (range, 25–74 years). All patients had anemia at initial examination, 7 had neutropenia, and 5 had thrombocytopenia. The median absolute lymphocyte count was 1,300/μL (1.3 x 109/L; range, 700–3,600/μL [0.7–3.6 x 109/L]). Immunophenotypic analysis showed a CD8+ T-cell population, and molecular analysis showed monoclonal T-cell receptor gene rearrangement in every case. The MDS was classified as refractory cytopenia with multilineage dysplasia (RCMD, n = 5), refractory anemia (n = 2), RCMD with ringed sideroblasts (n = 1), and chronic myelomonocytic leukemia (n = 1). We compared the data for these patients with T-LGL/MDS with a group that had only T-LGL. The median hemoglobin level and absolute lymphocyte count were lower in patients with T-LGL/MDS (P < .05). The frequency of coexistent T-LGL and MDS at our institution suggests an etiologic relationship rather than simple coincidence.

Key Words:
  • T-cell large granular lymphocyte leukemia
  • Myelodysplastic syndrome
  • Cytopenia

Large granular lymphocyte leukemia (LGL) is characterized by a proliferation of morphologically distinctive large granular lymphocytes named for their large azurophilic cytoplasmic granules.1,2 There are at least 2 types of LGL: T-cell (T-LGL) and natural killer cell. T-LGL is the most common form and is an indolent disease that requires no treatment at the time of initial diagnosis in most patients, although up to 70% will ultimately require therapy.35 At diagnosis, patients with T-LGL usually have neutropenia, anemia, and an increased number of large granular lymphocytes in the peripheral blood.6,7 Most patients remain asymptomatic and are diagnosed incidentally by routine examination of the blood.4 However, some patients have symptoms related to neutropenia, such as recurrent infections, or splenomegaly that can lead to diagnosis.

Occasionally, patients with T-LGL have bone marrow failure with a severity comparable to that of patients with a myelodysplastic syndrome (MDS). Furthermore, others have reported small numbers of patients who have both MDS and T-LGL.8 In the largest study, Saunthararajah and colleagues8 described 9 patients who had both MDS and T-LGL and proposed a common pathogenesis for these 2 diseases.

We describe 9 additional patients who had simultaneous T-LGL and MDS. We describe the clinicopathologic, immunophenotypic, and cytogenetic findings in detail, unlike most other studies, and we compared the data for this subset of patients with data for a group of patients with T-LGL without MDS.

Materials and Methods

Study Group

We identified 42 patients with a possible diagnosis of T-LGL among patients evaluated at the University of Texas M.D. Anderson Cancer Center, Houston, between January 1998 and June 2006. After reviewing the clinical records, results of bone marrow morphologic examination, flow cytometry immunophenotypic findings, and results of molecular studies to assess for T-cell receptor (TCR) gene rearrangements, 26 patients were confirmed as having T-LGL leukemia, and their cases were previously reported.5 Within this group, 7 patients also fulfilled the World Health Organization (WHO) classification criteria for MDS. Subsequently, we encountered 2 additional patients who had simultaneous T-LGL and MDS, and the data for these patients are also included in this study.

Review of Morphologic Findings

A complete blood cell count (CBC) with a manual differential count was performed in all cases. May-Grünwald-Giemsa–stained peripheral blood smears were examined for the presence of large granular lymphocytes. Bone marrow core biopsy and aspirate specimens were obtained from the iliac crest in all cases. Bone marrow cellularity and the pattern of lymphocyte infiltration were determined on core biopsy or aspirate clot specimens stained with H&E; 500-cell differential counts and morphologic assessment for dysplasia were determined on Wright-Giemsa–stained aspirate smears. Iron stain was performed on bone marrow aspirate smears in all cases.

Immunophenotypic, Cytogenetic, and Molecular Methods

Flow cytometric immunophenotypic analysis was performed in our laboratory on peripheral blood or bone marrow aspirate specimens by standard multiparameter methods using antibodies specific for CD2, CD3, CD4, CD5, CD7, CD8, CD56, CD57, TCRαβ, and TCRγδ as previously described.9 In 3 cases, TCR Vβ chain analysis was performed using a TCR-Vβ Repertoire Kit (Beckman Coulter, Miami, FL). Immunohistochemical studies were performed in selected cases for CD3, CD4, CD5, CD8, CD20, TIA-1 (cytotoxic granule–associated RNA binding protein), and granzyme B using fixed, paraffin-embedded tissue sections of bone marrow core biopsy or clot specimens using methods previously described.10

Conventional cytogenetic analysis was performed on bone marrow aspirate samples of all patients using standard techniques as previously described.11 DNA was extracted from bone marrow aspirate specimens and analyzed for TCR gene rearrangements by various methods. Southern blot for TCRβ gene rearrangement studies was performed in 4 cases. Polymerase chain reaction methods were used to analyze the TCRγ and/or TCRβ chain genes in 7 cases using a mixture of 4 family-specific, multicolored, fluorescently labeled variable region (V) primers and 4 unlabeled joining region (J) primers, as described previously.12

Diagnostic Criteria for MDS and T-LGL

The criteria we used for the diagnosis of T-LGL were based on the WHO classification2 and the criteria proposed by Semenzato et al.7 These criteria include the following: persistent (>6 months) large granular lymphocytosis in peripheral blood, cytopenias (anemia, thrombocytopenia, and/or neutropenia), presence of more than 15% CD3+ large granular lymphocytes with coexpression of CD57, or detection of a monoclonal TCR gene rearrangement.

The criteria used for the diagnosis and classification of MDS were those specified in the WHO classification.13,14 These include morphologic evidence of dysplastic changes in the erythroid, megakaryocytic, or granulocytic lineages; percentage of blasts in peripheral blood or bone marrow or both; and the presence or absence of ringed sideroblasts (RS). We also included 1 case of chronic myelomonocytic leukemia (CMML), which is classified as a myelodysplastic/myeloproliferative disease (MDS/MPD) in the WHO classification.14

Statistical Analysis

The Fisher exact test was used to analyze differences between the T-LGL and T-LGL/MDS groups.


Clinical and Laboratory Findings

The clinical features for the 9 patients with T-LGL/MDS are summarized in Table 1. There were 6 men and 3 women with a median age of 60 years (range, 25–74 years). Three patients (cases 2, 6, and 9) were referred with the diagnosis of T-LGL ranging from 1 to 8 years’ duration (mean, 5.7 years). One of these patients (case 2) had a history of hemochromatosis and had been followed up with phlebotomies and observation only. Two patients (cases 6 and 9) had worsening neutropenia and anemia despite therapy. Six patients were referred to our institution with the diagnosis of MDS ranging from 2 months’ to 3 years’ duration (mean, 8.8 months). One patient (case 6) had rheumatoid arthritis, and 2 patients had a history of cancer: colon (case 1) and prostate (case 8). Physical examination revealed hepatosplenomegaly in 1 patient (case 9) and mild splenomegaly in another patient (case 6). There was no evidence of hepatosplenomegaly in the remaining 7 patients. There was no evidence of lymphadenopathy in any of the patients.

The laboratory findings are summarized in Table 2. The hemoglobin level ranged from 6.7 to 11 g/dL (67–110 g/L) with a median of 8.5 g/dL (85 g/L; reference range, 14–18 g/dL [140–180 g/L] for men and 14–16 g/dL [140–160 g/L] for women). All patients had anemia at initial diagnosis, and 3 patients (cases 5, 7, and 8) had Coombs-negative hemolytic anemia. The platelet count ranged from 3 to 438 x 103/μL (3–438 x 109/L) with a median of 138 x 103/μL (138 x 109/L; reference range, 140–440 x 103/μL [140–440 x 109/L]); 5 patients had thrombocytopenia. The absolute neutrophil count ranged from 20 to 2,700/μL (0.02 to 2.7 x 109/L) with a median of 1,220/μL (1.22 x 109/L). Seven patients (cases 1, 2, 4–6, 8, and 9) had neutropenia with an absolute neutrophil count of less than 1,700/μL (1.7 x 109/L). The absolute lymphocyte count ranged from 700 to 3,600/μL (0.7–3.6 x 109/L) with a median of 1,300/μL (1.3 x 109/L); lymphopenia was present in 2 patients (cases 3 and 8). No patients had lymphocytosis (>4,000/μL [4.0 x 109/L]). For 7 patients, peripheral blood smears were available for review to determine the percentage of large granular lymphocytes and calculate an absolute count. Large granular lymphocytes showed an intermediate size with oval nuclei, clumped chromatin, and variable number of cytoplasmic granules. The absolute large granular lymphocyte count ranged from 0.4 to 2.1 x 109/L with a median of 0.9 x 109/L (reference range, <0.80 x 109/L). Three patients (cases 1, 3, and 8) had a low large granular lymphocyte count.

View this table:
Table 1
View this table:
Table 2

Morphologic Findings

The bone marrow morphologic findings are summarized in Table 3. The T-LGL and MDS features are described here separately for the sake of clarity.

The diagnosis of T-LGL was supported by the presence of lymphocytosis and large granular lymphocytes. In bone marrow aspirate smears, the median lymphocyte count was 33% (range, 15%–64%). In 8 (89%) of 9 cases, the lymphocyte percentage was increased (reference range, 3%–17%). Lymphocytes were small, with a subset of slightly larger lymphocytes displaying moderately abundant clear cytoplasm with a variable number of granules Image 1A. In bone marrow biopsy specimens, the median cellularity was 50% with a range from 20% to 95%. All cases showed an interstitial lymphocytic infiltrate, with distinct lymphoid aggregates in a nonparatrabecular and nodular pattern identified in 6 cases. These aggregates were composed of small round to slightly irregular hyperchromatic lymphocytes; no large transformed cells were identified. Germinal centers were not identified.

The diagnosis of MDS in bone marrow aspirate smears was supported by the presence of dysplastic features in every case. These changes involved erythroid precursors and megakaryocytes in 7 cases each and granulocytes in 6 cases. Erythroid dysplasia consisted mostly of dyssynchronous nuclear to cytoplasmic maturation; less often, there were irregular nuclear outlines and nuclear budding. RS were identified and numerous in case 9. Megakaryocytic dysplasia consisted of hypolobulation or hyperlobulation and occasional loose clustering (Image 1D). Granulocytic dysplasia consisted mostly of nuclear hyposegmentation and cytoplasmic hypogranularity. The median blast count was 1%; 8 cases had a blast count of less than 5%, and 1 case had a blast count of 6% (reference range, <5%). In bone marrow biopsy specimens, occasional loose clusters of megakaryocytes were identified. The MDS was classified as follows: refractory cytopenia with multilineage dysplasia (RCMD; n = 5), refractory anemia (RA; n = 2), and RCMD-RS (n = 1), and 1 case of MDS/MPD was classified as CMML.

View this table:
Table 3
Image 1

(Case 9) Bone marrow and peripheral blood findings in a case with refractory cytopenia with multilineage dysplasia with ringed sideroblasts. A, Large granular lymphocytes are present in a peripheral blood smear (Wright-Giemsa, x1,000). B, Dysplastic megakaryocyte (inset) and erythroid cells are seen in a bone marrow aspirate smear (Wright-Giemsa, x500). C, Ringed sideroblasts are seen with iron stain in a bone marrow smear (x500). D, Bone marrow core biopsy specimen showing dysplastic megakaryocytes and interstitial lymphoid infiltrate (H&E, x400). E and F, Interstitial lymphocytes are positive for CD3 by immunoperoxidase stain (E, x400) and for CD8 (F, x400).

Immunophenotypic Findings

The immunophenotypic results supporting the diagnosis of T-LGL are summarized in Table 3. By using flow cytometric immunophenotypic analysis, in each case, an aberrant T-cell population was identified Image 2. Except for 1 case (case 5), T cells expressed CD3 and CD8 and were negative for CD4, CD19, and CD20. CD2 was expressed in all 6 cases tested, and CD5 was expressed in 5 of 6 cases tested, with dim intensity in 3 cases. CD7 was expressed in all 7 cases tested, and the intensity was dim in 3 cases. All 7 cases tested were positive for TCRαβ and negative for TCRγδ. Of 8 cases tested, 5 (63%) were CD57+. The expression of CD57 was variable; it was expressed in more than 50% of CD3+ lymphocytes in 2 cases and in a subset (range, 21%–29%) of CD3+ lymphocytes in 3 cases. CD56 was negative in the 8 cases tested. TCR Vβ chain analysis was performed in 3 cases, and the results were monoclonal in each case, with preferential use of Vβ8, Vβ9 and Vβ13.6, or Vβ7.1 Image 3. CD55 and CD59 were assessed in 1 case (a patient with hemolytic anemia) and were within the reference range.

Immunohistochemical studies were performed in 7 cases. An interstitial infiltrate of CD3+ T cells was identified in all 6 cases tested (Image 1E), and the infiltrate was also focally nodular in 5 cases. TIA-1 and CD8 were positive in an interstitial pattern in all cases assessed, 5 and 4, respectively (Image 1F). Nodules of lymphocytes had relatively fewer TIA-1+ and CD8+ cells. Granzyme B was negative in 2 tested cases. CD20 highlighted cells mainly within lymphoid nodules and was negative in interstitial lymphocytes.

Conventional Cytogenetics and Molecular Results

Conventional cytogenetic analysis showed abnormal karyotypes in 2 patients. In case 4, del(6)(q21q25) was identified. In case 6, a complex karyotype was shown: 47,XY,del(6)(q21),t(6;22)(q13;q13),add(7)(p22),add(9) (p24),+14,–18,+mar. The remaining 7 patients had a normal diploid karyotype.

Molecular analysis to assess for TCR gene rearrangements was performed in all 9 cases and detected monoclonal TCR gene rearrangement in all cases (Table 3). In 4 cases, Southern blot hybridization was used to demonstrate monoclonal TCRβ chain rearrangements. In 8 cases, PCR methods were used to demonstrate TCRγ gene rearrangements that used the Vγ1 (n = 5), Vγ2 (n = 2), Vγ3 (n = 3), and Vγ4 (n = 1) families.

Clinical Follow-up

The median follow-up was 58 months from the time of diagnosis (range, 4–138 months). Three patients died. A 68-year-old woman (case 1) was referred to our institution with the diagnosis of CMML, and we established the coexistent diagnosis of T-LGL. She had pancytopenia that was refractory to drug therapy and had a splenectomy. She died 4 months after being seen at our hospital. A 54-year-old woman (case 4) was referred to our institution with the diagnosis of MDS, and we further classified her disease as T-LGL and RCMD. She had a history of deep venous thrombosis and asthma for which she received high-dose prednisone, complicated by avascular necrosis of multiple bones. She had a 3-year history of pancytopenia and was recently dependent on erythrocyte and platelet transfusions. This patient died of disease and of pulmonary complications 4 years after diagnosis. A 25-year-old man (case 9) was referred to our hospital with an 8-year history of T-LGL, and we established the simultaneous diagnosis of T-LGL and RCMD-RS. He was treated with a variety of agents at various times for neutropenia and anemia before referral to our institution. However, his cytopenias were refractory to therapy. Despite additional therapy, this patient died with pancytopenia and infectious complications 9 years after initial diagnosis. The therapies for these 3 patients at our institution included azacitidine, cyclosporin, prednisone, and erythropoietin (EPO) (case 1); alemtuzumab, EPO, and intravenous immunoglobulin (case 4), and pentostatin with alemtuzumab (case 9).

Image 2

Flow cytometric immunophenotype. Atypical T cells are CD3+/CD5–/CD8+/CD56–/CD57+/TCRαβ+. APC, allophycocyanin; FITC, fluorescein isothiocyanate; PE, phycoerythrin; PerCP, peridinin chlorophyll protein; SSC, side scatter; TCR, T-cell receptor.

Six patients were alive at last follow-up, ranging from 18 months to 11.5 years. The MDS component in these patients was classified as RA (cases 2 and 3) and as RCMD (cases 5–8). All patients received EPO, and 6 patients received cyclosporin at some point. Three patients who had hemolytic anemia responded to prednisone, one of whom also was treated with rituximab (case 7) and another with azacitidine (case 8).

Comparison of T-LGL/MDS and T-LGL Groups

A comparison of the clinicopathologic and laboratory findings in patients with T-LGL/MDS and T-LGL alone is summarized in Table 4. The median hemoglobin in the T-LGL/MDS group was 8.5 g/dL (85 g/L; range, 6.7–11 g/dL [67–110 g/L]) vs 10.3 g/dL (103 g/L; range, 7.5–16 g/dL [75–160 g/L]) in the T-LGL group (P = .0138). All 9 patients (100%) were anemic in the T-LGL/MDS group vs 8 (42%) of 19 in the T-LGL group (P = .004). The median absolute lymphocyte count was 1,300/μL (1.3 x 109/L; range, 700–3,600/μL [0.7–3.6 x 109/L]) in the T-LGL/MDS group vs 2,900 (2.9 x 109/L; range, 600–2,400/μL [0.6–24 x 109/L]) in the T-LGL group (P = .029). Lymphopenia was found in 2 (22%) of 9 patients with T-LGL/MDS vs 2 (11%) of 19 patients with T-LGL. Lymphocytosis, not present in any of the 9 patients with T-LGL/MDS, was identified in 6 (67%) of 9 patients with T-LGL, but the difference was not statistically significant (P > .12). Patients with T-LGL/MDS were more frequently men (67%) than were patients in the T-LGL group (47%) (P, not significant). The frequencies of neutropenia and thrombocytopenia were similar in both groups. No significant differences in patient age, bone marrow cellularity, percentage of blasts, percentages of lymphocytes or erythroid precursors, and the frequency of lymphoid aggregates were observed between the T-LGL/MDS and T-LGL groups.

Image 3

T-cell Vβ chain analysis by flow cytometry; 91.5% of T cells are positive for Vβ8 only in vial D out of 24 antibodies in the panel.

View this table:
Table 4


We have described a heterogeneous group of 9 patients who had T-LGL and MDS simultaneously. The coexistence of T-LGL and MDS has been reported rarely. Dhodapkar and colleagues15 identified 5 patients with T-LGL and concomitant MDS. Saunthararajah et al8 described 9 patients with coexistent T-LGL and MDS. Although rare, it seems likely that the simultaneous occurrence of T-LGL and MDS may be somehow related rather than a chance event. Saunthararajah et al8 suggested 2 possible explanations: (1) T-LGL might arise from a clonal MDS stem cell. (2) T-LGL could represent an autoimmune response to an antigen expressed by normal or MDS bone marrow cells. In both of these scenarios, one would hypothesize that the MDS arose first, followed by T-LGL. In the series of cases we report, we cannot establish a sequence to the development of T-LGL and MDS. The morphologic features of MDS in our cases, although present, were not overt, nor did they affect most of the cells of an affected lineage. In addition, 3 patients had long histories of T-LGL before the diagnosis of MDS was established. If anything, these findings suggest that T-LGL can precede the onset of MDS.

Saunthararajah and colleagues8 have suggested that T-LGL associated with MDS may be more frequent than is appreciated. They suggest that analysis for clonality of TCR genes in all patients with MDS would be useful in identifying patients with simultaneous MDS and T-LGL. However, the detection of a monoclonal T-cell population by molecular methods in this patient group is not specific for T-LGL or other T-cell neoplasm. Scott et al16 reported that 31% of patients without hematologic disorders can have a persistently increased percentage (>25%) or number (>1 x 109/L) of large granular lymphocytes that are CD8+. Furthermore, in approximately 50% of these cases, a monoclonal TCR gene rearrangement can be detected.16 For these monoclonal, nonleukemic borderline cases, the term “T-cell clonopathy of undetermined significance” has been proposed.15 Some elderly people have persistent monoclonal expansions of CD8+ T cells, and overt T-LGL usually does not develop in them.17 Others have observed that monoclonal CD8+ T large granular lymphocyte proliferations can be associated with autoimmune disorders, such as rheumatoid arthritis, Sjögren syndrome, and Hashimoto thyroiditis.18 Thus, detection of a monoclonal TCR gene rearrangement is not, in itself, sufficient to establish the diagnosis of T-LGL in a patient with MDS. The patients we describe were all symptomatic and required therapy during their clinical course, and only 2 patients had autoimmune disease.

The T-LGL component in the cases we report was of low tumor burden. Only 4 (44%) of 9 patients had lymphocyte counts of more than 2,000/μL (2 x 109/L). Furthermore, on review of peripheral blood smears in 7 patients, only 1 patient (14%) had a large granular lymphocyte count greater than 2 x 109/L, 3 patients had large granular lymphocytes in the normal range (<0.8 x 109/L), and 2 patients had lymphopenia. Semenzato et al7 described 11 patients with T-LGL with LGL counts ranging from 0.5 to 1.85 x 109/L and bone marrow lymphocytosis occurring in only 4 cases (36%). Saunthararajah et al8 described 9 patients with T-LGL/MDS in whom the absolute number of LGLs ranged from 0.08 to 1.01 x 109/L. Chan and colleagues2 state in the WHO classification that there is currently no agreement on the minimum level of lymphocytosis required for the diagnosis of T-LGL. They note that patients with T-LGL usually have an increased number of peripheral blood lymphocytes, usually between 2,000 and 20,000/μL (2–20 x 109/L), with large granular lymphocytes being the predominant cells. In a clinicopathologic study of patients with T-LGL by Lamy and Loughran,19 LGL counts were greater than 5 x 109/L in 50% of cases, 1 to 4 x 109/L in 40% of cases, and less than 1 x 109/L in 8% of cases.

We compared this group of patients with a group of patients at our hospital who had only T-LGL and were described previously.5 The median hemoglobin level and absolute lymphocyte count were lower in the T-LGL/MDS group compared with the T-LGL group (P < .05). The frequency of neutropenia, lymphopenia, lymphocytosis, and thrombocytopenia and bone marrow findings were similar (P > .05) for the T-LGL/MDS and T-LGL groups. Therefore, it seems that lymphocytes and large granular lymphocytes are present at lower levels in patients with T-LGL/MDS. The explanation for fewer lymphocytes in these patients is unknown. It is possible that the presence of MDS somehow modulates the number of large granular lymphocytes in the blood. Another possibility is that the availability of sensitive flow cytometry immunophenotypic and molecular techniques allows the diagnosis of T-LGL at an earlier point in its disease evolution.7

Conventional cytogenetic studies in this series showed del(6)(q21) in 2 patients. This abnormality previously has been reported associated with T-LGL, but not with MDS.20 In addition, the t(6;22)(q13;q13) found in 1 patient in this study has been reported in a patient with leukemic myeloma.21 Saunthararajah et al8 identified cytogenetic abnormalities in 2 of 9 patients with T-LGL/MDS in their study, a frequency identical to that for our patient group. However, the abnormalities in their study were different: add(2)(p13) and –18 in a case of RA and 20q– in a case of refractory anemia with excess blasts. The frequency of cytogenetic abnormalities in our study is in keeping with that reported for patients with T-LGL. Semenzato et al7 reported cytogenetic abnormalities in 13% of 195 patients with T-LGL. The frequency of cytogenetic abnormalities in our patient group and in the study by Saunthararajah et al8 is lower than would be expected in patients with MDS overall but is consistent with the frequency of cytogenetic abnormalities in low-grade MDS. Of interest, 6 patients in our study had follow-up bone marrow examinations at a mean interval of 10 months after initial diagnosis. All 6 patients had persistent morphologic changes of myelodysplasia, suggesting that these changes were not transient or of a reactive nature.

T-cell LGL and MDS can occur simultaneously. Patients have heterogeneous clinical manifestations and outcome. Patients with T-LGL/MDS tend to have low hemoglobin levels and absolute lymphocyte counts, significantly lower than patients with T-LGL alone. The MDS component was characterized by a low blast count (median, 1%), and RCMD was the most frequent morphologic category. Immunophenotypic and molecular studies were very helpful for establishing the diagnosis of T-LGL. Conventional cytogenetic abnormalities were infrequent, but del(6q21) was identified in both patients with abnormal cytogenetic results in this study.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
View Abstract