OUP user menu

An Updated Approach to the Diagnosis of Myeloid Leukemia Cutis

Danielle M. P. Cronin MD, Tracy I. George MD, Uma N. Sundram MD, PhD
DOI: http://dx.doi.org/10.1309/AJCP6GR8BDEXPKHR 101-110 First published online: 1 July 2009


The diagnosis of myeloid leukemia cutis can be difficult, particularly in the context of an initial skin biopsy with a malignant hematopoietic neoplasm. We studied the immunohistochemical characteristics of 33 cases of myeloid leukemia cutis diagnosed at Stanford University Medical Center, Stanford, CA, 1996–2007, and compared them with the corresponding bone marrow blast immunophenotype and World Health Organization classification (2008). In the skin, CD43 marked 97% of cases (32/33), myeloperoxidase marked 42% (14/33), CD68 marked 94% (31/33), CD163 marked 25% (7/28), and CD56 marked 47% (14/30). CD34 and CD117 were predominantly negative. In 19 cases in which myeloperoxidase was negative, all marked with CD68 and CD43. The flow cytometric immunophenotype of the leukemic blasts in the bone marrow was discordant with the immunohistochemical profile in the skin in all cases, showing loss or gain of at least 1 antigen. Given the immunophenotypic differences between skin and bone marrow blasts, we provide an updated immunohistochemical approach to the diagnosis of myeloid leukemia cutis.

Key Words:
  • Myeloid leukemia cutis
  • Myeloid sarcoma
  • Immunohistochemistry
  • CD34
  • Flow cytometry
  • World Health Organization classification
  • Blastic plasmacytoid dendritic cell neoplasm
  • CD4+/CD56+ hematodermic neoplasm

Leukemia cutis (LC) refers to the specific infiltration of the skin by neoplastic leukemic cells, most often in conjunction with systemic leukemia.1,2 There is debate about the appropriate diagnostic terminology for LC, and various authors have used historic, related, and/or overlapping terms, including chloroma, extramedullary myeloid tumor, granulocytic sarcoma, and myeloid sarcoma. The term leukemia cutis is perhaps the most broad and is frequently used in the dermatopathology literature. In studies using biopsy-confirmed cases of LC, the prevalence of this disease is 2% to 3% in the setting of systemic acute leukemia. Skin involvement in myeloid leukemia is associated with monocytic differentiation, central nervous system involvement, and aneuploidy of chromosome 8.24 Very rarely, LC may precede clinically detectable systemic leukemia (so-called aleukemic leukemia cutis).1

Clinically and histologically, myeloid LC can be confused with a wide array of skin conditions. Clinically, lesions of myeloid LC are characterized by single or multiple papules, nodules, or plaques that may range from violaceous to red-brown.5 The clinical differential diagnosis is broad, as a wide range of neoplastic, inflammatory, and infectious skin lesions may be associated with hematologic malignancies and their treatment. In addition, myeloid LC lesions are histologically heterogeneous but typically characterized by a primarily interstitial infiltrate of leukemic cells involving the dermis, sometimes extending into the subcutis. Perivascular and/or periadnexal arrangements can also be seen.6 The blasts are medium sized to large with smooth chromatin, prominent nucleoli, and increased mitotic activity.7 Among cutaneous neoplastic hematologic conditions, the histologic differential diagnosis includes lymphomas with high-grade morphologic features (B- or T-lineage lymphomas, CD30+ lymphoproliferative disorders), blastic plasmacytoid dendritic cell neoplasm (also known as agranular CD4+/CD56+ hematodermic neoplasm), mast cell sarcoma, and high-grade plasma cell neoplasms.

In the absence of established systemic acute leukemia, confirming the diagnosis in cases of suspected myeloid LC can be difficult. In all cases, judicious use of immunohistochemical staining is of paramount importance.813 With this in mind, we examined a series of cases of myeloid LC by immunohistochemical staining and compared these findings with the flow cytometric and cytogenetic findings in the corresponding bone marrow specimens. Our goal was to identify an appropriate panel of commercially available immunohistochemical stains that may be helpful in the diagnosis of myeloid LC in paraffin-embedded tissues. We also correlated these findings with the current World Health Organization (WHO) classification of acute myeloid leukemia.9,14

Materials and Methods

Case Selection

The pathology database of Stanford University Medical Center, Stanford, CA, was searched for cases of myeloid LC. For the study, 33 cases of myeloid LC diagnosed between 1996 and 2007 had paraffin-embedded materials available. The cases selected all met the common dermatopathology definition of LC, ie, they all had specific blastic infiltrates of varying densities.1 Corresponding bone marrow aspirate, cytogenetic and/or molecular data, and flow cytometric data were obtained when available, and the diagnosis was classified by using the 2008 WHO classification of tumors of hematopoietic and lymphoid tissues.14 Other clinical data obtained about the patients included age, sex, location of the biopsy, clinical appearance of the lesions, duration between skin and bone marrow biopsies, and the presence or absence of intervening therapy. This study was approved by the Stanford University Institutional Review Board.

Immunohistochemical Studies

H&E-stained sections and appropriate immunohistochemical studies performed at the time of diagnosis were reviewed to confirm the original findings. Additional immunohistochemical studies were performed on formalin-fixed, paraffin-embedded tissue using a Ventana Benchmark semiautomated stainer (Ventana Medical Systems North America, Tucson, AZ) or an automated DAKO polymer-based detection system (DAKO North America, Carpinteria, CA), per the manufacturer’s protocol as previously described.15 Primary antibodies directed against CD3, CD20, CD34, CD43, CD56, CD68, CD117, CD163, and myeloperoxidase (MPO) were used. The antibody sources, dilutions, epitope retrieval procedures used before incubation with the primary antibody, automated stainer used, and staining patterns are summarized in Table 1. We use 2 protocols for CD34 and CD117 staining at our institution: a standard protocol titered against vascular endothelium (CD34 standard) or mast cells (CD117 standard) and a bone marrow protocol optimized for the detection of bone marrow blasts (CD34 BM and CD117 BM) (Table 1). Staining for all markers was defined as follows: positive, moderate to intense staining of at least 5% of lesional cells; and negative, faint staining of fewer than 5% of lesional cells to no staining of lesional cells.15 Diffuse faint staining was not seen in any of the cases. For the purposes of this study, we evaluated immunohistochemical staining in malignant infiltrates only and did not include reactive conditions occurring in patients with leukemia.

View this table:
Table 1

Positive and negative control samples were present in all cases for all antibodies. Lymphocytes served as positive internal control samples for CD20, CD3, and CD43; basal melanocytes for CD117; cutaneous vessels for CD34; and cutaneous nerves for CD56. Neutrophils in tonsils served as positive external control samples for MPO and histiocytes in tonsils for CD68 and CD163. The epidermis (except for basal melanocytes) served as a negative internal control sample for all antibodies.


The mean age of the 33 patients selected for this study was 53 years (range, 22 days to 90 years). There was a slight male predominance, with 20 males and 13 females. Using available data, the results of the bone marrow specimens for 29 patients were examined and the lesions classified according to published WHO criteria Table 2.14 The corresponding clinical information, bone marrow biopsy diagnosis (29 patients), and cytogenetic and/or molecular information (18 patients) is shown in Table 3. As in previous studies, a subset of cases (5/29) showed monocytic differentiation,2,6,13 and 5 of 18 cases with available cytogenetic information had numeric abnormalities of chromosome 8.2,16,17 In addition, 7 cases were associated with myelodysplasia-related changes, and 3 cases had normal bone marrow findings. Outside of these findings, WHO subtype and/or cytogenetic specific correlations with development of myeloid LC were not seen.

The histologic findings of myeloid LC were relatively uniform and similar to previously published reports.7,13 Sections showed an unaffected epidermis with a grenz zone Image 1. The proliferation was composed primarily of an interstitial infiltrate of immature malignant cells with increased nuclear/cytoplasmic ratios. Rarely, there was folliculotropism, but, in general, the infiltrate was arranged in periadnexal and/or perivascular configurations. High-power examination revealed enlarged nuclei with finely dispersed chromatin, variably prominent nucleoli, and, frequently, irregular nuclear contours. The histologic characteristics could not be used to further classify the lesions.

The results of the immunohistochemical studies are shown in Table 4 and Image 1. CD43 was expressed in nearly all cases (32/33 [97%]). We also found that a high percentage of cases expressed CD68, regardless of their WHO classification and MPO status (31/33 [94%]). In contrast, the antibody for CD163 marked only 7 (25%) of 28 cases. We found that MPO was present in 14 (42%) of 33 cases. The 19 cases that did not express MPO, however, all expressed CD68 and CD43. CD56 was expressed in 14 (47%) of 30 cases and was also not associated with a specific WHO classification. In the group consisting of cases of acute monocytic leukemia, we found that CD163 expression was overrepresented: 3 of the 7 cases found to be positive overall were in this group. In addition, of the remaining 4 cases positive for CD163, 2 were associated with chronic myelomonocytic leukemia or acute myeloid leukemia arising in a patient with a history of chronic myelomonocytic leukemia (cases 23 and 32). This finding further confirms the relative specificity of CD163 expression for monocytic differentiation.

View this table:
Table 2

It is interesting that CD34 was expressed in only 2 (6%) of 31 cases tested. CD117 was also not expressed often and was present in 3 (10%) of 30 cases. Approximately half of the cases were tested with CD117 (standard) and CD34 (standard), while the remainder were tested with CD117 (bone marrow) and CD34 (bone marrow). For both antibodies, a similar number tested positively with either protocol and constituted a minority of cases (overall 7%–10%), regardless of whether the standard or bone marrow protocol was applied, with CD34 expressed by 1 (6%) of 16 and 1 (7%) of 15 cases and CD117 expressed by 1 (7%) of 14 and 2 (13%) of 16 cases by the standard and bone marrow protocols, respectively.

To further examine the relationship between the bone marrow immunophenotype and that of the blasts of myeloid LC, we compared the flow cytometric data with data obtained from immunohistochemical analysis of biopsy specimens of myeloid LC Table 5. In all cases, the immunohistochemical profile of the skin blasts was discordant from the bone marrow flow cytometric antigen expression profile, even in temporally concurrent studies with similar blast cell morphologic features. For example, 8 of 13 cases were found to express CD34 by flow cytometry, but only 1 matched case expressed CD34 by immunohistochemical analysis. Similarly, 8 cases expressed CD117 by flow cytometry, and only 2 matched cases expressed CD117 by immunohistochemical analysis. Ten cases expressed MPO by flow cytometry, and only 6 matched cases showed expression by immunohistochemical analysis. Six cases expressed CD56 by flow cytometry, and 3 matched cases expressed CD56 by immunohistochemical analysis. In addition, several cases showed gain of expression of different antigens when the flow cytometric analysis of the bone marrow was compared with antigen expression in skin. In 1 case, we found that the skin blasts expressed CD56, even though this was not documented in the flow cytometric data analysis (case 33). Similarly, in 1 case, MPO was expressed by the skin blasts but was not expressed when the bone marrow was analyzed by flow cytometry (case 12).


While myeloid LC is a relatively rare manifestation of acute leukemia, its development can pose difficult diagnostic problems. In most cases, the patients are known to have systemic leukemia with circulating blasts. In these cases, we have found it expedient to correlate flow cytometric information of systemic acute leukemia (obtained from bone marrow and/or peripheral blood studies) to support the diagnosis of myeloid LC. In some cases, however, myeloid LC may be the first manifestation of acute leukemia, in which the bone marrow biopsy demonstrates a precursor lesion (myeloproliferative or myelodysplastic syndrome) or normal findings. In these cases, definitive diagnosis and/or classification of the disorder using current criteria requires immunohistochemical analysis of the skin lesions and correlation with cytogenetic information.

We also confirmed that immunophenotypic markers normally used to define blasts in the bone marrow (ie, CD34 and/or CD117) could not be used in the setting of skin lesions. These findings are similar to those previously reported in the literature.10,13 In our study, only a fraction of cases of myeloid LC expressed CD34 or CD117, even when temporally coincident flow cytometric analysis of the bone marrow showed that bone marrow blasts expressed these antigens (Table 5). Several possibilities are hypothesized to explain this discordance.

One argument may be that acute leukemias that involve the skin are often of monocytic lineage, and these usually lack expression of CD34.13 While acute monocytic leukemias are overrepresented in our study, the majority are not of monocytic lineage, and a significant number expressed CD34 on bone marrow analysis.

A second possible explanation for discordant antigen expression may be related to clinical therapy. In several cases in our study, there was interim therapy (usually systemic chemotherapy) between the bone marrow biopsy and the diagnosis of LC (15 of 33 cases; Table 3). There is, therefore, a possibility that the blasts manifesting as LC represent phenotypically distinct populations selected for by treatment or that the immunophenotype has been otherwise altered by therapy. The phenomenon of an immunophenotypically distinct blast subpopulation that persists following chemotherapy has been attributed to persistence of “minimal residual disease,” the idea being that this subpopulation may account for clinical relapse and may display a different phenotype than was characterized before initiation of therapy.1821

A third explanation may be that the blasts in myeloid LC express CD34 and/or CD117, but immunohistochemical methods are not of sufficient sensitivity for their detection. Previous studies have documented the differing sensitivities of flow cytometric and immunohistochemical methods in detecting expression of some antigens. For example, CD117 has been shown to have increased sensitivity when detected by flow cytometric vs immunohistochemical methods.2224 Other studies, however, have shown that flow cytometric and immunohistochemical studies generate reproducible immunophenotyping in acute leukemias diagnosed by bone marrow biopsy, especially with regard to CD34 expression.22,23,2527

Many of our cases expressed CD34 or CD117 in temporally concurrent bone marrow analysis but were frequently negative for these markers in the skin. In at least 1 case (case 21), immunohistochemical analysis for CD34 was performed on the bone marrow and skin lesion and showed blast expression of CD34 in the bone marrow but not in the skin. This is not necessarily linked to titering differences, as titers directed specifically toward blast expression failed to demonstrate an increase in staining in skin lesions. What we document for the first time in our studies, therefore, is relative loss of CD34 and CD117 in skin, in comparison with temporally concurrent bone marrow analysis. In these cases, these differences do not seem to be necessarily linked to therapeutic changes or differences in antigenic detection with differing methods.

In addition to CD34 and/or CD117 discrepancies, we noted discordance of expression of MPO and/or CD56 in skin lesions relative to the immunophenotype of the bone marrow. We saw a gain of MPO and/or CD56 in a minority of cases. While some of these findings may be related to therapy and/or sensitivity differences in detection methods, it is also possible that an immunophenotypically distinct subgroup of blasts occupies extramedullary sites, with gain or loss of certain antigens in the process of extramedullary involvement. Further work in this area may yield very interesting results about the pathogenesis of extramedullary involvement by acute leukemia.

Image 1

Leukemia cutis. A, Low power examination of H&E-stained sections shows a primarily interstitial infiltrate of cells separated from the overlying epidermis by a grenz zone (×100). B, Higher power examination reveals a monomorphous population of cytologically malignant cells with increased nuclear/cytoplasmic ratios and significant mitotic activity (H&E, ×400). Inset, Very high power shows nuclei with finely dispersed chromatin, multiple nucleoli, and irregular nuclear contours (×1,000 oil immersion). CF, The leukemic cells express CD43 (C, ×600), CD68 (E, ×600), and CD56 (F, ×600) but are negative for myeloperoxidase (D, ×600).

CD163 is an acute phase–regulated transmembrane protein that binds haptoglobin-hemoglobin complexes and is implicated as a hemoglobin scavenger receptor.28,29 It is thought to be relatively specific for monocytes and tissue macrophages and has been proposed to function in the innate immune response and resolution of inflammation. Preliminary studies have shown that CD163 has more specificity for histiocytes than CD68, which is an organelle-specific marker that stains lysosomes. In our study, this marker was found to be present in only a minority of cases of myeloid LC but was overrepresented in the group of cases of monocytic lineage. This finding shows that while CD163 may not be helpful in cases of unknown lineage, it may be useful in confirming monocytic differentiation in infiltrates of myeloid LC.28

Given the discrepancy between what is usually observed in the immunophenotyping of bone marrow specimens and what is seen in skin lesions, we devised an updated approach to immunophenotyping of myeloid LC that would confirm this diagnosis Figure 1.30 We found that regardless of leukemic phenotype or flow cytometric expression profile, a panel of antibodies that included MPO, CD3, CD20, CD43, CD56, CD68, and CD117 would correctly identify cases of myeloid LC in the vast majority of cases. This approach would be especially useful in the analysis of malignant cutaneous infiltrates with immature morphologic features in which a hematologic malignancy is suspected on morphologic grounds. In this context, negative CD20 and CD3 results would exclude most neoplasms of lymphocytic origin.

While MPO expression strongly supports the diagnosis of myeloid LC in the context of malignant hematopoietic skin infiltrates, we found that 58% of our cases were negative for this marker.13 Cases that were MPO– (19/33) had the CD43+/CD68+ immunophenotype. In this context, the differential diagnosis would include myeloid LC, mast cell sarcoma, and blastic plasmacytoid dendritic cell neoplasms. If the lesion is MPO– and CD68–, one could consider CD20–/CD3– lymphomas, such as a B-cell lymphoma recurring after rituximab therapy or CD30+ lymphoproliferative disorders. These considerations can be confirmed by using CD79a or CD30, respectively. Other markers that may be useful to identify difficult lymphoid neoplasms include PAX5 (B cells) and CD2 (T cells). The very rare case of acute lymphoblastic LC may be identified with these additional markers.

View this table:
Table 4

We found CD56 and CD117 to be helpful in further evaluation of MPO– cases that expressed CD43 and CD68. In CD68+/MPO–/CD56– cases (9 of 19 cases), the differential diagnosis would include mast cell sarcoma and myeloid LC. In these cases, mast cell tryptase would help distinguish between myeloid LC and mast cell sarcomas.31 In the CD68+/MPO–/CD56+ cases (10 of 19 cases), staining for CD4 and CD123 will often distinguish between myeloid LC and blastic plasmacytoid dendritic cell neoplasms.32

View this table:
Table 5
Figure 1

Proposed algorithm for the immunohistochemical diagnosis of leukemia cutis (LC). The numbers shown in brackets represent the number of cases positive or negative for each marker over the number of test cases in that arm of the algorithm.

* One case (case 22) was negative for CD43, but a myeloperoxidase (MPO) stain was positive, confirming myeloid LC. The concurrent bone marrow biopsy was diagnostic of acute promyelocytic leukemia.

To determine if we could find cases of blastic plasmacytoid dendritic cell neoplasms that had been initially classified as myeloid LC, we tested 9 of the 10 cases (cases with material available for further study) that were CD68+/MPO–/CD56+ with CD4. Only 1 case (case 5) had an overlapping immunophenotype, being MPO–/CD56+/CD4+. The corresponding bone marrow biopsy in case 5 confirmed myeloid differentiation, demonstrating acute myeloid leukemia with myelodysplasia-related changes. In addition, we noted that expression of CD4 in this case was weak, in contrast with the strong and uniform staining seen with blastic plasmacytoid dendritic cell neoplasms. However, it is important to note that differentiating between these 2 entities is not always possible using histologic findings and paraffin-based immunohistochemical stains.33 Markers more recently reported to be of use in characterizing blastic plasmacytoid dendritic cell neoplasms include BDCA-2, TCL1, CLA, and MxA, although the WHO recommends “acute leukemia of ambiguous lineage” for tumors having some but not all of the immunophenotypic characteristics of these tumors.14,3236 Frozen section staining with CD4 may also prove useful in distinguishing these entities.37 Therefore, distinguishing between myeloid LC and blastic plasmacytoid dendritic cell neoplasms confined to the skin may require extensive immunohistochemical studies, with ambiguous cases best classified as acute leukemia of ambiguous lineage. Regardless, cutaneous CD56+ hematologic neoplasms as a group are known to have a poor prognosis.3840

While an extensive workup may be indicated in patients with LC, a more abbreviated panel may be sufficient for patients who have a documented diagnosis of systemic leukemia. In these patients, it is less critical to exclude other malignancies such as nonhematologic malignancies and lymphomas because the likelihood of a second malignancy is relatively low. Expression of MPO alone or a combination of CD43 and CD68 expression in the face of MPO negativity may be sufficient to identify most cases of myeloid LC in these circumstances.

This study demonstrated that the diagnosis of myeloid LC is best made in conjunction with bone marrow findings and flow cytometry because markers thought to be specific for blasts (ie, CD34 and CD117) are not detected by immunohistochemical staining of myeloid LC in the vast majority of cases. We also propose an immunophenotypic algorithm that correctly classified all 33 cases in our series. There can be some immunophenotypic overlap between myeloid LC and blastic plasmacytoid dendritic cell neoplasms, which may be rectified by examining the findings in the corresponding bone marrow. We were surprised to find such disparate immunophenotypic profiles between the skin infiltrates and the bone marrow lesions, with loss and gain of antigens documented. Whether this represents a true alteration in the immunophenotype of the blasts present in myeloid LC (due to prior chemotherapy or other factors) or simply sensitivity differences between flow cytometric and immunohistochemical antigen detection has yet to be determined. Because of these differences, caution should be used when drawing diagnostic conclusions based on the comparison of these profiles.


We thank Amarjeet Grewall for laboratory assistance and Anet James for assistance with the graphics.


  • Supported by the Stanford Department of Pathology.


  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.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
View Abstract