The presence of myxoid stroma in malignant uterine

ORIGINAL ARTICLE Myxoid Leiomyosarcoma of the Uterus A Clinicopathologic Analysis of 30 Cases and Review of the Literature With Reappraisal of Its Distinction From Other Uterine Myxoid Mesenchymal Neoplasms Carlos Parra-Herran, MD,* John K. Schoolmeester, MD,w Liping Yuan, PhD,z Paola Dal Cin, PhD,z Christopher D.M. Fletcher, MD, FRCPath,z Bradley J. Quade, MD, PhD,z and Marisa R. Nucci, MDz Abstract: Myxoid leiomyosarcoma (mlms) of the uterus is a rare neoplasm; thus, knowledge of its clinical behavior and morphology remains limited. We therefore reviewed 30 cases initially diagnosed as uterine mlms to better characterize its clinicopathologic features. Diagnosis was confirmed in 25 subjects (average age 51.5 y), of whom 80% were stage 1 at presentation. The average tumor size was 10.8 cm. An infiltrative tumor border was observed microscopically in 24 cases (96%); the border in 1 case could not be assessed. Fourteen cases (56%) had >10 mitoses per 10 high-power fields, 8 (32%) had between 2 and 10, and 3 cases (12%) had <2 mitoses. Geographic tumor necrosis and moderate to severe nuclear pleomorphism were seen concurrently in 12 cases (48%). All tumors expressed smooth muscle markers. Estrogen receptor was expressed in 29.4% of cases. Eighteen of 21 cases (85.7%) were negative for ALK by immunohistochemistry. Follow-up information was available in 18 subjects: 8 died of disease, 6 were alive with local and/or distant recurrence, and 4 were alive with no evidence of disease. Individuals who died of disease tended to have tumors with >10 mitoses per 10 high-power fields. Among cases with Z5-year follow-up, overall survival was 11.1%, significantly worse compared with reported survival rates for conventional LMS. The initial diagnosis of mlms was revised in 5 cases. Four had a distinctive loose myxoid appearance, nuclei with vesicular chromatin, and ALK positivity by immunohistochemistry, suggesting, suggesting inflammatory myofibroblastic tumor. This diagnosis was confirmed by in situ hybridization in 2 cases. One additional myxoid tumor lacked smooth muscle or myofibroblastic features and could not be classified further. mlms is an aggressive neoplasm characterized by infiltrative tumor borders From the *Department of Pathology, University of Ottawa and The Ottawa Hospital, Ottawa, ON, Canada; wmayo Clinic, Rochester, MN; and zbrigham and Women s Hospital, Harvard Medical School, Boston, MA. B.J.Q. and M.R.N. contributed equally as senior authors. Conflicts of Interest and Source of Funding: The authors have disclosed that they have no significant relationships with, or financial interest in, any commercial companies pertaining to this article. Correspondence: Marisa R. Nucci, MD, Brigham and Women s Hospital, Harvard Medical School, 75 Francis St., Boston, MA 02115 (e-mail: mnucci@partners.org). and variability of other features (mitotic count, atypia, and necrosis). The differential diagnosis includes myxoid leiomyoma and inflammatory myofibroblastic tumor. Attention to distinguishing morphologic features and immunohistochemistry will aid in the interpretation. An illustrated algorithm with criteria for diagnosis is proposed. Key Words: uterine leiomyosarcoma, myxoid leiomyosarcoma, inflammatory myofibroblastic tumor, myxoid leiomyoma, myxoid smooth muscle tumor of unknown malignant potential (Am J Surg Pathol 2016;40:285 301) The presence of myxoid stroma in malignant uterine smooth muscle tumors is a rare but well-recognized phenomenon. When abundant, the myxoid stroma separates smooth muscle fibers, giving a distinct appearance and warranting the diagnosis of myxoid leiomyosarcoma (mlms). Characterization of uterine mlms as a distinct neoplasm, separate from its conventional (spindle cell) counterpart, may have important diagnostic and prognostic implications. Although data thus far are limited, published studies suggest that mlmss are more aggressive, having a worse 5-year overall and disease-free survival rate compared with conventional LMSs. 1,2 In addition, established diagnostic features of malignancy applied to conventional LMSs namely, tumor cell necrosis, severe nuclear pleomorphism, and mitotic activity exceeding 10 mitoses per 10 high-power fields (HPFs) may not be present, which can be a source of diagnostic difficulty. In particular, mitotic activity has been reported to be low in most mlmss because of wide separation of the neoplastic cells by myxoid matrix. As a result, a mitotic count as little as 2 per 10 HPF is regarded as indicative of malignancy because some tumors with low mitotic counts have resulted in adverse outcome (recurrence, metastases, or fatality due to disease). 3 5 mlms of the uterus is rare; a total of 69 individuals with this disease have been reported in the English literature to date. 2 4,6 34 Thus, much of the histopathology and biologic behavior of this entity is yet to be fully described. Therefore, in this study we describe the clinical Am J Surg Pathol Volume 40, Number 3, March 2016 www.ajsp.com 285

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016 and pathologic characteristics of 30 additional subjects initially diagnosed with uterine mlms on the basis of established criteria. 5,35 We also discuss the usefulness of established morphologic features and supplementary studies in determining malignancy and separating mlms from its mimics. MATERIALS AND METHODS This study was approved by the Human Research Committee (the local Institutional Research Board) at Brigham and Women s Hospital. Case Selection and Review We reviewed cases seen primarily or in consultation at Brigham and Women s Hospital with an initial diagnosis of mlms of the uterus. Available clinical information was recorded, including subject age, clinical presentation, and follow-up data. Macroscopic findings were obtained from surgical pathology reports. Histologic material in each case (hematoxylin-eosin and immunohistochemistry slides) was reviewed by 4 gynecologic pathologists (C.P.-H., J.K.S., B.J.Q., and M.R.N.). Diagnosis of mlms was confirmed if any of the following diagnostic criteria outlined in the current literature were present 5,36 : (1) Nuclear atypia, defined as the presence of hyperchromasia and nuclear enlargement. Degree of nuclear enlargement was recorded as 1+ (mild; nucleus similar in size to that of adjacent nonneoplastic myometrium), 2+ (moderate; nucleus 2 the size of adjacent non-neoplastic myometrium), and 3+ (severe; nucleus Z3 the size of adjacent nonneoplastic myometrium). Presence of nuclear hyperchromasia was determined by comparison with nonneoplastic myometrial nuclei. (2) Mitotic activity of 2 mitoses per 10 high-power microscopic fields or more. Mitoses were counted in at least 50 HPF in each case (magnification field, 400; field diameter, 0.62 mm). A mitotic figure was defined as a nuclear structure with well-formed chromatids appreciable at HPF (either clumped or progressively separated depending on the phase of the division) in a cell without a visible nuclear membrane and without hypereosinophilic cytoplasm. Highest count per 10 HPF and atypical forms, if present, were recorded. (3) Coagulative geographic tumor cell necrosis. Cellularity of the tumor was assessed in comparison with adjacent non-neoplastic myometrium and was determined to be hypercellular or hypocellular. Other histopathologic features were assessed, including tumor borders and percentage of myxoid component. Alcian blue histochemical staining was reviewed if part of archival material, or staining was performed if formalinfixed, paraffin-embedded tissue was available; the staining was given a score depending on its intensity and distribution: negative (absent) or positive (1+ for weak intensity/focal distribution, 2+ for moderate intensity/ multifocal distribution, and 3+ for strong intensity/ diffuse distribution). Immunohistochemistry and In Situ Hybridization Studies Archival immunohistochemical staining was reviewed in each case. Additional immunohistochemical analysis was performed in cases with available tissue material. The panel chosen for review included the following antigens: anaplastic lymphoma kinase 1 (ALK; clone 3633, dilution 1:300, citrate buffer pressure cooked; Cell Signaling, Danvers, MA), smooth muscle actin (clone 1A4, dilution 1:20,000, no antigen treatment; Sigma, St. Louis, MO), desmin (clone DEU-10, dilution 1:5000, citrate buffer pressure cooked; Sigma), h-caldesmon (clone h-cd, dilution 1:300, citrate buffer pressure cooked; Dako, Carpinteria, CA), estrogen receptor (clone SP1, cat.#rm-9101-s, dilution 1:40, citrate buffer pressure cooked; Thermo Scientific, Waltham, MA), CD10 (clone VP-c328, dilution 1:20, citrate buffer pressure cooked; Vector Lab, Burlingame, CA), Ki67 (MIB-1, cat.#m7246, dilution 1:200, citrate buffer pressure cooked; Dako), and ROS1 (clone D4D6, cat.#3287, dilution 1:500, EDTA pressure cooked; Cell Signaling). Staining was interpreted in terms of intensity and distribution as negative (absent staining) or positive (1+ for weak/focal, 2+ for moderate/multifocal, and 3+ for strong/diffuse). Fluorescence in situ hybridization (FISH) analysis was performed on 5-mm tissue sections from formalinfixed paraffin-embedded specimens. For the detection of ALK rearrangement, we used the LSI ALK (Abbott Molecular, Des Plaines, IL) dual-color, break-apart rearrangement probe mixture, which contains 2 different labeled probes (Spectrum Orange and Spectrum Green) on opposite sides of the breakpoint of the ALK gene, located at 2p23. Probes and nuclei were codenatured simultaneously, followed by hybridization and washing, according to the manufacturer s directions (Abbott Molecular). At least 50 cells with strong, discrete signals were analyzed in each case. An intact ALK locus appears as 2 colocalized orange and green signals or as a single yellow signal when both probes overlap. If ALK rearrangement is present, the orange probe appears distinct from the green probe (seen as separate signals with loss of the expected yellow combined signal). Statistical Analysis Statistical inferential analysis was performed on cumulative data using the Statistics Online Computational Resource from University of California (Los Angeles, CA), and GraphPad Prism 6 for Windows (La Jolla, CA) to identify associations between recorded variables and outcome at follow-up. Numerical variables (such as subject age, tumor size, immunohistochemistry intensity scores) were compared using the Wilcoxon Mann-Whitney test. Categorical variables (such as presence of tumor cell necrosis and lymphovascular space invasion) were compared using the Fisher exact test. 286 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus One-sided and 2-sided P-values were calculated for each variable. RESULTS Thirty-three cases diagnosed as mlms of the uterus were identified in our search. Three cases did not have histologic material available for review and were excluded. The remaining 30 subjects constituted the cohort of this study (27 consultation and 3 in-house cases). Age distribution, stage at time of initial diagnosis, and outcome data are presented in Table 1, pathologic characteristics in Table 2, and immunohistochemical and in situ hybridization results in Table 3. Morphologic Features of Confirmed Cases of Uterine mlmss After secondary review of available material and ancillary studies, the diagnosis of mlms was confirmed in 25 cases (83%, cases 1 to 25, Figs. 1 3). The average age was 51.5 (range, 33 to 77) years. Twenty subjects (80%) had a pathologic tumor stage I (lesion confined to the uterus) at the time of presentation. Conversely, 5 subjects (20%) presented with extrauterine disease (abdominal and pelvic cavities). Tumor size was available in 21 cases and ranged from 3 to 33 cm (mean, 10.8). The gross appearance was available in 13 cases (52%, Fig. 1). The tumor was commonly described as gelatinous, mucoid, or myxoid and tan to yellow, or gray (10/13, 77%). Three tumors were grossly hemorrhagic: 1 was partly calcified, 1 was described as firm, and 1 was rubbery. One tumor was described as multilobulated and 1 as having satellite nodules surrounding the main mass. All tumors in which the tumor border could be assessed had an infiltrative growth pattern on microscopic examination (24 cases, 96%). In the remaining case, the tumor border could not be assessed. Different patterns of infiltration into the surrounding myometrium and endometrium were seen: 19 cases showed a markedly irregular (geographic) tumor interface with angulated and branching borders, 2 invaded the myometrium in the form of discontinuous, finger-like projections, and 3 had focal areas of destructive infiltration at the tumormyometrial interface. On low-power microscopic examination, mlmss were either uniformly hypercellular (16/25, 64%) or variably cellular (7/25, 28%). Tumors within the latter category contained hypercellular areas that merged with those that were distinctly hypocellular; the hypocellular foci typically contained mucin pools or had a myxoid matrix that imparted a lacy or reticular appearance. In only 2 cases (8%) was the tumor uniformly hypocellular when compared with the adjacent myometrium. Tumors were consistently composed of spindle cells arranged in a myxoid extracellular matrix; in 2 cases, the stroma had a myxohyaline appearance. Myxoid stroma represented TABLE 1. Patient Age, Extrauterine Extent at Presentation and Follow-up Data Case No. Age Extrauterine Extent* Follow-up Period Status/Site of Recurrence 1 62 No 10 mo Dead of disease 2 45 No 1 y, 3 mo Dead of disease 3 45 No 2 y Alive with disease/lungs 4 52 No 2 y Dead of disease 5 48 No 2 y, 10 mo Alive with disease/abdomen, pelvis 6 35 No 4 y Alive with disease/pelvis 7 73 No 3 y Dead of disease 8 58 Abdomen, pelvis 4 y Dead of disease 9 74 No 4 y Alive with disease/lungs, liver 10 47 No 3 y Alive with disease/abdomen, pelvis 11 39 No 2 y, 6 mo Alive free of disease 12 33 No Not known Not available 13 52 No Not known Not available 14 39 No 12 y Alive free of disease 15 46 Omentum Not known Not available 16 45 No 2 y Alive free of disease 17 42 Abdomen, pelvis Not known Not available 18 49 Pelvis Not known Not available 19 77 No 1 y, 6 mo Alive free of disease 20 60 No 8 mo Dead of disease 21 51 No Not known Not available 22 62 No Not known Not available 23 45 No 3 y Dead of disease 24 56 Abdomen, ovaries, tubes 3 y, 2 mo Dead of disease 25 53 No 1 y, 1 mo Alive with disease/ribs, wrist 26 53 No 2 y, 3 mo Dead of disease 27 54 No Not known Not available 28 56 No Not known Not available 29 54 No 4 y Dead of disease 30 39 No 4 y Alive with disease/pelvis *At the time of initial diagnosis. www.ajsp.com 287

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016 TABLE 2. Macroscopic and Histopathologic Characteristics of Myxoid Mesenchymal Tumors of the Uterus Case No. Pattern Size (cm) Tumor- Myometrial Interface Nuclear Enlargement Mitoses (in 10 HPF) Atypical Mitoses 1 LMS like 7.4 Infiltrative Y 3+ Y 21 Y N Yes No 2 LMS like 9 Infiltrative Y 1+ Y 20 N N No Yes 3 LMS like 7.5 Infiltrative Y 2+ Y 19 N N No Yes 4 LMS like 3 Infiltrative Y 2+ Y 6 N Y No No 5 LMS like 16 Infiltrative Y 3+ Y 25 Y Y Yes No 6 LMS like 20 Infiltrative N 2+ Y 1 N Y No No 7 LMS like 5.5 Infiltrative Y 3+ Y 28 Y N No No 8 LMS like 8 Infiltrative Y 1+ Y 12 N N No No 9 LMS like 14 Infiltrative Y* 1+ Y 2 N N No No 10 LMS like 33 Infiltrative Y* 2+ Focally Y 1 N N Yes Yes 11 LMS like 9 Infiltrative Y* 1+ Y 4 N Y No Yes 12 LMS like 6.5 Infiltrative Y* 1+ Y 2 N N No Yes 13 LMS like 6 Infiltrative Y 1+ Y 16 N Y Yes No 14 LMS like 8.5 Infiltrative Y* 3+ Y 1 N Y Yes No 15 LMS like ND Infiltrative Y* 3+ Y 20 Y N Yes No 16 LMS like 10 Infiltrative Y* 1+ Y 5 N N No No 17 LMS like 14 Unknown Y 1+ Y 2 N Y No No 18 LMS like ND Infiltrative N 1+ Y 15 N Y Yes No 19 LMS like 5.8 Infiltrative Y 2+ Focally Y 30 N N No Yes 20 LMS like 12 Infiltrative Y 1+ Y 30 N N Yes Yes 21 LMS like ND Infiltrative Y 1+ Y 2 N N No No 22 LMS like 12 Infiltrative Y 1+ Y 4 N Y Yes No 23 LMS like 4.5 Infiltrative Y 3+ Y 25 Y Y Yes Yes 24 LMS like 15 Infiltrative Y 1+ Y 18 Y N Yes No 25 LMS like ND Infiltrative Y 3+ Y 14 Y N Yes Yes Mean Infiltrative Hypercellular 1+, 2+ = 72% Hyperchromasia Mean 12.9 Yes 28% Yes 40% Yes 48% Yes 40% 10.79 96% 92% 100% Median 9 Smooth 4% Hypocellular 8% 3+ = 28% Median 14 No 72% No 60% No 52% No 60% 26 IMT like 9 Infiltrative Y 1+ N 10 N Y No Yes 27 IMT like 14 Infiltrative Y 1+ N 0 N Y No No 28 IMT like 10 Infiltrative Y 2+ Y 8 N Y No No 29 IMT like 3.5 Infiltrative Y 2+ N 6 N Y No No 30 Myxomatous 1.5 Infiltrative N 1+ Y 1 N N No No *Cellularity was variable, predominantly hypercellular in comparison with non-neoplastic myometrium. CI indicates chronic inflammation; IMT, inflammatory myofibroblastic tumor; LVI, lymphovascular space invasion; ND, not determined. CI Hypercellularity Hyperchromasia Necrosis LVI >50% of the overall tumor volume in all cases, and Alcian blue stain was consistently positive with moderate to strong intensity (average score, 2.45). Neoplastic cells were typically arranged in bundles and fascicles; more infrequently, cells were individually dispersed within the myxoid matrix. Neoplastic cells consistently had typical smooth muscle morphology with eosinophilic cytoplasm and elongated (cigar shaped to tapered) nuclei. The nuclear size was scored as 1+ (comparable to adjacent non-neoplastic myometrium) in 13 of 25 (48%) cases, as 2+ (nucleus 2 the size of adjacent non-neoplastic myometrium) in 5 of 25 (20%) cases, and as 3+ (nucleusz3 the size of adjacent non-neoplastic myometrium) in 7 of 25 (28%) cases. Of note, in 3 cases designated as 2+, the nuclear enlargement was a focal finding. Tumor cell nuclei were hyperchromatic in all cases. The mitotic rate ranged from 1 to 30 per 10 HPF (average, 12.9). Three cases (12%) had only 1 mitosis, 8 (32%) had between 2 and 10 mitoses, and 14 (56%) had >10 mitoses per 10 HPF. Geographic tumor cell necrosis was observed in 12 (48%), and lymphovascular space invasion was present in 9 (36%). Ten cases (40%) contained variable numbers of inflammatory cells within the tumor. Nine had a chronic inflammatory infiltrate composed of lymphocytes and occasionally plasma cells dispersed throughout the tumor; in 2 of these, both of which also had necrosis, polymorphonuclear leukocytes were present. One case showed only perivascular lymphocytes. Ancillary Studies in Uterine mlms All histologically confirmed cases of mlms were positive for at least 1 smooth muscle marker (smooth muscle actin, desmin, or h-caldesmon, Fig. 2). Nine (36%) expressed all 3 markers, and 12 (48%) expressed 2. Smooth muscle actin was the most sensitive stain, being detected in all 24 cases tested with an average intensity score of 2.08. Desmin was positive in 21 cases with a score 288 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus TABLE 3. Immunohistochemical and FISH Results for Uterine mlmss (Cases 1 to 25) and Reclassified Uterine Myxoid Lesions (Cases 26 to 30) Case No. Alcian Blue SMA DESMIN h-cald ALK ROS1 CD10 ER (%) Ki67 (%) ALK FISH 1 1+ 3+ 3+ 1+ 0 0 3+ 0 90 2 2+ 1+ 0 1+ 0 0 3+ 0 40 3 3+ 1+ 3+ 3+ 0 0 3+ 90 20 4 3+ 3+ 3+ 2+ 0 0 2+ ND 40 5 3+ 1+ 3+ 0 ND ND 2+ ND 60 6 3+ 1+ 0 0 0 0 0 0 0 7 3+ 3+ 2+ 1+ 0 0 3+ 0 0 8 ND 2+ 1+ 0 ND ND ND ND ND 9 ND 3+ 1+ 0 0 0 ND ND ND 10 ND 3+ 3+ 2+ 0 0 ND 50 ND 11 3+ 1+ 2+ 0 0 0 0 10 0 12 3+ 3+ 2+ 1+ 0 0 0 ND 0 13 2+ ND 3+ 0 ND ND 3+ ND ND 14 3+ 1+ 0 0 0 0 0 0 0 15 3+ 2+ 2+ 0 0 0 3+ 0 0 16 2+ 2+ 3+ 3+ 0 0 3+ 70 0 17 ND 3+ 3+ 3+ 0 0 ND 80 0 18 2+ 1+ 0 0 0 0 0 0 0 19 1+ 3+ 3+ 1+ 0 0 ND 0 20 20 3+ 3+ 2+ 0 0 0 3+ 0 0 21 ND 3+ 3+ 0 ND ND ND ND ND 22 3+ 2+ 3+ 0 0 0 ND ND ND 23 2+ 1+ 1+ 0 1+ 0 1+ 0 0 Negative 24 1+ 3+ 1+ 0 3+ 0 0 0 0 Negative 25 3+ 1+ 1+ 0 3+ 0 1+ 0 5 ND Average 2.5 2.1 1.9 0.72 0.33 0 1.7 17.6 14.5 26 3+ 0 0 0 1+ 0 3+ ND 50 ND 27 ND 3+ 3+ 0 2+ 0 1+ 75 ND Negative 28 3+ 3+ 3+ 0 3+ 0 0 20 0 Positive* 29 2+ 1+ 3+ 0 3+ 0 0 0 0 Positivew 30 3+ 0 0 0 0 0 0 20 ND *Positive for 5 0 centrometric probe (unbalanced rearrangement). wpositive for classic ALK1 rearrangement. ER indicates estrogen receptor; H-CALD, h-caldesmon; ND, not done; SMA, smooth muscle actin. of 1.92. Conversely, h-caldesmon was less frequently expressed (10 cases; average score, 0.72). Immunohistochemical analysis for ALK was performed in 21 cases. Most (18, 85.7%) were negative for this marker; however, 3 were positive (cases 23 to 25, Fig. 3). FISH was performed in cases 23 and 24, and both were negative for ALK rearrangement. Case 25 did not have material available for further testing. All 21 cases tested for ROS1 expression by immunohistochemistry were negative. Twelve of 18 cases were CD10 positive (66.7%). Estrogen receptor was expressed in only 5 of 17 cases tested (29.4%). Finally, data on Ki67 immunohistochemistry were available in 19 cases. Tumors tended to have absent to low proliferation indexes; the percentage of positive cells ranged from 0% (12 cases) to 90%, with an average of 14.5%. Follow-up Data Follow-up information was available on 18 subjects (Table 1). The follow-up period ranged from 8 to 144 (average, 35.2; median, 32) months. At follow-up, 8 subjects (44.5%) died of disease, 6 (33.3%) were alive with recurrent or metastatic disease, and 4 (22.2%) were alive with no evidence of disease. When assessing cases with a follow-up period of 5 years or more, 8 of 9 subjects died of disease, for an overall 5-year survival rate of 11.1%. Table 4 shows the relationship between clinical and pathologic variables and clinical outcome, and Figure 4 displays the stratification of the study cases according to histologic features of malignancy. There were no statistically significant associations with overall survival when corrected for multiple comparisons. Of note, in our series, the mitotic rate was higher in subjects who had died of disease: all who died had tumors with a mitotic rate >5 mitoses per 10 HPF (cases 1, 2, 4, 7, 8, 20, 23, and 24). In contrast, 6 of 10 subjects alive at follow-up had tumors with mitotic rates of <5 per 10 HPF (cases 3, 5, 6, 9, 10, 11, 14, 16, 19, and 25). Nonetheless, 3 subjects with tumors with mitotic rates of r2 mitoses per 10 HPF (cases 6, 9, and 10) had local or distant recurrence (2 with pelvic and 1 with lung/liver recurrence). Morphologic Characterization of Cases With Revised Diagnosis After a review of available histologic material and immunohistochemistry and in situ hybridization results, www.ajsp.com 289

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016 FIGURE 1. Macroscopic appearance of mlms (A, B; case 11). Notice the yellow and gelatinous cut surface, distinct from an adjacent conventional leiomyoma (B, right lower aspect). Tumor interface is focally irregular. Histologic appearance of mlms (C, D; case 12). The neoplasm has an infiltrative interface with finger-like projections into the surrounding myometrium. Cellularity ranges from low to markedly increased depending on the amount of myxoid matrix (C, hematoxylin and eosin; D, hematoxylin and eosin). Images A and B courtesy of Dr. R. Tucker Burks, M.D., Carolinas Pathology Group, Charlotte, NC. the diagnosis of mlms could not be confirmed and was revised in 5 cases (cases 26 to 30). Cases 26 to 29 had a similar morphologic appearance. All were hypercellular in comparison with nonneoplastic myometrium and had a predominant loose myxoid component with a minor fascicular smooth muscle like pattern (Fig. 5). The former was composed of plump spindle cells with oval nuclei and vesicular chromatin dispersed in a prominent myxoid matrix, giving a fasciitis or tissue culture like appearance. Only 1 case contained cells with enlarged, epithelioid cells with nuclei that ranged from hyperchromatic to pale and vesicular (case 28). An inflammatory infiltrate composed of lymphocytes and plasma cells was observed in variable amounts in all tumors. All 4 cases had infiltrative borders. Two cases each had 1+ and 2+ nuclear enlargement. Three had mitotic activity >2 mitoses per 10 HPF; 1 had no mitotic activity. None had coagulative tumor cell necrosis or atypical mitotic figures. Case 26 lacked smooth muscle marker expression; in the other 3 cases, smooth muscle actin and desmin were positive. None of these cases expressed h-caldesmon. On the contrary, ALK expression by immunohistochemistry was observed in all 4 cases. Given the highly suggestive histomorphology and presence of ALK staining, the diagnosis of inflammatory myofibroblastic tumor was rendered. Three cases had material available for FISH studies. FISH was negative in case 27, positive for classic ALK gene rearrangement in case 29, and positive for an unbalanced translocation in case 28 (positive for 5 0 centromeric probe alteration only). Follow-up information was available for cases 26 and 29: both died of disease 2 and 4 years after initial diagnosis, respectively. Lastly, case 30 showed a third histologic pattern characterized by spindle cells with scant eosinophilic cytoplasm 290 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus FIGURE 2. mlms (case 3). The tumor has irregular infiltrative borders (A) cells show moderate (2+) nuclear pleomorphism, hyperchromasia, and conspicuous mitotic activity (B). Cells are separated by pale myxoid extracellular matrix, highlighted by Alcian blue stain (C). Tumor cells are strongly positive for Desmin (D) and h-caldesmon (E) and negative for ALK (F) by immunohistochemistry (A, hematoxylin and eosin; B, hematoxylin and eosin; C, Alcian blue; D, Desmin; E, h-caldesmon; F, ALK). www.ajsp.com 291

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016 FIGURE 3. Submucosal mlms (Case 24). The tumor interface is infiltrative with angulated irregular projections into the adjacent endometrium and myometrium (A). The neoplasm shows fascicular growth and is composed of bundles of hyperchromatic spindle cells with tapered ends, surrounded by a prominent myxoid matrix (B). This tumor showed strong ALK expression (C) and was negative for h-caldesmon (D) by immunohistochemistry. FISH for ALK gene rearrangements was negative (A, hematoxylin and eosin; B, hematoxylin and eosin; C, ALK; D, h-caldesmon). and elongated hyperchromatic wavy nuclei individually dispersed in an abundant myxoid matrix (myxoma-like pattern, Fig. 6). The tumor cells contained nuclei that were not enlarged in comparison with non-neoplastic myometrium, and there was no mitotic activity. Tumor borders were infiltrative; however, coagulative necrosis or lymphovascular space invasion were not observed. Tumor was positive for estrogen receptor; all smooth muscle markers and ALK were negative. On the basis of the morphologic and immunohistochemical features, a diagnosis of myxoid spindle cell neoplasm of uncertain malignant potential was rendered. The subject was alive with recurrent disease in the pelvis 4 years after hysterectomy. DISCUSSION Reflecting the rarity of uterine mlms, English literature reports account for only 69 documented cases in aggregate (Table 5). 2 4,6 34 To the best of our knowledge, our study represents the largest single study of uterine mlms to date. Our study confirms that mlms can occur over a wide range of ages, but most commonly occurs in postmenopausal women, similar to conventional uterine LMS. In addition, a distinctive mucoid or gelatinous gross appearance is common and usually is the first indication of the diagnosis. Size is variable, but tumors are typically large (mean, 10.8 cm). Since its original description in 1986, a low mitotic rate has been recognized as an important and potentially challenging feature of uterine mlms. 3,4,22 A mitotic rate as low as 2 per 10 HPF has been associated with aggressive behavior and is therefore considered a criterion for malignancy. 5 In our study, the mitotic rate showed wide variation. Interestingly, over half of our subjects 292 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus TABLE 4. Correlation Between Clinical and Pathologic Characteristics and Survival at Follow-up Variables AFD+AWD DOD P Patient age (y)* 50.2 56.4 0.18 Stage I at presentationw 10/10 6/8 0.18 Tumor size (cm)* 13.7 8.05 0.07 Lymphovascular space invasionw 5/10 3/8 0.66 Nuclear enlargement (score)* 2 1.8 0.13 Tumor cell necrosisw 4/10 4/8 1.00 Mitoses (number per 10 HPF)* 10.2 20 0.05 No. positive smooth muscle markers* 2.2 2.37 0.72 Smooth muscle actin (intensity score)* 1.7 2.37 0.18 Desmin (intensity score)* 1.9 1.62 0.59 h-caldesmon (intensity score)* 0.9 0.62 0.92 Ki67 (percentage of positive cells)* 13.12 24.2 0.81 *Analyzed as continuous variable: average for each survival group is presented. Two-sided P-value calculated using the Wilcoxon Mann-Whitney test. P-value corrected for multiple comparisons (corrected probability threshold P = 0.005). wanalyzed as categorical variable: proportion of cases with the variable is presented. Two-sided P-value calculated using the Fisher exact test. AFD indicates alive free of disease; AWD, alive with disease; DOD, dead of disease. (14/25, 56%) had tumors with mitotic rates of >10 per 10 HPF. Seven cases (28%) had a rate between 2 and 10 (cases 9, 11, 12, 16, 17, 21, and 22). Three of these subjects had follow-up information. One individual (case 9) was alive with disease in the lungs and liver; this individual s tumor had 2 mitoses per 10 HPF. The other 2 subjects (cases 11 and 16) were alive without evidence of disease at last follow-up (2 y, 6 mo; 2 y); these cases had 4 and 5 mitoses per 10 HPF, respectively. Thus, we confirm that use of a mitotic threshold of 2 mitoses/10 HPF, which is different from the threshold of 10 per 10 HPF used for conventional LMS, is associated with aggressive behavior. Nonetheless, tumor recurrence also was observed in 2 cases (which had follow-up information) with a mitotic count below this threshold (mitotic rate of 1 per 10 HPF; cases 6 and 10). Therefore, low or even absent mitotic activity does not exclude adverse outcome by itself and must be considered in light of other morphologic features. Moderate to severe (2+ and 3+) nuclear enlargement and geographic tumor cell necrosis were seen with the same frequency in our series (48%). When present, these features are helpful in diagnosing malignancy. However, assessment of tumor necrosis can be difficult and suffers from considerable interobserver variability. 39 In addition, nuclear enlargement and atypia can be a focal finding only, as evidenced in 2 of our cases. An infiltrative tumor interface, in contrast, was a more consistent feature of malignancy in our series. Thorough evaluation of these characteristics requires careful examination that often includes extensive sampling. Immunohistochemical Studies in Uterine mlms All cases with a histologic appearance diagnostic of mlms expressed at least 1 smooth muscle marker in our series. As expected, smooth muscle actin and desmin were the most sensitive. Conversely, h-caldesmon was seen only in 40% of cases tested, and its expression was weak. Overall, although variation was observed in our study, expression of a smooth muscle phenotype appears to be consistent in mlmss, and its absence should raise the possibility of an alternative diagnosis. Interestingly, CD10 was frequently observed (66.7% of cases tested), which is in keeping with the reported low specificity of CD10 in the diagnosis of mesenchymal uterine lesions. 40 42 This limitation should be considered when the differential diagnosis includes endometrial stromal tumors. Estrogen receptor was seen only in a minority of subjects (29.4% of cases tested), which contrasts with expression rates for conventional LMS (of up to 87%). 43 45 Estrogen receptor immunohistochemistry does not appear to have diagnostic utility, because leiomyomas and other myxoid lesions of the uterus can also express this marker. However, it is prudent to consider testing for hormone receptors for therapeutic purposes. Lastly, Ki67 was low in our series: more than half of the cases tested were completely negative, and the average index was 14.5%. FIGURE 4. Distribution of uterine mlms cases by mitotic count in 10 HPF, presence of geographic tumor cell necrosis, and presence of significant nuclear atypia (2+ or 3+ nuclear enlargement and hyperchromasia). DOD indicates dead of disease; AWD, alive with disease; AFD, alive free of disease; NA, follow-up information not available. www.ajsp.com 293

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016. 294 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus Biologic Behavior of mlms Our study shows that mlms portends a poor prognosis, in keeping with previous reports. 1,2,46 Demise due to disease was documented in almost half of our cases with follow-up information, and an additional third of the cohort developed local or distant tumor recurrences. Five-year survival in our study was 11.1%, significantly lower than the survival rates reported for conventional LMS. We recognize that the number of cases with a follow-up period of at least 5 years is small in our study; thus, more experience and documentation in the literature is needed. Interestingly, adverse outcome was observed despite the rather subtle morphologic features observed in some cases. A fatal outcome has been reported in tumors with low mitotic rates in several references (Table 5). Of note, all patients who died in our study cohort had tumors with a brisk mitotic rate (7 cases had >10 mitoses, and 1 had 6 mitoses per 10 HPF), which suggests a potentially underappreciated association between high mitotic rate and prognosis. Combining our data with the published English-language literature, 14 of 21 patients (67%) who had died from disease had tumors with a mitotic rate of Z5 mitoses per 10 HPF (range, 5 to 30; mean, 18 mitoses/ 10 HPF). Differential Diagnosis of Uterine mlms and Proposed Diagnostic Approach The differential diagnosis of uterine mlms includes a spectrum of benign and malignant lesions. A proposed algorithm for the diagnosis of mlms and its distinction from smooth muscle tumors with unknown malignant potential and benign lesions is outlined in Figure 7. In the routine assessment of a smooth muscle neoplasm with myxoid stroma, the first step is to determine the extent of myxoid change. By convention, a tumor is designated as myxoid if the neoplasm contains an abundant myxoid extracellular matrix in between muscle bundles and individual cells; often times, this correlates to a gelatinous appearance on gross examination, which was present in the majority of cases in our series. We suggest using a cutoff of 50% for the amount of myxoid matrix identified histologically to designate the tumor as a mlms, as this is a threshold that can easily be evaluated on histologic examination and represents the minimum amount of myxoid matrix identified histologically in our series, including those cases that were grossly gelatinous in appearance. It is also important to distinguish hydropic degenerative changes, which frequently occur in uterine leiomyomas and carry no clinical importance, from a truly myxoid extracellular matrix. 47 The latter is rich in glycosaminoglycans and proteoglycans, 48 which can be demonstrated by the Alcian blue stain; conversely, hydropic change and edema will be Alcian blue negative. Also, hydropic change is commonly accompanied by collagen deposition ( hyaline degeneration). 47 Once the myxoid nature of the lesion has been established, assessment of the tumor border is necessary. The presence of infiltrative tumor borders was a consistent finding in our series. This observation has been previously reported by Burch and Tavassoli, 4 who observed that an infiltrative tumor border is related to potential for aggressive behavior. In fact, the 3 cases with smooth borders recorded in their series were alive with no evidence of disease at follow-up, and the same outcome was reported in 4 additional case reports documenting a smooth tumor border (Table 5). 10,11,23,32 On the basis of our results and this cumulative evidence, we recommend a diagnostic algorithm based on an initial assessment of the tumor interface, followed by examination of the degree of nuclear atypia (nuclear pleomorphism and hyperchromasia), the mitotic rate in at least 50 HPF, and the presence or absence of coagulative tumor cell necrosis. If the tumor border is infiltrative and there is at least one of the following 2 or more mitoses per 10 HPF, moderate to severe (2+/3+) nuclear atypia, or unequivocal coagulative tumor necrosis a diagnosis of mlms can be rendered. An infiltrative tumor border in the absence of any other feature of malignancy should be viewed with caution; none of our cases had such a presentation, and experience in this scenario is limited. Review of the literature revealed 3 cases with an infiltrative border, 1+ atypia and no or only 1 mitosis per 10 HPF. 3,4,21 One subject died of disease at 8 years, 1 subject was alive with disease at 92 months, and 1 subject was alive and free of disease at 113 mo. Consequently, the diagnosis of mlms may be advisable in this situation, although a myxoid smooth muscle tumor of uncertain malignant potential could be considered given the paucity of data for this scenario. As myxoid smooth muscle tumors with a circumscribed border are rare, limited data are available, which hampers prognostication. By convention, if the tumor borders are smooth, there is no to minimal nuclear atypia, there is no necrosis, and mitoses are <2 per 10 HPF, the tumor can be diagnosed as a myxoid leiomyoma. It is less clear how to approach well-circumscribed tumors with moderate to severe nuclear atypia and/or mitotic rates of Z2 per 10 HPF. There are only 3 cases in the literature that describe a well-circumscribed mlms with moderate to severe nuclear atypia; all of these had mitotic rates of >2 per 10 HPF. 4 All 3 patients are alive without evidence of disease at 35, 90, and 105 months. One might argue that these tumors should be called of uncertain malignant potential; FIGURE 5. Tumor reclassified as inflammatory myofibroblastic tumor (Case 28; A F). The lesion has irregular borders (A) and contains spindle cells individually dispersed in a myxoid stroma, admixed with multiple chronic inflammatory cells (B). Alcian blue positive myxoid matrix is abundant, giving the tumor a loose hypocellular appearance (C). Tumor cells are strongly positive for Desmin (D) and ALK (E). FISH (F) was positive for dual-color, break-apart ALK rearrangement (A, hematoxylin and eosin; B, hematoxylin and eosin; C, Alcian blue; D, Desmin; E, ALK; F, ALK fluorescent in situ hybridization). www.ajsp.com 295

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016 FIGURE 6. Uterine myxoid neoplasm, not otherwise specified (Case 30; A F). The lesion has irregular borders (A and B) and contains small and uniform wavy spindle cells individually dispersed in an abundant myxoid matrix, strongly positive for Alcian blue stain (C and D). Tumor cells are negative for h-caldesmon (E) and ALK (F) by immunohistochemistry (A, hematoxylin and eosin; B, hematoxylin and eosin; C, hematoxylin and eosin; D, Alcian blue; E, h-caldesmon; F, ALK). 296 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus TABLE 5. Reported Cases of Uterine mlms in the English Literature to Date References No. Cases Age (y) Border Size (cm) Mitoses Atypia Necrosis FU Interval FU Status King et al 3 6 49 NA 8 0 1+ NA 11 y DOD 51 Infiltrative 9.5 2 1+ NA 9 y AWD 47 Infiltrative NA 1 1+ NA 8 y DOD 67 Infiltrative 5 2 1+ NA 2 y DOD 57 Infiltrative 15 2 2+ NA 1 y AWD 68 NA 15 1 1+ NA 1 y DOC Kim et al et al 37 1 50 NA NA 4 NA NA 24 mo AFD Shroff et al 6 1 50 Infiltrative 9 4 3+ NA 24 mo AFD Chen 7 1 57 NA 12 0 1+ NA 5 y DOD Yu et al 38 2 59 NA NA 0 NA NA 24 mo AFD 76 NA NA 1 NA NA NA NA Pounder and Iyer 8 2 32 Infiltrative 7 22 2+ NA 59 mo AWD 65 Infiltrative 5 15 2+ NA 12 mo DOD Salm and Evans 9 2 67 Infiltrative 16 19 NA NA 9 mo AWD 68 Infiltrative 8 8 NA NA NA NA Peacock and Archer 10 1 30 Circumscribed NA 2 1+ NA 18 mo AFD Aida et al 11 1 76 Circumscribed 10.5 30 NA Yes 12 mo AFD Kunzel et al 12 1 49 Infiltrative 9 30 3+ NA 5 mo AWD Fukunishi et al 13 1 39 NA 14 30 NA NA 16 mo DOD Fraga et al 14 1 63 Irregular 6.2 Low 1+ NA 4 mo AWD Schneider et al 15 1 65 NA 9.5 30 3+ focal NA 8 mo DOD Jones and Norris 31 2 64 NA 10 4 2+ Yes 30 mo DOD 44 NA 12 11 2+ Yes 12 mo AWD Sprogøe-Jakobsen and Hølund 16 1 48 Infiltrative 4 0 NA NA 36 mo AFD Chang and Shim 18 1 54 Infiltrative 17 8 3+ focal Present 18 mo DOD Oruc et al 19 1 54 Irregular 20 2 NA Yes 1.5 mo AFD Kasahara et al 21 1 56 Infiltrative (f) 10 0 1+ NA 15 mo DOD Takano et al 32 1 70 Smooth 12 6 NA NA 70 mo AFD Mittal et al 20 1 73 Infiltrative 7 2 1+ NA 13 mo AFD Kaleli et al 23 1 39 Circumscribed 45 2 1+ NA 34 mo AFD Ng et al 29 1 74 Infiltrative 20 8 1+ Yes NA NA Kagami et al 22 1 20 NA 8.5 1 1+ NA 92 mo AWD Malhotra et al 24 1 60 Infiltrative NA > 10 2+ Yes NA AFD Vigone et al 25 1 59 Infiltrative 4.5 5 2+ Yes 6 mo DOD Botsis et al 26 1 69 Infiltrative 5 3 1+ NA 84 mo AFD Karpathiou et al 27 1 30 Infiltrative 7 0 1+ No NA NA Abeler et al 2 18 39-85 Infiltrative: 12 < 5: 3 <5: 11 1+: 6 Yes: 12 NA NA 6-10: 9 6-10: 5 2+: 11 Smooth: 6 > 10: 5 11-15: 2 3+: 1 No: 6 Burch and Tavassoli 4 12 29 Infiltrative NA 1 3+ NA 24 mo AWD 46 Infiltrative 9 2 1+ NA 28 mo DOD 42 Infiltrative 7 1 3+ NA 41 mo AFD 50 Infiltrative NA 6 1+ NA 79 mo AFD 57 Infiltrative 12 0 1+ NA 113 mo AFD 64 Infiltrative 10 7 2+ NA 30 mo DOD 31 Infiltrative (f) NA 0 2+ NA 19 mo AWD 59 Infiltrative (f) 4.2 3 2+ NA 36 mo AFD 69 Infiltrative (f) 7.5 8 1+ NA 44 mo AFD 63 Circumscribed 8 5 3+ NA 105 mo AFD 64 Circumscribed 7.5 57 2+ NA 90 mo AFD 78 Circumscribed 13 6 3+ NA 35 mo AFD Ritchie and Kumari 28 1 38 NA 9.5 +++ 3+ Yes NA AWD Barone et al 30 1 66 Infiltrative 11.5 20 NA Yes NA NA Mitrache et al 34 1 61 NA 17 20 2+ Yes NA NA Definition of nuclear atypia was not uniform among references. In most, it was stated as mild (or 1+), moderate (or 2+), or severe (or 3+). +++, Number of mitoses as reported in reference 28. AFD indicates alive free of disease; AWD, alive with disease (recurrence); DOC, dead of other causes; DOD, dead of disease; f, focally; FU, follow-up; NA, not available/not specified. however, data in scenarios such as this are still insufficient to support such a recommendation. The diagnosis of myxoid smooth muscle tumor of uncertain malignant potential seems appropriate if the mitotic rate is <2 per 10 HPF, whereas the diagnosis of mlms should be strongly consideredifthemitoticrateisz2 per10hpf. There are 2 cases diagnosed as mlms in the literature that have borders described as smooth or circumscribed, with 1+ atypia and a mitotic rate of 2 per 10 HPF. 10,23 Both of these individuals are alive without evidence of disease at 18 and 34 months. Although there is no documented adverse outcome, the follow-up www.ajsp.com 297

Parra-Herran et al Am J Surg Pathol Volume 40, Number 3, March 2016 FIGURE 7. Algorithm and proposed criteria for the diagnosis of uterine myxoid smooth muscle neoplasms. *Nuclear atypia is defined as the presence of hyperchromasia and enlargement defined as 1+ (similar in size to the nuclei of adjacent non-neoplastic myometrium), 2+ (2X size), and 3+ (Z3X size). a The term myxoid applies to tumors with a prominent myxoid stroma. A proposed cutoff of 50% of the tumor volume is proposed. b By convention, the diagnosis of myxoid leiomyoma is made in the absence of any worrisome features for malignancy, including infiltrative borders. 35 c None of the cases in our series had a circumscribed border. In contrast, none of the cases of mlms in the literature with a documented circumscribed/smooth border had documented recurrences or death (see also Table 5). The diagnosis of mlms can be made if 2 or more of the 3 features of malignancy are present (similar to conventional LMS). If there is only 1 feature of malignancy in a tumor with well-defined borders, the diagnosis of myxoid smooth muscle of unknown malignant potential (STUMP) could be considered, although experience is very limited. d None of the cases in our series fell into this category. Although the diagnosis of myxoid leiomyosarcoma should be considered, experience in this scenario in the literature is scarce. interval is too short to draw a definitive conclusion that these tumors are benign; therefore, the diagnosis of a myxoid smooth muscle tumor of uncertain malignant potential (vs. mlms as they were diagnosed and presented in the literature) also may be considered for this combination of findings. As shown in our study, other entities with myxoid appearance can mimic mlms and should be routinely considered. An illustrative diagram with the most important entities in the differential is depicted in Figure 8. In particular, inflammatory myofibroblastic tumor is an important confounder, given its morphologic overlap with smooth muscle neoplasms. Inflammatory myofibroblastic tumor is an uncommon neoplasm. Although its anatomic distribution is wide, its prevalence in the female genital tract is low; only few series and case reports are available in the English literature. 49,50 Histologically, this lesion is characterized by a spindle cell population in a myxoid to collagenous stroma admixed with a lymphoplasmacytic infiltrate. 51,52 Inflammatory myofibroblastic tumor is associated with rearrangement of the ALK gene at 2p23, which is identified in 50% to 70% of tumors in children and young adults but is less frequent in older patients. 53,54 This entity is currently classified as having an intermediate malignant potential, given its low but well-documented risk for local recurrence and metastasis. 51,52 Although most published cases of uterine inflammatory myofibroblastic tumor had a benign clinical course, aggressive behavior has been recently reported. 55 In fact, 2 of the 4 cases reclassified as inflammatory myofibroblastic tumor in the present series (1 confirmed by molecular studies) died of disease. These 2 cases have not been previously reported and represent further evidence of the potential for aggressive behavior in uterine inflammatory myofibroblastic tumor. Distinction between inflammatory myofibroblastic tumor and mlms is critical. Although surgical excision is the standard of treatment for both conditions, targeted therapy with tyrosine-kinase inhibitors is a valid option in recurrent or unresectable inflammatory myofibroblastic tumor. Such distinction, however, can be difficult, because both entities are characterized by a prominent myxoid extracellular matrix and a variably prominent chronic inflammatory infiltrate. Tumor borders can be infiltrative in both lesions. In addition, inflammatory myofibroblastic tumor can contain fascicular areas reminiscent of smooth muscle differentiation. The presence of more classic loose myxoid areas containing myofibroblastic cells with a 298 www.ajsp.com

Am J Surg Pathol Volume 40, Number 3, March 2016 Myxoid Leiomyosarcoma of the Uterus FIGURE 8. Differential diagnosis of mesenchymal myxoid tumors of the uterus and relevant histopathologic features. tissue culture like appearance and open vesicular nuclei should raise the possibility of inflammatory myofibroblastic tumor and prompt consideration for further testing (ALK immunohistochemistry and in situ hybridization). ALK overexpression by immunohistochemistry may assist in the distinction between uterine inflammatory myofibroblastic tumor and mlms, because it is consistently present in the former. 50,53 In contrast, on the basis of our results most cases of mlms are negative for this marker (85.7%). Nonetheless, ALK does not appear to be entirely specific for inflammatory myofibroblastic tumor, and caution must be exercised in its interpretation. In our series, we observed ALK positivity in 3 tumors otherwise consistent with the diagnosis of mlms. The underlying mechanism for ALK expression in these cases is unclear. In situ hybridization was negative in 2 of such cases, arguing against a genetic mechanism and suggesting ALK accumulation unrelated to classic ALK gene rearrangements. Importantly, ALK overexpression by immunohistochemistry and ALK rearrangements have been documented in several tumors other than inflammatory myofibroblastic tumor, particularly rhabdomyosarcoma and malignant peripheral nerve sheath tumor. 56 ROS1, another tyrosine-kinase receptor, appears to have a role in the pathogenesis of ALK-negative inflammatory myofibroblastic tumors. ROS1 kinase fusions have been recently reported in 4 cases, and successful therapy with a tyrosine-kinase inhibitor has been documented in one of them. 57 In addition, ROS1 immunohistochemistry appears to predict gene rearrangement in inflammatory myofibroblastic tumors. 58 Abnormalities in the ROS1 gene have not been reported in any uterine neoplasm to date. In our study, ROS1 immunohistochemistry was negative in all cases of mlms, which suggests that these tumors are not related to ROS1 alterations. Likewise, cases of inflammatory myofibroblastic tumor (either confirmed or suspected) lacked ROS1 expression. Further investigation is necessary to determine the frequency of ROS1 gene rearrangements and the value of ROS1 immunohistochemistry in uterine inflammatory myofibroblastic tumors. Uterine neoplasms with myxoid stroma, other than those with smooth muscle or myofibroblastic origin, are exceedingly rare. Endometrial stromal sarcomas may occasionally have a predominant fibromyxoid pattern, and a case of purely myxoid endometrial stromal sarcoma has been previously reported. 59 Attention to the classic architectural and cytologic features of endometrial stromal neoplasms will assist in the diagnosis, in particular the absence of severe nuclear pleomorphism and the presence of a uniform vascular network composed of thin-walled vessels. In addition, negativity of smooth muscle markers will favor an endometrial stromal neoplasm. The distinction is clinically relevant, as low-grade endometrial stromal sarcomas have an indolent course with long-term relapses and are amenable to endocrine therapy. Isolated cases of myxoid liposarcoma arising in a leiomyoma and myxoid malignant fibrous histiocytoma have been previously described. 60,61 Other entities, such as myxoma or low-grade fibromyxoid sarcoma, have yet to be reported. One case in our series (case 30) lacked histologic and immunohistochemical features of smooth muscle or myofibroblastic differentiation and resembled www.ajsp.com 299