Histological and molecular analysis of cellular leiomyoma with sclerosis: linked to HMGA2 overexpression

Abstract

HMGA2 overexpression is found in 10–15% of leiomyomas (LM). HMGA2 overexpression is common in variants of hydropic, intravenous and lipo-LM. Cellular or highly cellular LM (CLM) is a LM variant with a less well-defined molecular nature. In this study, we identified and examined 52 hypercellular LM with sclerotic collagen, herein defined as cellular leiomyoma with sclerosis (CLM-S). CLM-S shows large tumour size (average 12.2 cm) and characteristic histology of tumour cells, arranged in cellular fascicles, sheets and trabeculae with abundant dense, pink sclerotic extracellular matrix in bands and nodules and increased vascularity. Tumour cells are uniform with small, round–oval nuclei and scant, pale–eosinophilic to vacuolated cytoplasm reminiscent of pericytes. The differential diagnosis of CLM-S includes conventional CLM, endometrial stromal tumours and perivascular epithelioid cell tumour. Immunohistochemical profile [HMGA2, fumarate hydratase, smooth muscle markers, Melan A and HMB-45] and molecular alterations [by HMGA2 mRNA reverse transcription–polymerase chain reaction (RT–PCR), HMGA2 fluorescence in-situ hybridisation and MED12 sequencing] were analysed in comparison to matched myometrium and CLM controls. Remarkably, 96% (50 of 52) of CLM-S demonstrated diffuse positive immunoreactivity for HMGA2 and up to an 80-fold increase in HMGA2 mRNA, determined by RT–PCR. FISH analysis with break-part probes at intron 3 and the 5’ UTR detected HMGA2 rearrangements in 47% (18 of 38) of CLM-S. All CLM-S retained expression of fumarate hydratase. No MED12 mutations were found in any CLM-S. Our findings show that CLM-S has unique and characteristic histomorphology probably driven by HMGA2 overexpression.

Introduction

Leiomyoma (LM) is the most common benign neoplasm of women’s reproductive organs. LM displays heterogeneous histology, and many variants are recognised by their characteristic morphological features. The most recent World Health Organisation tumour classification (WHO 2020) summarises the histological LM variants.¹ Recent molecular advances have demonstrated that most LMs are driven by single gene alteration(s)²; however, the association between gene mutations and histological variants is only partially understood.

In the past decade, significant progress in the molecular characterisation of LM has shown that approximately 70% of the LMs harbour MED12 mutations, 10–15% have HMGA2 overexpression and < 1% show fumarate hydratase (FH) alterations.³ Awareness of such molecular findings and their relation to histomorphology may significantly improve the diagnosis of certain LM variants. For example, LM with FH alterations demonstrates some characteristic histological and nuclear features including haemangiopericytoma-like vessels, oedema, hyaline globules, pink macronucleoli, perinucleolar halos and neurilemoma-like growth patterns. Immunostaining helps to further classify tumours as fumarate hydratase-deficient LM (FH-LM).⁴ LM with MED12 mutations (MED12-LM) is the prototypical tumour type and displays a mixture of overt smooth muscle tumour cells and tumour associated fibroblasts, which can be less pronounced in other LM variants.⁵ LM with HMGA2 overexpression (HMGA2-LM) is the second most common group, and comprises several histological variants including hydropic leiomyoma,⁶ intravenous leiomyomatosis⁷ and lipoleiomyoma.⁸

Cellular or highly cellular leiomyoma is a histological variant of LM with a largely unknown histopathogenesis and molecular landscape. Early studies of cellular LM (CLM) provide morphological and immunostaining patterns which aid in differentiation from endometrial stromal tumours. CLM shows hypercellularity, round–oval and spindled cells with high nuclear cytoplasmic ratio, thick-walled vessels and an irregular blended tumour–myometrial interface.⁹ However, background sclerosing collagens are not described in CLM.

In this study, we retrospectively review CLM with characteristic extracellular dense and pink, sclerotic collagen, defining them as CLM with sclerosis (CLM-S). We analyse CLM-S’s pathological and clinical presentation, immunoprofile and molecular findings. Our study shows that CLM-S harbours HMGA2 overexpression in > 96% of tumours, but no FH deficiency or MED12 mutations. We report our histomorphology and molecular findings of CLM with prominent sclerosis, which are different from those of conventional CLM.

Materials and methods

CASE SELECTION

This study was approved by Northwestern University’s institutional review board. In our previous large cohort study of randomly selected/unprocured LM, we found that 10% displayed HMGA2 overexpression.¹⁰ Further evaluation of these HMGA2-LM demonstrated unique histological features: (1) increased cellularity and (2) increased dense sclerosing extracellular matrix. We therefore defined them as CLM-S. CLM-S samples were collected from 2013 to 2021. Two pathologists (B.B.G. and J.J.W.) independently screened selected haematoxylin and eosinstained representative sections of candidate tumours. All slides of tumour sections were examined. Tumours with diagnostic agreement and diffuse presence of CLM-S microscopic features (see below) were included for further analysis. A total of 52 CLM-S was included, as well as 22 CLM without sclerosing collagen (CLM) and matched myometrium as controls.

SLIDE REVIEW AND HISTOLOGICAL FEATURES

Microscopic features of CLM-S that served as inclusion criteria at screening were the following: (1) tumour hypercellularity; (2) spindled tumour cells with increased nuclear cytoplasmic ratio and small, slightly elongated/round–oval nuclei; (3) background dense pink and sclerotic collagen bundles; and (4) increased thick-walled vessels. For each included case, two pathologists (B.B.G. and J.J.W.) recorded tumour cellularity, cellular growth pattern, tumour cytology and nuclear features (shape/size/chromatin/nucleoli), background extracellular matrix and vasculature. Mitotic counts were performed among 10 ×40 high-power fields (HPF, equal to square 0.16 mm). Background sclerosing collagen was defined as mature collagen (dense, dark pink, discrete and organised). Recorded data for collagen comprised extent within the tumour section (percentage) and arrangement/relation to tumour cells. Vessel density was assessed using parameters from the authors’ previous study: evaluation among three ×10 medium-power fields (MPF, equal to square 4.0 mm) and counted as number of thick-walled vessels per MPF.⁶

IMMUNOHISTOCHEMISTRY (IHC)

Full-mounted 5-μm tumour tissue sections were prepared for immunostaining for HMGA2 (BioCheck, Foster City, CA, USA; #59170AP 1:500 antibody dilution) in all CLM-S cases (to ensure a complete assessment for HMGA2 immunoreactivity). A tissue microarray (TMA) was prepared for CLM-S, CLM and myometrial controls to use for additional markers: desmin, h-caldesmon, CD10, Melan A, HMB-45 and FH. IHC procedures were performed by Northwestern University’s Path Core facilities on a Ventana Nexus automated system (Basel, Switzerland), as described previously.⁶ Stains were scored, when applicable, by percentage of positive tumour cells from 0 to 100% and by intensity of expression as negative (0), weak (1+), moderate (2+) or strong (3+). Specifically, strong and diffuse nuclear positive immunoreactivity for HMGA2 (2+ and ≥ 50%) was considered as HMGA2 overexpression. All other patterns, including no staining and weak and/or focal immunoreactivity (0–1+ and < 50%), were considered as negative. Lastly, the results were semi-quantitatively analysed.

RNA EXTRACTION AND RT – PCR

Total RNA was extracted from formalin-fixed paraffin-embedded tissue (FFPE) blocks of leiomyoma and matched myometrium by the high pure FFPET RNA isolation kit (Roche, Indianapolis, IN, USA), according to the manufacturer’s directions and as described previously.⁶ cDNA was generated by reverse transcription from 10 ng of total RNA using qScript^® cDNA synthesis kit according to the manufacturer’s instructions (Quantabio, Beverly, MA, USA). Relative expression levels of HMGA2 were determined by quantitative RT–PCR (qRT–PCR) using the QuantStudio^™ 3 RT–PCR system from Applied BioSystems (Waltham, MA, USA) and normalised to beta-actin mRNA levels. qRT–PCR was performed using Power SYBR^™ Green PCR Master Mix in an Applied Biosystems 7900HT RT–PCR system. Fold change values were calculated using the comparative Ct method using endogenous control glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The experiments were repeated in triplicate.

FLUORESCENCE IN-SITU HYBRIDISATION (FISH)

A two FISH break-apart probe set was used in this study to evaluate HMGA2 gene rearrangement, with the 5’ HMGA2 labelled in spectrum-orange (5’ end) and the 3’ HMGA2 labelled in spectrum-green (Empire Genomics, Buffalo, NY, USA), as described previously.⁶ The first HMGA2 probe set was designed to detect alterations in the mid-portion (intron 3) of HMGA2. The second HMGA2 probe set was designed to detect alterations in the genomic region 20Kb upstream of 5’ UTR HMGA2. FISH using the break-apart probes for HMGA2 was performed on a TMA of FFPE tissue samples for both tumours and controls following the manufacturer’s instructions and assessing HMGA2 rearrangement status and/or copy numbers. Probe integrity and localisation were first evaluated on normal metaphase spreads from a peripheral blood sample by the Clinical Cytogenetics Laboratory at Northwestern Memorial Hospital. FISH signals were assessed at 100× HPF magnification to ensure intracellular location, and for each case or control an attempt was made to count 200 cells in total. FISH patterns showing only yellow fusion signals with colocalised/direct juxtaposition of spectrum-orange (red) and spectrum-green (green) signals were scored as intact HMGA2 and negative for gene rearrangement. Fusion signals co-existing with separate hybridisation signals showing a single green (1G) or red signal (1R) were scored as positive for HMGA2 gene rearrangement. Cells meeting this criterion showed 1F/1R/1G, 1F/1G or 1F/1R signals, and cases were scored as positive if a threshold of ≥ 10% was reached. Cases were excluded as uninterpretable if no signals could be visualised within the tumour tissue present.

MED12 MUTATION ANALYSIS

As described previously,⁶ genomic DNA was extracted from tumour FFPE tissue sections with a DNA extraction kit (Qiagen, Boston, MA, USA), and 50 ng of DNA were loaded into PCR. DNA from exon 2 of MED12, with flanking exon–intron junction sequences, was amplified with primers 5’-GCC CTT TCA CCT TGT TCC TT-3’ (forward) and 5’-TGT CCC TAT AAG TCT TCC CAA CC-3’ (reverse). PCR products were purified by ExoSAP-IT reagent (Affymetrix, Inc., Santa Clara, CA, USA) following the manufacturer’s instructions. Sequencing of the purified DNA products was performed at Northwestern University’s NUSeq Core using Applied Biosystems’ BigDye version 3.1 (ThermoFisher Scientific, Waltham, MA, USA). The reactions were then run on Applied Biosystems’ 3730xl DNA Analyser. Mutations/variations were analysed by DNASTAR Lasergene version 9 software (Madison, WI, USA).

DATA INTERPRETATION AND STATISTICAL ANALYSIS

GraphPad Prism software (version 6; La Jolla, CA, USA) was used for statistical analysis. Age and tumour size data were presented as medians and ranges, while other data were presented as mean and standard deviation. Either a Student’s t-test or χ² test was used to determine statistical significance. A P-value less than 0.05 was considered statistically significant.

ETHICS APPROVAL STATEMENT

This study was approved by Northwestern University’s institutional review board.

Results

CASE AND SPECIMEN FINDINGS

Patients with CLM-S were diagnosed at a mean age of 46 years, but spanning a wide range from 25 to 78 years. A similar mean age and range were noted for patients with CLM (46.9, aged 29–72 years). Approximately 65% of CLM-S (34 of 52) were from hysterectomies; the remaining cases were from myomectomies. CLM-S showed larger average and more variable tumour size than CLM: 12.2 ± .27 cm for CLM-S versus 8.1 ± 0.8 cm for CLM (P < 0.01). Most CLM-S were uterine in origin (~77% in corpus and ~ 2% in lower uterine segment) with no significant difference of submucosal, intramural and subserosal locations (data not shown). The cervix and adnexa/pelvis were the second most common sites of origin for CLM-S (~10% each) and, very rarely, tumours were vaginal in origin (~2%). In comparison, all CLM were uterine in origin. We observed CLM-S to occur more often in the setting of multiple LM than CLM; however, no significant difference (n = 23, ~44% for CLM-S and n = 8, ~36% for CLM) was observed. Interestingly, CLM-S manifested as a solitary lesion or as a distinct subset in a majority background of usual-type LM when encountered in the multiple LM setting, suggesting an independent tumorigenesis from other LM. Biodemographic and pathological specimen data of patients with CLM-S and CLM are displayed in Table 1 and S1.

Table 1.

General information for the selected cases for this study

Title	Subtitle	CLM-S	CLM	P-value
No. of cases		52	22
Age in years (range)		46 (25–78)	46.9 (29–72)	0.257*
Clinical procedure	Myomectomy	18 (34.6%)	3 (14%)	0.092**
	Hysterectomy	34 (65.4%)	19 (86%)
Tumour size (cm)		12.2 3.27	8.1 ± 0.8	0.021*
Mitotic count/10 HPF,	Mean (range)	2.32 (0–7)	2.27 (0–7)	0.931*
Anatomical location	Vagina	1 (1.9%)	0
	Cervix	5 (9.6%)	0
	LUS	1 (1.9%)	0
	Uterus	40 (76.9%)	22 (100%)
	Adnexa/pelvic	5 (9.6%)	0
No. of tumours	Solitary	29 (55.8%)	14 (63.6%)	0.611**
	Multiple	23 (44.2%)	8 (36.3%)

CLM, cellular leiomyoma; CLM-S, cellular leiomyoma with sclerosis; HPF, high-power field; LUS, lower uterine segment.

^*Student t-test

^**χ² test.

GROSS AND MICROSCOPIC FINDINGS

A total of 35 CLM-S had gross photographs available. Tumour were grossly large, bulging and nodular with a mainly well-demarcated border. Cut surfaces were variegated light yellow and pink-to-tan, soft in character and vaguely whorled (Figure 1). Some tumours were richly vascular and showed spotty haemorrhage (Figure 1C).

Figure 1.

Gross photographs of cellular leiomyoma with sclerosis (CLM-S). CLM-S of cervical origin in the setting of multiple leiomyomas (A) and of adnexal/pelvic (B) and uterine origins – intramural (C, D). Large tumour size is indicated by the ruler for comparison.

Microscopic sections of CLM-S displayed striking hypercellularity and characteristic background collagen and extracellular matrix-forming bands, cords and nodules with crisp edges. CLM-S showed collagen scattered between hypercellular tumour cell groups with occasional concentration around vessels (Figure 2). A starburst-like collagen pattern was seen, similar to extracellular matrix in endometrial stromal tumours (Figure 2I). Collagenous extracellular matrix in CLM-S was present across all tumour slides, and the amount and density of sclerosing collagen varied from case to case. In comparison, mature well-formed collagen was dense and darker pink, discrete and well-organised with crisp borders, forming spheres and jagged/irregular shapes. Mature well-formed collagen involved, on average, 20.7% of tumour tissue (range = 1–60%, Supporting information, Table S1).

Figure 2.

Characteristic collagen and extracellular matrix seen in cellular leiomyoma with sclerosis. Collagen and extracellular matrix are abundant forming bands, cords (A–D) and nodules (E–G). Edges are well-delineated. The pink sclerotic matrix varies in density. Extracellular matrix is located between tumour cells groups and occasionally concentrated around vessels (H). A starburst-like pattern can be seen mimicking a feature of endometrial stromal tumours (I).

CLM-S showed increased density of thick-walled vessels compared to typical leiomyoma. CLM-S demonstrated on average 14 thick-walled vessels/10×-MPF (range = 3–29 vessels), slightly higher than the average vessel density of typical leiomyomas (9/10× MPF, range = 0–21 vessels). Ultimately, increased vessel density was not a hallmark feature of CLM-S and was excluded as a diagnostic criterion accordingly. CLM-S also displayed a mainly well-demarcated, pushing border with surrounding myometrium. Rare cases of CLM-S showed an irregular border and blending into surrounding myometrium, as previously reported in CLM.⁹ Lastly, occasional ischaemic, infarct-type or degenerative necrosis was present in CLM-S.

Tumour cells of CLM-S grow in a wide range of patterns beyond the typical fascicular arrangement of smooth muscle neoplasms. Tumour cells are arranged in mainly organised cords, fascicles, sheets and trabeculae with a streaming or intersecting/perpendicular character (Figure 3). These architectural patterns differ from the well-organised solid sheets and thick bundles/fascicles of CLM. The cytology of CLM-S is similar to cellular or highly cellular LM. Tumour cells of CLM and CLM-S are small, round–oval with high nuclear cytoplasmic ratio and occasionally elongated nuclei. Chromatin is evenly distributed, ranging dark and granular to vesicular with inconspicuous or pinpoint nucleoli (Figure 4A). Tumour cell cytoplasm is relatively scant and pale–eosinophilic, with occasional perinuclear vacuolisation, in keeping with a feature of smooth muscle neoplasms (Figure 4A). We observed low mitotic counts in the range of 0–7 mitotic figures/10 HPFs (Supporting information, Table S1). All CLM-S displayed uniform, no more than mild nuclear atypia, except for one case which showed degenerative-appearing bizarre nuclei (Figure 4B).

Figure 3.

Growth patterns in cellular leiomyoma with sclerosis. Tumour cells are arranged in mainly organised cords (A, B) with occasional palisading (A), fascicles and sheets (C, D) with streaming. Also seen are disorganised bundled (E) and trabecular (F) patterns with intersecting/perpendicular feature.

Figure 4.

Cytological features and histological heterogeneity of cellular leiomyoma with sclerosis. A, Tumour cells display high nuclear cytoplasmic ratio with small, round–oval nuclei. Chromatin is dark and granular to vesicular with even distribution and inconspicuous–pinpoint nucleoli. Cytoplasm is scant and pale–eosinophilic with occasional perinuclear vacuolisation. B, One tumour with remarkable nuclear atypia. C, D, Examples of cellular and sclerosing leiomyoma display predominantly cellular and sclerotic pattern with focal features of lipoleiomyoma (C) and hydropic leiomyoma (D).

Upon thorough examination of all our tumour slides, we found that CLM-S contained purely cellular and sclerosing pattern in 75% (39 of 52) of cases. The remaining 25% (13 of 52) showed a predominantly cellular and sclerosing pattern admixed with few features of lipoleiomyoma and hydropic leiomyoma (Figure 4C,D, Table 2, Supporting information, Table S1). These findings suggested that a common, similar histogenesis of CLM-S, lipoleiomyoma and hydropic LM related to HMGA2 overexpression. Similar to the hypocellularity phenomenon of hydropic leiomyoma from abundant oedema,⁶ CLM-S contained scattered less cellular areas that were diluted by large amounts of collagenous extracellular matrix.

Table 2.

Histological heterogeneity of cellular leiomyoma with sclerosis

Histology	No. of cases	%
CS	39	75.0
CS + LP	7	13.5
CS + HD	3	5.8
CS + LP + HD	3	5.8

CS, cellular with sclerosing; HD, hydropic; LP, lipo.

IMMUNOHISTOCHEMISTRY AND MOLECULAR ANALYSIS

To evaluate potential molecular changes associated with CLM-S, we first performed IHC for HMGA2 and FH on all cases. Remarkably, almost all CLM-S (96.2%, 50 of 52) showed strong and diffuse positive expression of HMGA2 (Figure 5, Table 3). In comparison, 18.2% (four of 22) CLM showed HMGA2 overexpression. Two CLM-S exhibiting typical histomorphology were negative for HMGA2, showing absent to only weak, focal staining (Figure 6). All cases demonstrated diffuse positive for desmin or h-caldesmon. Immunostains for CD10, Melan-A and HMB45 were performed on all cases, including two CLM-S immunonegative for HMGA2. Approximately 21% (11 of 52) of CLM-S showed focal immunopositivity for HMB45 (in a range of 1–15% of tumour cells) and no detectable Melan-A expression was found (Table 3, Supporting information, Table S1). The findings were in keeping with the diagnosis CLM-S, and not supportive of endometrial stromal tumour or perivascular epithelioid cell tumour (PEComa). In addition, all cases of CLM-S presented normal/intact expression of FH immunostain, indicating no loss or altered fumarate hydratase.

Figure 5.

Cellular leiomyoma with sclerosis (CLM-S) with HMGA2 overexpression. A, B, Close view of CLM-S with strong and diffuse immunoreactivity for HMGA2. C, Representative histology (left) and HMGA2 immunostain (right) slides of 48 CLM-S with HMGA2 overexpression.

Table 3.

Immunohistochemical and molecular findings of cellular leiomyoma with sclerosis

Biomarkers	Testing modality	Category	No. of cases	% of cases
HMGA2	IHC	Positive	50/52	96.2
	FISH	Positive total	18/38	47.4
		Positive for probe 1*	6/38	15.8
		Positive for probe 2**	7/38	18.4
		Positive for both probes	5/38	13.2
HMB45	IHC	Positive	11/52	21.2
		Mean (range %)	0.731 (0–15%)
Melanin-A	IHC	Positive	0/52	0
FH	IHC	Loss of FH	0/52	0
MED12	Sequencing	MED12 mutation	0/36	0

IHC, immunohistochemistry; FISH, fluorescence in-situ hybridisation.

^*Probe 1 = break-apart at intron 3 of HMGA2

^**probe 2 = break-apart at 5’ UTR region of HMGA2.

Figure 6.

Cellular leiomyoma with sclerosis (CLM-S) without HMGA2 overexpression. Representative histology (left) and HMGA2 immunostain (right) slides of two CLM-S which lacked HMGA2 overexpression.

Next, we investigated the level of HMGA2 mRNA expression in CLM-S through real-time PCR performed on 40 cases with HMGA2-positive immunostaining. CLM-S showed higher HMGA2 mRNA expression, with up to 80-fold increase relative to matched myometrium controls (Figure 7A, Table 3, Supporting information, Table S1). We then evaluated the association of HMGA2 overexpression with a chromosomal translocation or HMGA2 gene rearrangement. We performed FISH with break-apart probes for HMGA2 (Figure 7B,C) on 38 cases of CLM-S with HMGA2-positive immunostaining. FISH showed that 47% (18 of 38) of CLM-S demonstrated HMGA2 gene rearrangement (Figure 7D–G, Table 3), confirming that nearly 50% of CLM-S gain HMGA2 overexpression due to gene rearrangement. Further analysis revealed that seven cases of gene rearrangement were detected by the 5’ UTR break-apart probe, six cases by the intron 3 break-apart probe and five cases detected by both break-apart probes (Table 3, Supporting information, Table S1). Overall, 31.6% (12 of 38) of tumours had gene rearrangement at the 5’ UTR and 15.8% (six of 38) at intron 3. For the five cases of HMGA2 rearrangement detected by both break-apart probes, a gene rearrangement close to the starting point of exon 1 may be plausible. The intron 3 probe contains 98 Kb between the spectrum-orange and spectrum-green components, whereas HMGA2’s first three exons extend only 14 Kb, making overlap possible (Figure 7B,C). Lastly, there was no difference of mRNA expression level in CLM-S with (log² = 2.399) and without (log² = 2.283) HMGA2 rearrangement.

Figure 7.

Real-time polymerase chain reaction (RT–PCR) and fluorescence in-situ hybridisation (FISH) analysis in cellular leiomyoma with sclerosis (CLM-S). A, The histoplot illustrates an increase of HMGA2 mRNA expression, detected by RT–PCR, in CLM-S (orange bars) relative to matched myometrium (blue bars). Small T-bars represent standard error. B,C, Diagrams illustrating the break-apart probes used to evaluate the HMGA2 gene at intron 3 (B) and 20Kb upstream of the 5’ UTR region (C). D–G, Corresponding representative images highlight four CLM-S with HMGA2 rearrangement by intron 3 probe (D, E) and 5’ UTR probe (F, G). Tumour cells show one yellow overlapping hybridisation signal or direct juxtaposition of red and green signals, and separate hybridisation signals producing distinct green or red colours with lack of signal overlap.

We examined MED12 mutations by Sanger sequencing in our first 36 consecutive, unprocured CLM-S. No MED12 mutations were identified in any tumours (Table 3). These findings are limited, but indicate that genetic/genomic alterations of HMGA2 and MED12 genes are mutually exclusive in these tumour types.

Discussion

LM is a benign neoplasm with heterogeneous histology. CLM with dense and sclerosing collagen is not well-described in gynaecological pathology. Distinguishing this entity can be a diagnostic challenge due to overlapping features with entities: usual-type LM, conventional CLM, hydropic LM, endometrial stromal tumour or PEComa. Our current series of 52 cases highlights an LM that displays tumour hypercellularity and abundant background dense, sclerotic extracellular matrix. We define this LM as CLM-S and establish its molecular relation to HMGA2 overexpression. We found that: (1) ~96% of CLM-S display HMGA2 overexpression by IHC, confirmed with RT–PCR for HMGA2 mRNA level; and (2) almost half (~47%) of CLM-S contain a HMGA2 gene rearrangement detectable by FISH. All CLM-S showed neither loss of fumarate hydratase (FH) expression nor MED12 mutations, suggesting that HMGA2 abnormality represents a mutually exclusive event in CLM-S’s tumorigenesis. Our findings ultimately provide a potential biological association to CLM-S tumour morphology, aiding the pathological diagnosis. Correlation of molecular findings with tumour morphology is beneficial to recognising LM’s histological variants outlined in the WHO 2020 classification.¹

We found CLM-S to contain certain gross and microscopic features. Tumours from our series were, on average, larger and more variable in size than the established, conventional CLM (12.2 ± 3.27 versus 8.1 ± 0.8 cm, respectively). Grossly, CLM-S are mainly well-demarcated, bulging, nodular tumours showing variegated light yellow and pink-to-tan cut surfaces with softening and spotty haemorrhage. CLM-S occurred mainly in the uterus (~77% corpus) and occasionally in the cervix, vagina and adnexa/pelvis. Microscopic features of CLM-S that are most striking and indicative of the diagnosis are tumour hypercellularity and background dense, pink sclerotic collagen appreciable from low-power magnification (4×). Hypercellular groups in CLM-S form a wide range of patterns beyond those encountered in typical smooth muscle neoplasms. Tumour cells are arranged in mainly organised cords, fascicles, sheets and trabeculae with a streaming or intersecting character. Tumour cells of CLM-S are uniformly small, round–oval with high nuclear cytoplasmic ratio; however, chromatin is evenly distributed, and mitotic activity is low. We found only one case in our series to contain bizarre nuclei (Figure 4B). Cytoplasm is scant with pale eosinophilia and occasional perinuclear clearing. Overall, the cytological features of CLM-S should convey a benign nature.

Present between tumour cell groups in CLM-S are characteristic background extracellular matrix-forming bands, cords and nodules with crisp edges and variable density. Occasionally, the pink sclerotic collagen is concentrated around vessels or produces a starburst-like pattern. Ultimately, CLM-S’s histological findings are unique among smooth muscle tumours, except cellular/highly CLM, and instead mimic endometrial stromal tumours (EST) or PEComa. Careful review for areas of more typical and heterogeneous histology may be helpful. An immunohistochemical panel of relevant markers should clarify any diagnostic confusion.¹¹ These actions are especially important considering that CLM-S and highly CLM share the following gross features with EST: soft consistency, tan-to-yellow cut surface, striking vascularity and irregular margins with the surrounding myometrium.⁹ Additionally, CLM-S will display thick-walled vessels that are a hallmark differentiation from the small arterioles of EST. Although CLM-S and PEComa share morphological findings, especially cytoplasmic features and background sclerosis, PEComa tumour cells grow in a perivascular nested pattern not present in CLM-S and express melanocytic markers unlike CLM-S. This study suggests that HMGA2 can be used as a surrogate marker to differentiate CLM-S from EST and PEComa.

As expected, CLM-S predominantly shares overlapping features with conventional cellular/highly CLM. CLM displays increased cellularity relative to uterine myometrium and tumour cells in diffuse/sheeted or dense fascicular patterns with a blended–merging border to native myometrium. Tumour cells show small, uniform nuclei without nuclear atypia and only few mitoses.¹ While CLM-S and CLM have similar tumour cytology, CLM lacks the characteristic background extracellular matrix and variable architectural patterns seen in CLM-S. Nuclear palisading reminiscent of peripheral nerve sheath tumour Verocay bodies¹² and thick-walled vessels⁹ found in CLM are also common in CLM-S. In contrast, tumour outgrowths termed ‘seedling leiomyomas’ in CLM⁹ were not observed in CLM-S.

We reported that CLM-S is characterised by HGMA2 overexpression and/or genomic alterations, and that MED12 exon 2 mutation or loss of FH expression by IHC do not occur in CLM-S. In comparison, the molecular landscape of conventional CLM is not well understood. Prior studies have shown that CLM has cytogenetic changes involving in t(1;10)¹³ and MED12 mutations in 15% of tumours.¹⁴ A study of 12 uterine CLM demonstrated that two tumours harboured gene fusions involving chromosome 12 at the HMGA2 locus and three tumours contained hyalinisation/collage plaques.¹⁵ A large molecular analysis of HMGA2, MED12 and FH mutations demonstrated that 32% of CLM showed frequent HMGA2 overexpression.¹⁶ However, histological details beyond degree of cellularity are not provided for these tumours.¹⁶ An investigation of 52 CLM found ~37% of tumours to display HMGA2 overexpression by IHC, confirmed by increased mRNA expression in all cases.¹² One pictured tumour included in this study shows morphological features characteristic of CLM-S.¹² Other previous studies demonstrated that CLM may have specific cytogenic alterations¹³ different from usual-type LM and are less likely to show MED12 mutations.¹⁷ Upon review of the literature, many reported CLM with HMGA2 overexpression may represent CLM-S.

We utilised two break-apart probes of HMGA2 for our study, emphasising the large size of the gene and complexity of its aberrations. FISH analysis of HMGA2 revealed detectable abnormalities in almost half of CLM-S, and approximately 66.7% (12 of 18) harboured gene rearrangement at the 5’ UTR. The aetiology of HMGA2 overexpression in our cases without HMGA2 rearrangement by FISH is unknown. The phenomenon that ~96% of CLM-S in our series display HMGA2 overexpression by IHC, yet more than half did not contain a detectable abnormality in HMGA2 by FISH, suggests that other mechanisms potentially lead to HMGA2 overexpression in CLM-S similar to hydropic LM.⁶ Based on recent studies on whole genome sequencing, HMGA2 overexpression can be caused by copy number gain or gene amplification in chromosome 12q15 where HMGA2 is located.^18,19 Prior works have demonstrated that down-regulation of the 3’ UTR for HMGA2 and disrupting tumour suppression by microRNA let-7 contributes to oncogenesis.^20–22 Ultimately, FISH testing and related data/results may be limited for HMGA2, and interpretations should incorporate variables accordingly.

Successfully identifying CLM-S would be helpful for biopsy or myomectomy specimens with limited tissue available for diagnostic work-up. It has been established that HMGA2-LM are usually larger and faster-growing tumours than those without HMGA2 overexpression.²³ Similarly, HMGA2 overexpressing tumours can cause an increased clinical burden on patients who may benefit from surgical intervention over symptomatic treatments.⁶ Resection provides definitive management with no disease recurrence, in keeping with LM’s benign nature.

To our knowledge, our study is the first and largest series to describe CLM-S and HMGA2 overexpression. Our findings are helpful to further delineate this unique subset of CLM. Our data suggest that assessing HMGA2 expression by IHC or performing FISH testing for HMGA2 gene rearrangement may help in the diagnostic evaluation of CLM-S. Further studies are warranted to assess how our investigation’s results apply to CLM-S’s biogenesis, clinical translation and diagnostic reproducibility.

Supplementary Material

supplementary material

Table S1. Summary of clinical, histologic and molecular data in cellular leiomyoma with sclerosis.

Data availability statement

Data available in article supplementary material