Low grade endometrial stromal sarcoma presents a diagnostic challenge due to its significant morphological overlap with other spindle cell mesenchymal tumors of the uterus. Recent studies re-evaluating the pathological diagnoses of uterine mesenchymal tumors have reported that LG-ESS can be misdiagnosed in up to 20% of cases [22]. Current knowledge about the immunohistochemical profile of LG-ESS is derived from studies with a relatively low number of examined cases, with the largest immunohistochemically analyzed series including fewer than 50 cases [23]. Much of the published data comes from single-case studies and small antibody panels, limiting the availability of robust data. Additionally, LG-ESS is burdened with a relatively high interobserver variability, as some cases can only be reliably diagnosed using molecular testing. This introduces bias into older data, particularly due to evolving terminology and definitions, as some older studies only use the term “endometrial stromal sarcoma,” which makes the comparison of these results problematic. Given these limitations, our aim was to provide reliable data from a carefully selected, molecularly examined series of LG-ESS, the largest series described in the literature for IHC profiling to date. Reliable data on the immunoprofile of LG-ESS are especially critical in routine practice, particularly for cases with equivocal features or cases where only limited tumor tissue is available (e.g., small diagnostic biopsies or tru-cut biopsies).

The main IHC markers for diagnosing endometrial stromal tumors are CD10 and the more recently introduced marker IFITM1 [24, 25]. The expression of CD10 has long been established as one of the key diagnostic markers, as it is sensitive (75–100%) and has been reported in 258/294 (88%) of all cases of LG-ESS with the available IHC results to date. However, CD10 is not highly specific for LG-ESS, as a small portion of LG-ESS can be negative, especially in cases with poor fixation [26]. CD10 expression is also reported in up to 30% of smooth muscle tumors (including cellular leiomyoma), as well as in PEComa and the sarcomatous component of adenosarcoma [10, 27,28,29,30].

Currently, there is limited data on the expression of IFITM1 in LG-ESS. Two studies report IFITM1 expression in 83% (10/12) and 100% (16/16) of LG-ESS cases, compared with 30% (6/20) and 40% (12/30) of smooth muscle tumors [25, 31]. Both studies concluded that IFITM1 specificity (70% and 86,7%) surpasses CD10, although each study examined a relatively small series of cases. IFITM1 expression, while not entirely restricted to LG-ESS, tends to be weak and focal in smooth muscle tumors. Zhao et al. also emphasized IFITM1’s superior sensitivity and specificity over CD10, although also based on a small series [32]. Our study found IFITM1 expression in 98/143 cases (69%), mostly extensive or diffuse with variable intensity. IFITM1 seems to outperform CD10 in distinguishing LG-ESS from smooth muscle tumors; however, larger studies are still needed to confirm its utility. A useful approach would therefore include using both CD10 and IFITM1 as a part of the IHC panel. Interestingly, our study found that both CD10 and IFITM1 were significantly more commonly expressed in LG-ESS cases with a recurrent fusion, which underscores the diagnostic challenges of fusion-negative tumors given the lack of recurrent genetic alterations and potentially more equivocal IHC results.

Hormone receptors, specifically ER (ERα) and PR, are another crucial part of the diagnostic panel. The available literature indicates that ER is expressed in 84,5% of LG-ESS, with PR expression reaching 87% [15, 23, 33,34,35,36,37,38]. Most studies report more extensive, stronger PR expression than ER, which is consistent with our findings. The expression of androgen receptor (AR) has also been described in a high percentage of LG-ESS, though the number of studies is limited [35]. While ER and PR are typically less useful for distinguishing smooth muscle neoplasms, they can aid in ruling out other non-Müllerian origin tumors, particularly for extrauterine LG-ESS.

Hormone marker expression (especially ER and PR) may also have predictive significance, with their high expression rates in LG-ESS supporting hormone therapy such as high-dose progestins, aromatase inhibitors, or GnRH analogues as a viable treatment option [39,40,41]. A recent meta-analysis found that adjuvant hormone therapy can reduce recurrence risk in patients with FIGO stages I–II disease, but with no benefit concerning overall survival [42]. Another study suggested that hormone therapy can also reduce recurrence risk even in stages II–IV LG-ESS, recommending 12 months of high-dose progestin treatment post-surgery [43]. While the effectiveness of hormone therapy is debated, reporting tumor hormone status remains essential as it can help guide treatment decisions. Conflicting evidence also exists regarding ER/PR expression as a prognostic factor, requiring validation through a molecularly confirmed series [35, 43, 44].

In routine practice, one of the common diagnostic pitfalls lies in differentiating cellular leiomyoma (CL) from LG-ESS. The correct diagnosis of CL versus LG-ESS is of extreme clinical importance, given the different biological natures and behavior of these tumors. In which case, a combination of smooth muscle markers such as α-SMA, desmin, h-caldesmon, calponin, transgelin, and smoothelin, and endometrial stromal markers, such as CD10 and IFITM1, is essential. Smooth muscle markers are frequently positive in LG-ESS, especially in cases with smooth muscle differentiation [10, 12, 28, 30, 45, 46]. In contrast, endometrial stromal markers can be expressed in smooth muscle tumors, especially cellular leiomyomas. In our previous study on CL, the expression of CD10 was seen in 65% of cases, and the expression of IFITM1 in 36.5% of cases [13]. This can lead to a potential misclassification if an immunohistochemical (IHC) profile is not rigorously interpreted within the morphological context. The most expressed smooth muscle marker in LG-ESS is α-SMA (50% of cases reported in literature), closely followed by desmin (47%). While relatively common, the expression of all smooth muscle markers tends to be mostly focal and weak.

Based on the literature, h-caldesmon has emerged as the most specific marker for distinguishing smooth muscle tumors from LG-ESS [30, 46, 47]. The published data shows the expression of h-caldesmon in 9% of LG-ESS, and in our study, h-caldesmon was also the least expressed (14%) of the traditionally examined smooth muscle markers [10, 15, 27, 30, 32, 38, 46, 48,49,50]. Transgelin, an actin-binding protein of the calponin family, has similarly been noted as a highly sensitive and specific marker of smooth muscle differentiation. However, its expression in LG-ESS has been examined in just two studies, with a combined total of 32 cases [47, 51]. Our findings showed a 22% expression rate in LG-ESS, notably higher than previously published data, likely due to our larger series size. These findings highlight the need to validate the “characteristic” immunoprofile of LG-ESS to support the accurate diagnosis of challenging cases encountered in routine practice.

In our series, 14–44% of LG-ESS cases exhibited smooth muscle marker expression, underscoring the need for more specific antibodies. One such emerging marker could be smoothelin, initially identified in 1996 by van der Loop et al., which is a cytoskeletal protein primarily found in fully differentiated, contractile smooth muscle cells [52, 53]. This sets it apart from other smooth muscle markers such as calponin, α-SMA, smooth muscle myosin, and h-caldesmon, which are also present in less differentiated, proliferative smooth muscle cells. Additionally, smoothelin’s expression is notably lacking in cells that exhibit smooth muscle-like characteristics, such as myofibroblasts, myoepithelial cells, and striated muscle cells—cells that frequently express other smooth muscle proteins. Studies examining smoothelin’s utility in smooth muscle tumors of the gastrointestinal tract, uterus, and other soft tissues have indicated that cytoplasmic expression is highly sensitive and specific for benign leiomyomas [54, 55]. Notably, it has been suggested that the staining location (cytoplasmic vs. nuclear) may vary based on the biological behavior of smooth muscle tumors, as aberrant nuclear expression has been reported in a subset of leiomyosarcomas and occasionally in GIST. Our study is the first to examine smoothelin expression in LG-ESS, where we observed rare, moderate nuclear expression in one case and weak cytoplasmic expression in another. This suggests that smoothelin could be an extremely valuable marker in cases with equivocal results from the traditional smooth muscle markers. In our previous study, its expression was found in 61.5% of cellular leiomyomas, but it was commonly focal and weak, which limited its practical use [13].

A key takeaway from our findings is the awareness of smooth muscle marker expression in a subset of LG-ESS and the importance of interpreting results with caution, as they should always be evaluated together with endometrial stromal markers. When differentiating smooth muscle tumors from LG-ESS, a combined panel of high-sensitivity markers with specific markers (such as h-caldesmon, transgelin, and, potentially, smoothelin) is recommended, keeping in mind that α-SMA and desmin in particular do not serve as good discriminators between endometrial stromal and smooth muscle lineage. The staining’s extent and intensity are also essential; tumors with diffuse, strong expression of multiple smooth muscle markers should be, in a proper morphological context, classified as smooth muscle tumors, even if there is a focal expression of CD10 or IFITM1.

Of the other markers useful in differential diagnosis, cyclin D1 and BCOR are often part of the panel used for distinguishing LG-ESS from HG-ESS. Although HG-ESS typically exhibits a distinct morphology, rare cases of LG-ESS can present with epithelioid, round cell morphology which may lead to the consideration of HG-ESS. The reported rate of cyclin D1 expression in LG-ESS is generally low, reaching 28%. Our results indicate a relatively high proportion of LG-ESS with positive cyclin D1 expression (56%), though this expression was typically occasional or focal, and weak in intensity. Although a diffuse expression of cyclin D1 was seen in 14/141 (10%) cases, strong and diffuse cyclin D1 expression was present in only a single case (which showed a usual morphological pattern for LG-ESS). BCOR expression was significantly rarer (5%, 7/142) but, when present, the extent of staining ranged from occasional to nearly diffuse (albeit weak). Similarly, the expression of BCORL1, which can also be present in some cases of HG-ESS, was seen in 24% (34/141) of LG-ESS, where it was also only rare and weak. None of the cases with BCORL1 expression harbored a fusion involving the BCORL1 gene. While BCOR and BCORL1 expression likely represent non-specific staining insufficient for HG-ESS diagnosis, they could pose a diagnostic challenge in certain ambiguous cases.

One of the less common differential diagnoses of LG-ESS is GIST, especially when found in extrauterine locations. In which case, a diagnostic panel of CD117 combined with CD10, ER, and PR is most effective, as CD34 can be positive in both LG-ESS and GIST [56]. Although some reports indicate a small portion of LG-ESS exhibit CD117 expression, the staining tends to be weak and focal [57, 58]. Consistent with prior reports, only a single case from our LG-ESS showed occasional, weak expression of CD117.

Another very important differential diagnosis is the entity of KAT6B/A::KANSL1 fused sarcomas, described in 2022 by Agaimy et al. [16, 16, 17, 19]. According to the current knowledge, due to the overlapping features between endometrial stromal and smooth muscle differentiation, sarcomas with the KAT6B/A::KANSL1 fusion cannot be diagnosed based only on the morphological and immunohistochemical features, and molecular testing is needed. The correct diagnosis is important in these cases as, despite their usually bland morphology, these tumors have propensity for aggressive behavior.

There are limitations to our study. The main one being the use of TMA, which brings the risk of underestimating/overestimating the IHC scoring. However, this risk was reduced by using two cores from each tumor to increase the amount of tissue, and this approach is widely used in the literature and allows for the examination of several markers on large sample collections. Additionally, a small portion of cases (n = 14) lack a known fusion status (as they could not undergo NGS RNA analysis due to insufficient amount or quality of the material), preventing molecular confirmation. Due to the low number of rare morphological variants, it was impossible to analyze the differences in expression between these and cases with conventional morphology.