SIX1 as a Novel Immunohistochemical Marker in the Differential Diagnosis of Rhabdomyosarcoma

Abstract Background: Differential diagnosis of rhabdomyosarcoma (RMS) is challenging. Sineoculis homeobox homolog 1 (SIX1) is an oncogene involved in skeletal muscle differentiation. We compared protein expression patterns of SIX1 in RMS and its most common differential diagnoses. Methods: SIX1 immunohistochemistry in 36 RMS and in 33 tumors from seven differential diagnostic subtypes were evaluated. The fraction of SIX1 positive tumor cells was scored by three independent observers. Results: A majority (75%) of the evaluated RMS expressed SIX1 in at least 50% of tumor cells and all except one RMS had more than 25% positive tumor cells. Neuroblastoma had less than 1% SIX1 positive tumor cells. Gonadoblastoma, malignant rhabdoid tumor, and Ewing sarcoma had 10% or less positive tumor cells. Pleuropulmonary blastoma exhibited 26–50% positive tumor cells and synovial sarcoma >50% positive cells. Conclusion: SIX1 immunohistochemistry is positive in most RMS, and occasionally in some tumors within the differential diagnoses of RMS.


Introduction
Rhabdomyosarcoma (RMS) is the third most common malignant solid tumor in children outside the CNS, only surpassed in incidence by Wilms tumor and neuroblastoma [1].It accounts for about 3% of all childhood cancers [2], arising in the mesenchymal soft tissues from malignant cells with a propensity toward skeletal myogenic differentiation [3].There are currently four recognized histological subtypes of RMS according to the WHO [4].Embryonal RMS (E-RMS) and alveolar RMS (A-RMS) constitute the two major subtypes, followed by pleomorphic RMS, that are typically uncommon in children [5], and spindle cell/sclerosing RMS, that was previously considered a subtype to E-RMS [6].
Diagnosis of RMS is based on the histopathological appearance of tumor tissue together with immunohistochemical and molecular profiles.Differential diagnosis can pose a challenge, and it is common that several experts are consulted before arriving at a final diagnosis [7].The current gold standard in the differential diagnosis of RMS is to demonstrate the presence of rhabdomyoblasts through the immunohistochemical expression of MyoD1 and myogenin [8].For A-RMS molecular golden standard also necessitates FOXO1-rearrangement for diagnosis.The protein expression of MyoD1 and myogenin can be focal and are consequently unreliable in small tumor samples, such as material from core needle biopsies [9].To complicate matters further, various other malignancies can exhibit differentiation toward skeletal muscle.Among the difficult differential diagnoses are other small round blue cell tumors of childhood that can arise in soft tissues, such as malignant rhabdoid tumor, neuroblastoma, and extra-skeletal Ewing sarcoma.
Even when a diagnosis of RMS is made, the sheer diversity in RMS morphologies can make subtyping difficult when based solely on histopathological examination [8].In this regard, molecular diagnostics have so far come to play a major part, as opposed to conventional immunohistochemistry on its own [3].High resolution molecular analysis is costly, bioinformatically work-intensive and requires high-volume, high-quality tumor samples.Increased immunohistochemical diagnostic possibilities would be advantageous, especially in the setting of sparse diagnostic samples, e.g.core needle biopsies.As there are considerable differences in outcome between RMS subtypes, discrimination is a matter of crucial clinical importance.The sometimes work-intensive differential diagnostic process as well as the multitude of presenting RMS morphologies, underlines a need for novel diagnostic tools.
The Sineoculis homeobox homolog 1 (SIX1) protein is part of the homeobox gene family, has an important role in the expansion of progenitor cells and is involved in many early embryonic processes, such as skeletal muscle differentiation [10].There is some evidence that SIX1 has a master regulatory role in the repression of RMS differentiation [11,12].Using immunofluorescence methodology, SIX1 expression has previously been suggested of potential value in the identification of blastemal elements in Wilms tumor [13].SIX1 protein overexpression has been linked to poor prognosis in various cancers such as gastric adenocarcinoma [14], breast cancer [15], hepatocellular carcinoma [16], pancreatic ductal adenocarcinoma [17], and glioma [18].This suggests that SIX1 overexpression is associated with a dedifferentiated or high-grade tumor state.We recently introduced SIX1 as a marker in our clinical work-up of Wilms tumors.Since then, we have made the incidental observations that also RMS frequently exhibits this marker.In the present study, we explore this phenomenon further.

Materials and methods
The study is part of a larger study of molecular profiles of childhood cancer, approved by the Regional Ethics Board (L289-11, updated 2018).

Study population
The biobank at Lund University/Regional Laboratories, Skåne, Sweden, in which all pediatric cancers diagnosed in the southern healthcare region of Sweden are registered, were searched from January 1998 to January 2023 for diagnoses of "rhabdomyosarcoma", "suspicion of rhabdomyosarcoma", "alveolar rhabdomyosarcoma", "embryonal rhabdomyosarcoma" and "pleomorphic rhabdomyosarcoma".All anatomical regions were included in the search.Cases were filtered to include only individuals below 18 years of age.Consult cases without any locally stored tissue material, revised diagnoses, or unclear subtypes were excluded.Common differential diagnoses were included from the most recently diagnosed cases from 2016 and onwards.In the case of malignant rhabdoid tumor, the tumor's rarity made us include cases dating as far back as 2000.

Immunohistochemistry
For cases that fulfilled the inclusion criteria, tumor samples in the form of formalin-fixed paraffin-embedded tissue blocks were obtained from the biobank.Tissue blocks suitable for immunohistochemical staining were identified through evaluation of archived hematoxylin-eosin (HE) slides, or when these were not available, or of poor condition, freshly made HE slides.Once tissue blocks with tumor tissue were identified, new three-micrometer-thick tissue sections were made and SIX1 immunohistochemistry was performed using a Ventana ultra-instrument.Pretreatment was performed using Ventana CC1-buffer solution (art.nr 950-224) at 100 °C for 64 min.The polyclonal SIX1 Atlas Antibody (art.nrHPA001893) was applied, diluted 1:500, and incubated at 36 °C for 32 min.The detection kit used was the Ventana UltraView Universal DAB Detection Kit (art.nr760-500).Negative and positive control tissues were included on every slide in the form of fetal kidney tissue.
In 13 cases SIX1/pan-cytokeratin stained tissue sections were previously available and were considered sufficient.In these cases, the same SIX1 antibody as detailed above was used with the only difference that it was diluted 1:400 and pretreated for 52 instead of 64 min.

Scoring of SIX1 positive cells
The slides were scanned with NanoZoomer S360 (Hamamatsu, Hamamatsu-city, Japan) and evaluated using the Aperio ImageScope version 12.4.6software (Leica Biosystems, Nussloch, Germany), NDP.view2 version 2.5 software (Hamamatsu, Hamamatsu-city, Japan), or for a few cases Sectra Workstation IDS7 version 24.1 (Sectra, Linköping, Sweden).The fraction of SIX1 positive viable tumor cells was semi-quantitatively determined by two junior observers (H.F. and S.L.), who were blinded to the diagnoses, and by a senior observer (D.E.).Six scoring categories were used as follows: less than 1%, 1-10%, 11-25%, 26-50%, 51-75%, and 75% or more.Scoring was performed autonomously, and results were not shared between observers.A minimum of 200 tumor cells had to be present in a sample for inclusion.Weak brown nuclear staining was considered as positive though cells with overwhelmingly blue nuclear contours were excluded, and cytoplasmic staining was not considered positive.

Statistical analyses
Descriptive statistics and statistical analyses were performed in SPSS Statistics 27 (IBM SPSS Statistics for Windows, Version 27.0.Armonk, NY: IBM Corp) on the compiled data.Cohen's Weighted Kappa (linear weight) was used to calculate the overall agreement between observers.In case of disagreement by one increment between all scorers, the mean value was considered as the consensus score.Cases with discrepancies, of more than one increment in scoring category between observers, were jointly reviewed by all observers and discussed for consensus with an experienced pediatric pathologist (D.G.).

Results
In total, 37 RMS were included for immunohistochemical analysis.One E-RMS tissue block proved to have no remaining tumor material and was subject to a late exclusion.Of the remaining RMS cases there were 19 female and 17 male patients, all were children aged from 0 to 17 years, and the mean age was 7 years.
For evaluation of SIX1 expression, tissue from 36 RMS were available for analysis, and these were distributed as follows: 26 E-RMS, eight A-RMS, one spindle cell RMS, and one DICER1-associated E-RMS.For 31 RMS cases, scoring was performed on whole sectioned excisional biopsies, as opposed to core needle biopsies.In addition, tissue sections from seven differential diagnoses were also assessed for SIX1 expression, including five malignant rhabdoid tumors, 12 neuroblastomas, two gonadoblastomas, one pleuropulmonary blastoma, nine Ewing sarcomas, two congenital mesoblastic nephromas, and two synovial sarcomas.The majority of samples were acquired from primary tumors prior to treatment.In three cases, biopsies had been acquired during ongoing treatment due to tumor progression (the DICER1-associated E-RMS one MRT, and one Ewing sarcoma).Five cases were relapses (one E-RMS, one MRT, two neuroblastomas, and one Ewing sarcoma).

SIX1 expression as compared to MyoD1 and myogenin
MyoD1 or myogenin IHC status was obtained from pathology reports from all except four older cases where it was unavailable.In all except one of the 35 SIX1 positive cases (defined as >25% SIX1 positive tumor cells), either MyoD1 or myogenin was  histological slides from representative non-rMS tumors stained with hematoxylin-eosin (upper) and SiX1 (lower) 20× magnification (inset 40×).(A1-2) a synovial sarcoma that was evaluated as more than 75% positive tumor cells.(B1-2) a malignant rhabdoid tumor that was evaluated as less than 1% positive tumor cells.(C1-2) neuroblastoma that was evaluated as less than 1% positive tumor cells.(D1-2) Pleuropulmonary blastoma that was evaluated as 26-50% positive tumor cells.rMS -rhabdomyosarcoma, SiX1 -sineoculis homeobox homolog 1 considered as positive and had facilitated in establishing the diagnosis according to the gold standard.The SIX1 positive case that had a negative MyoD1 and myogenin IHC status was the DICER1-associated E-RMS.The E-RMS case that was SIX1 negative had a positive MyoD1/myogenin IHC status.

Variation in scoring between observers
Inter-observer variation between evaluators, measured by weighted kappa in pairwise analyses, was in the range of 0.80-0.85(Table 1).Of all the 69 scored cases, only six differed by more than one increment in scoring category between observers and were subsequently jointly reviewed for consensus.

Discussion
In this study we have investigated the staining frequencies of a novel immunohistochemical marker in tissue sections from RMS, a high-grade malignancy of childhood, and in tissue sections from tumors that are challenging to distinguish from RMS by conventional histological staining as they can exhibit a small round blue cell morphology similar to poorly differentiated RMS.Our results indicate that the expression of SIX1 is more pronounced in RMS as a group, compared to the other tumors, with some notable exceptions discussed further below.

SIX1 expression in RMS
We find it most interesting that SIX1 seems to be expressed in RMS to the large extent displayed in this study.The frequency with which RMS tumor cells are SIX1 positive makes SIX1 suitable for identifying tumors lacking in obvious rhabdomyoblastic differentiation, particularly when limited tumor material is available.One RMS stood out as largely devoid of SIX1 expression (<1%).This case was reviewed, and the diagnosis of E-RMS was reconfirmed by performing new immunohistochemistry for MyoD1 and myogenin.The case in question thus constitutes a true outlier.One E-RMS case with more than 50% SIX1 positive tumor cells on excisional biopsy had follow-up biopsies made after adjuvant chemotherapy and brachytherapy.This case exhibited only 1-10% positive tumor cells after therapy (data not shown).On closer examination the tumor cells that were positive for myogenin (<1%) and MyoD1 (about 1%) were also positive for SIX1.The case exhibited notably more SIX1 positive cells than for the former two biomarkers.This suggests that SIX1 might be useful also in the post-adjuvant setting.On this topic, the only case that was considered negative for myogenin and MyoD1 was the DICER1-associated E-RMS, and that tumor had received adjuvant therapy.A recent study by Botiralieva et al. [19], published only in Russian, evaluated 201 RMS for SIX1 protein expression.Here, SIX1 was found to be expressed in 80-95% of RMS tumors.This is remarkable, especially as it reportedly was performed on tissue microarrays and not whole sectioned tissue.The scope of that article was rather to map the immunohistochemical profile of RMS and no comparison with SIX1 expression in differential diagnoses were performed [19].

SIX1 expression as a determinant of RMS subtype
In our limited material and with our method of choice, there was no difference in SIX1 expression between the different RMS subtypes (Figure 1).To determine RMS subtype, molecular diagnostics is likely the way forward and it is already redefining the RMS classification.For instance, most experts already consider that RMS is better classified as FOXO1 fusion-positive RMS and FOXO1 fusion-negative RMS as opposed to A-RMS and E-RMS [3].In addition, spindle-cell/sclerosing RMS have been shown to manifest with recurrent MYOD1 mutations [20].
In the study by Botiralieva et al. [19], SIX1 was found to be expressed in each of the evaluated RMS subtypes.Subtypes accounted for were E-RMS, A-RMS, spindle-cell/ sclerosing and epithelioid RMS (a subtype not currently recognized by the WHO) [19].It can be concluded that SIX1 as a determinant of RMS subtype will most likely not be relevant.

Tumors with low SIX1 expression
Our results indicate that SIX1 expression might be useful in differentiating RMS from neuroblastoma, malignant rhabdoid tumor, Ewing sarcoma, and gonadoblastoma since these four tumor types stained in a lower percentage category.It is suggestive that if only these four tumors are included in the analysis and synovial sarcoma, congenital mesoblastic nephroma, and pleuropulmonary blastoma are excluded, and if a SIX1 expression cutoff of more than 10% positive tumor cells is used, this would yield a specificity of 100% and a sensitivity of 97%.As previously stated, to draw any statistically significant conclusions regarding specificity and sensitivity, a much larger material is needed.
Neuroblastoma stems from neural crest cells [21] and is therefore not expected to express the SIX1 protein from an embryological perspective.All of the evaluated neuroblastomas scored less than 1% positive tumor cells (Figure 3).
Malignant rhabdoid tumor has an unknown cell of origin [22], even though its name implies a mesenchymal origin.Noteworthy, cells with the classical rhabdoid appearance were negative for SIX1 (Figure 4).This could be important as RMS too can exhibit cells with rhabdoid features.Evidently, rhabdoid inclusions are relatively nonspecific and not indicative of skeletal muscle differentiation [23].
Ewing sarcoma is hypothesized to stem from either neural crest cells or mesenchymal stem cell, and there are several lines of evidence to support either [24].Our results indicate that Ewing sarcoma does not express SIX1, suggesting that Ewing sarcoma is not durably dependent on SIX1, which could imply support of a neural crest cell origin.
Gonadoblastoma is derived from undifferentiated gonadal tissue [25] and is therefore not expected to express the SIX1 protein from an embryological perspective as was the case in our study.

Tumors with as high SIX1 expression as RMS
Synovial sarcoma has an early mesenchymal origin [26], thus it is not entirely surprising that this tumor expresses the SIX1 protein.Our results show that SIX1 likely cannot be used to differentiate synovial sarcomas from RMS. Synovial sarcoma can have some expression of myogenin or MyoD1, so neither can differentiate the two diagnoses.In these cases, making a correct diagnosis requires input from other diagnostic tools, such as the common desmin stain which is usually negative in synovial sarcomas as opposed to RMS, although exceptions have been described [27,28].In addition, there are specific gene-fusions coupled with synovial sarcomas that are present in more than 95% of cases [29].
Pleuropulmonary blastomas are also tumors of mesenchymal origin [30].They sometimes exhibit MyoD1 and myogenin [31,32] and in our study the single case included scored for SIX1 expression in one of the middle percentage categories (26-50%), which poses a diagnostic problem.Reassuringly though, RMS in the lung is very uncommon.
Finally, congenital mesoblastic nephroma also has a mesenchymal origin and the two cases evaluated scored positive for SIX1 either in the high or in the low range.The high-scoring sample was of the cellular type, as opposed to the lower scoring sample that was of the classical type.Interestingly, the cellular subtype seems to be associated with worse outcome [33].This could again suggest that SIX1 overexpression is associated with a high-grade tumor state.

Study strength and limitations
As we are dealing with such an unusual tumor group, the number of RMS examined in this exploratory study is actually to be considered a strength.At the same time, to draw any statistically significant conclusions regarding specificity and sensitivity of the methodology, a much larger study material is needed.In particular, it is important to confirm whether the strong expression in synovial sarcoma is a recurrent characteristic in this tumor type.The moderate staining in pleuropulmonary blastoma and congenital mesoblastic nephroma calls for further investigation, particularly to assess whether there are cases of these tumors that can exhibit an expression level of a magnitude similar to RMS.To solve this in the setting of a prospective study, the infrequency of some of these tumors warrant collection of samples from multiple centers.
As far as the choice of measurement method is concerned, there is a clear element of subjectivity associated with semi-quantitative methods.It would have been more desirable and objective to let a computer determine the proportion of positively staining tumor cells.Artificial intelligence needs to be trained and requires resources far beyond the means of this modest study.Semi-quantitative measurement methods are established practice both in the clinic and in this field of research and are satisfactory in answering our exploratory questions.That we had three independent observers with low inter-rater variability demonstrates the stability of the biomarker and reproducibility of results.As two of the observers were unfamiliar with diagnosing childhood tumors they were blinded to diagnoses and confirmation bias was effectively diminished.
Another important aspect to consider is that two different softwares, Aperio ImageScope version 12.4.6(Leica Biosystems) or NDP.view2 version 2.5 (Hamamatsu), were used and also different hardware, laptops or stationary computers.This was not ideal, although it simulates a real-world setting, where cases would be distributed to different consulting pathologist with their own diagnostic setups.Importantly, despite these differences in both hard-and software, the inter-observer variability was still low, indicating reproducibility of results.
What is also noteworthy is that some of the specimens were more than 20 years old.It is known that antigen retrieval from seasoned tissue blocks could benefit from more thorough processing.We prepared all tissues through the same protocol and there was low correlation between case age and the percentage of stained tumor cells for all tumors (Spearman's Rank correlation coefficient 0.46, (p < 0.01).Arguably this indicates biomarker stability.For the RMS there was no correlation to case age (Spearman's Rank correlation coefficient −0.13 although no evidence provided (p = 0.44) due to sample size.Finally, it should be noted that the used antibody, although directed toward SIX1, may cross-react with other related SIX1 proteins, which were not the focus of this study.

Conclusion
Our limited study indicates that SIX1 immunohistochemistry is often highly expressed in RMS as compared to some tumors that traditionally constitute the differential diagnoses of RMS.Neuroblastoma, malignant rhabdoid tumor, Ewing sarcoma and gonadoblastoma do not express or exhibit very low expression of the SIX1 protein.
Our findings are limited by the sample size, but merit further study for validation, preferably in a prospective cohort-like fashion.

Figure 3 .
Figure 3. SiX1 expression levels in differential diagnoses to rhabdomyosarcoma. in these boxplots the variation of expression across the six scoring categories are shown for all tumors.rhabdomyosarcomas (rMS) are for simplicity shown as a single group.the number of cases of each tumor type is given in parentheses.the boxes indicate the spread of the middle 50% of scores for each tumor group.the lines that divide the boxes into two parts is the median value.circles are outliers 1.5 times the interquartile range (iQr) outside the boxes.asterisks are extreme outliers 3 times the iQr outside the boxes.