Osteosarcomas (OS), the most common primary malignant bone tumours, are classified as low-grade (characterised by MDM2 amplification) or high-grade (with complex karyotypes). Accurate diagnosis is essential for treatment and prognosis. This study evaluates pre-analytical variables associated with the success or failure of epigenetic analyses in osteosarcoma samples and proposes a standardised preparation protocol.
MethodsRetrospective cohort study of adult patients with OS diagnosed at our sarcoma reference centre (CSUR) over the past 20 years. Pre-analytical variables: year of diagnosis, histological subtype, tissue type, site of origin, sample type (core needle biopsy or surgical specimen with or without chemotherapy), decalcification method (none, EDTA, or nitric acid), and FISH availability. Five 5-μm sections were obtained from each paraffin block. DNA methylation profiling was performed using the Infinium MethylationEPIC v2.0 platform (Illumina). Univariate and multivariate analyses were performed to identify failure predictors.
ResultsA total of 103 samples from 79 patients were analysed: 58 conventional OS, 14 extraskeletal, 24 parosteal, and 7 dedifferentiated OS. Of the 95 formalin-fixed, paraffin-embedded (FFPE) samples, 43 (45.2%) were suitable for epigenetic analysis, whereas all frozen samples were adequate (100%). Decalcification affected success rates, although not significantly: nitric acid was associated with the highest failure rate (68.97%), followed by EDTA (57.14%) and non-decalcified samples (46.15%).
ConclusionFFPE samples are suitable for epigenetic studies, although performance depends on pre-analytical factors. Frozen tissue remains the gold standard. Nitric acid should be avoided. A protocol is proposed that prioritises frozen tissue, documents decalcification methods, excludes strong acids, incorporates quality control measures, and favours samples less than five years old.
Los osteosarcomas (OS), los tumores óseos malignos primarios más frecuentes, se clasifican en: bajo grado (caracterizados por amplificación de MDM2), y alto grado (cariotipos complejos). Un diagnóstico preciso determina el pronóstico y el tratamiento. Este trabajo evalúa qué variables preanalíticas se asocian al éxito o al fallo de los análisis epigenéticos en estas muestras y propone un protocolo normalizado de preparación.
MétodosCohorte retrospectiva de OS en adultos de nuestro centro de referencia para sarcomas (CSUR) durante los últimos 20 años. Variables preanalíticas: año, subtipo histológico, tipo de tejido, origen, tipo de muestra (BAG o pieza quirúrgica con o sin quimioterapia), agente decalcificante (sin decalcificación, EDTA o ácido nítrico) y FISH. Se analizaron cinco secciones de 5μm de bloques de parafina. El perfil de metilación se realizó mediante la plataforma Infinium MethylationEPIC v2.0 (Illumina), con análisis univariantes y multivariantes.
ResultadosSe analizaron 103 muestras de 79 pacientes: 58 OS convencionales, 14 extraesqueléticos, 24 parostales y 7 desdiferenciados. De las 95 muestras FFPE, 43 (45,2%) fueron válidas para el estudio epigenético; las restantes correspondieron a muestras congeladas (todas aptas, 100%). La decalcificación influyó, aunque no de forma significativa: los casos tratados con ácido nítrico mostraron la mayor tasa de fallo (68,97%), seguidos de EDTA (57,14%) y de los no decalcificados (46,15%).
ConclusiónLas muestras FFPE son útiles, pero su rendimiento depende de las variables preanalíticas. Las muestras congeladas se consideran el método de preferencia por su mejor rendimiento. Debe evitarse el uso de ácido nítrico. Protocolo propuesto: uso de tejido congelado siempre que sea posible; documentación del agente decalcificante y exclusión de ácidos fuertes; establecimiento de controles de calidad (CC); e inclusión de casos<5 años de antigüedad.
Osteosarcoma (OS) is the most common primary malignant bone tumour across all age groups. Approximately 60% of patients are between 10 and 20 years of age, with OS being the second leading cause of death in this age group.1 Despite its low incidence, osteosarcoma is associated with a high disability rate and a 5-year survival rate of less than 60% in metastatic disease.
Osteosarcomas are classified into high- and low-grade tumours, with specific subtypes defined according to their morphology and anatomical location. Among high-grade lesions are conventional intramedullary OS, as well as periosteal and high-grade surface OS arising on the bone surface. Low-grade OS comprise two recognised subtypes: low-grade central (intramedullary) OS and parosteal OS,2 which occurs on the bone surface.
On the one hand, low-grade OS, irrespective of their location, exhibit similar morphology, characterised by a variable amount of parallel bony trabeculae and a paucicellular fibroblastic stroma, together with amplification of the Murine Double Minute Clone 2 (MDM2) oncogene.3MDM2 is an oncogene located on the long arm of chromosome 12, at 12q15, which encodes an oncoprotein that negatively regulates p53. As p53, one of the genes most frequently mutated in tumours, is responsible for activating genes involved in cell cycle arrest and apoptosis, overexpression of MDM2 limits this tumour-suppressive function.4
Parosteal osteosarcoma accounts for 4–5% of all osteosarcomas and is the most common surface osteosarcoma. Between 15% and 43% of low-grade osteosarcomas may undergo dedifferentiation, transforming into a high-grade osteosarcoma,1 termed dedifferentiated OS, which retains MDM2 amplification. Low-grade central osteosarcoma represents only 1–2% of OS cases and, to a lesser extent, may also undergo dedifferentiation. Both tumour types harbour ring chromosomes or giant marker chromosomes with amplified regions at 12q13–15, including the MDM2 and CDK4 genes,5,6 which are used to support the diagnosis.6 In a high proportion of cases, immunohistochemical expression of MDM2 and/or CDK4 can also be demonstrated. In this regard, the detection of MDM2 amplification/overexpression may be particularly useful in distinguishing these low-grade sarcomas from their often benign mimics.7
There is limited literature addressing this process of dedifferentiation and its clinical management. Studies from the Rizzoli Institute suggest that patients with dedifferentiated parosteal osteosarcoma have a relatively favourable prognosis; however, it remains unclear whether this difference is attributable to tumour-related or patient-related factors.8 Furthermore, as dedifferentiated chondrosarcomas exhibit a very poor response to chemotherapy,9 the degree of dedifferentiation in osteosarcomas may likewise influence treatment response.
Conversely, conventional osteosarcomas exhibit marked genomic complexity and instability, resulting in significant tumour heterogeneity. They are characterised by a complex karyotype, with chromosomal complexity, aneuploidy and structural rearrangements, few recurrent genetic alterations, and an absence of MDM2 gene amplification.10,11
For decades, the application of molecular alterations as biomarkers for the early diagnosis and prognosis of osteosarcoma has been extensively investigated. Genetic analyses have identified a range of mutations associated with this bone tumour; however, the findings to date remain inconclusive, as the complexity of the disease cannot be explained by a single molecular alteration or solely from a genetic perspective. The aetiology of osteosarcoma involves the interplay of multiple levels of gene expression regulation. Accordingly, in addition to genetic alterations, it is essential to consider regulatory mechanisms such as epigenetic factors. Epigenetics is defined as the set of mechanisms that regulate DNA without modifying its sequence and comprises flexible and reversible events with a substantial impact on tumorigenesis. In this context, DNA methylation plays a pivotal role, as it exhibits considerable variation in cancer cells and may be involved in processes recognised as the hallmarks of human cancer. Thus, the study of epigenetic factors, particularly DNA methylation, holds significant promise as a biomarker with clinical applications in diagnosis, prognosis, and response to chemotherapy.12
There is considerable variability in the acquisition and processing of biopsy specimens, which directly affects the suitability of the material for diagnostic, molecular and research applications. International clinical guidelines for bone sarcoma recommend the routine collection of fresh and frozen tissue for clinical molecular studies.13,14 Accordingly, the aim of this study is to review clinical practices in sample processing that influence subsequent clinical and research analyses, with a view to establishing optimised and standardised protocols.
Material and methodsA retrospective study was conducted including 79 adult patients diagnosed with OS and discussed at the Musculoskeletal Tumour Board of our institution over the past 20 years.
The cohort comprised patients with conventional, low-grade intramedullary, parosteal and dedifferentiated osteosarcomas, with a total of 103 samples (95 formalin-fixed paraffin-embedded [FFPE] and 8 frozen). For each sample, detailed metadata were collected, including year of acquisition, histological subtype, tissue type, anatomical location, sample type (core needle biopsy [CNB], resection specimen with or without prior chemotherapy), decalcification method (no decalcification, EDTA decalcification, or nitric acid decalcification), and the availability and result of FISH analysis. Decalcification methods were not prioritised due to the retrospective nature of the study; however, in more recent cases, the recommendations of the SEAP bone and soft tissue working group have been followed, favouring EDTA and formic acid and avoiding strong acids that compromise DNA integrity.15
For epigenomic preparation and analysis, the two types of samples were processed as follows. For frozen samples, DNA extraction was performed using column-based methods (E.Z.N.A. DNA Kit and DNeasy Blood & Tissue Kit, Qiagen, Hilden, Germany), in accordance with the manufacturer's instructions.
For formalin-fixed, paraffin-embedded (FFPE) samples, five 5-μm sections were cut from each block for automated extraction of genomic DNA using the 405 MagCore Genomic DNA One-Step Kit on the HF-16 biorobot (RBC Bioscience).
DNA samples obtained from frozen tissue were treated with RNase A for 1h at 45°C, quantified by fluorometry (Quant-iT PicoGreen dsDNA Assay, Life Technologies, CA, USA), and their purity assessed using a NanoDrop spectrophotometer (Thermo Scientific, MA, USA) based on the 260/280 and 260/230 ratios. DNA integrity in fresh samples was evaluated by electrophoresis on a 1.3% agarose gel.
For methylation analysis, all samples were processed using the Infinium DNA MethylationEPIC v2.0 BeadChip array with the iSCAN system (Illumina Inc., San Diego, CA, USA), launched in January 2023, which enables the detection of more than 950,000 CpG sites.12
A total of 600ng of purified DNA (for fresh tissue samples) was randomly distributed into 96-well plates and processed using the EZ DNA Methylation Lightning kit (Zymo Research Corp., CA, USA), in accordance with the manufacturer's recommendations. Bisulphite-converted DNA (bsDNA) was processed following the Infinium HD methylation assay protocol, compatible with the HumanMethylation 850K and 450K platforms, thereby ensuring reproducibility and reliability in the detection of epigenetic alterations.12,16 Additionally, FFPE-derived samples underwent a further restoration step using the Infinium FFPE DNA Restoration kit (Illumina Inc., San Diego, CA, USA).
Methylation data were analysed using the minfi package from Bioconductor. All sample data were processed in R, using the IlluminaHumanMethylationEPICv2anno.20a1.hg38 annotation package for probe annotation. For sample quality control, the mean detection p-value across all probes was calculated for each sample. Samples with a mean detection p-value greater than 0.05 were excluded from the analysis.
Ethical considerationsThe study was conducted in accordance with the principles of the Declaration of Helsinki and current regulations governing biomedical research. The protocol was reviewed and approved by the Research Ethics Committee for Medicinal Products, with reference number 2023-1219-1. As the study was based on previously archived biological samples and anonymised clinical data, no additional informed consent was required.
Statistical analysisData were analysed using R (version 4.4). A descriptive statistical analysis was performed on the cohort of 95 paraffin-embedded samples in order to assess the availability of high-quality biological material suitable for epigenetic analysis according to different clinical and sample-related variables. Qualitative variables were summarised as absolute frequencies (n) and percentages (%). Associations between categorical variables were evaluated using Pearson's Chi-squared test. A p-value<0.05 was considered statistically significant. Subsequently, multivariate logistic regression models were fitted, incorporating variable selection and regularisation strategies, with the aim of identifying the factors with the greatest predictive value.
ResultsA total of 103 OS samples were included: 58 conventional, 14 extraskeletal, 24 parosteal, and 7 dedifferentiated. These samples were obtained from 79 patients: 65 patients provided a single sample, corresponding to the primary tumour in 59 cases, recurrence in 3 cases, and metastasis in 3 cases. Fourteen patients contributed multiple samples, totalling 38 specimens, including primary tumours, recurrences, and metastases. The clinical and pathological characteristics of the cohort are summarised in Table 1, which shows a sex-balanced population with a predominance of conventional osteosarcoma and a majority of tumours located in the extremities, particularly the lower limbs.
Clinical characteristics of the study population.
| Variables | N=79 | % |
|---|---|---|
| Age at diagnosis | ||
| Mean | 48.62 | – |
| Range | (11–90) | – |
| Sex | ||
| Male | 38 | 48.10 |
| Female | 41 | 51.20 |
| Histology | ||
| Conventional osteosarcoma | 49 | 62.02 |
| Low-grade osteosarcoma | 13 | 16.46 |
| Extraskeletal osteosarcoma | 14 | 17.72 |
| Dedifferentiated osteosarcoma | 3 | 3.80 |
| Primary site | ||
| Lower limb | 46 | 58.22 |
| Upper limb | 12 | 15.19 |
| Trunk | 8 | 10.13 |
| Pelvis | 7 | 8.86 |
| Retroperitoneum | 3 | 3.80 |
| Maxillofacial region | 3 | 3.80 |
| Metastasis at diagnosis | ||
| Yes | 14 | 17.72 |
| No | 65 | 82.28 |
| Site of metastasis | ||
| Lung | 12 | 85.72 |
| Lung and pleura | 1 | 7.14 |
| Lung, L2 vertebra and ribs | 1 | 7.14 |
| Survival | ||
| Alive | 42 | 53.17 |
| Dead | 32 | 40.51 |
| NA | 5 | 6.32 |
Of the total samples analysed, all 8 fresh-frozen samples were suitable for epigenetic analysis (100%), whereas among 95 FFPE samples, only 43 (45.2%) met the required quality criteria for methylation analysis. These initial differences highlight the superiority of fresh-frozen tissue over formalin-fixed paraffin-embedded (FFPE) material.
Table 2 presents the descriptive analysis and the association between key clinical and pre-analytical variables and the outcome of methylation analysis in FFPE samples. A significant association was observed with the year of sample collection, with a higher success rate in more recent samples. Histology also had a relevant impact, with a higher failure rate in low-grade osteosarcomas. Tissue type showed a significant effect as well, with soft tissue-derived samples more frequently meeting quality criteria than those of bone origin. Regarding tumour source, metastatic samples showed a higher success rate compared with primary tumours. Representative examples are shown in Fig. 1.
Descriptive analysis and association assessment using Fisher's exact test.
| Variables | Methylation | Fisher's exact test | ||
|---|---|---|---|---|
| Ok | Failure | |||
| PARAFFINN=95 | ||||
| Year | 2004 | 1 (100.00%) | 0 (0%) | p value: 0.0129 |
| 2005 | 1 (100.00%) | 0 (0%) | ||
| 2006 | 0 (0%) | 1 (100.00%) | ||
| 2009 | 0 (0%) | 2 (100.00%) | ||
| 2010 | 2 (33.33%) | 4 (66.67%) | ||
| 2011 | 5 (41.67%) | 7 (58.33%) | ||
| 2012 | 0 (0%) | 6 (100.00%) | ||
| 2013 | 2 (50.00%) | 2 (50.00%) | ||
| 2014 | 1 (20.00%) | 4 (80.00%) | ||
| 2015 | 3 (50.00%) | 3 (50.00%) | ||
| 2016 | 4 (50.00%) | 4 (50.00%) | ||
| 2017 | 3 (37.50%) | 5 (62.50%) | ||
| 2018 | 1 (25.00%) | 3 (75.00%) | ||
| 2019 | 0 (0%) | 4 (100.00%) | ||
| 2020 | 2 (40.00%) | 3 (60.00%) | ||
| 2021 | 2 (66.67%) | 1 (33.33%) | ||
| 2022 | 3 (75.00%) | 1 (25.00%) | ||
| 2023 | 5 (100.00%) | 0 (0%) | ||
| 2024 | 8 (88.89%) | 1 (11.11%) | ||
| 2025 | 0 (0%) | 1 (100.00%) | ||
| Diagnosis | Conventional osteosarcoma | 28 (52.83%) | 25 (47.17%) | p value: 0.0056 |
| Dedifferentiated osteosarcoma | 3 (75.00%) | 1 (25.00%) | ||
| Low-grade osteosarcoma | 4 (16.67%) | 20 (83.33%) | ||
| Extraskeletal osteosarcoma | 8 (57.14%) | 6 (42.86%) | ||
| Tissue | Bone | 28 (38.36%) | 45 (61.64%) | p value: 0.0162 |
| Soft tissue | 15 (68.18%) | 7 (31.82%) | ||
| Bone type | Long | 21 (36.84%) | 36 (63.16%) | p value: 0.7720 |
| Flat | 7 (43.75%) | 9 (56.25%) | ||
| NA | 15 | 7 | ||
| Origin | Primary | 34 (43.59%) | 44 (56.41%) | p value: 0.0045 |
| Recurrence | 1 (12.50%) | 7 (87.50%) | ||
| Metastasis | 8 (88.89%) | 1 (11.11%) | ||
| Presentation at diagnosis | Localised | 37 (44.05%) | 47 (55.95%) | p value: 0.5376 |
| Metastatic | 6 (54.55%) | 5 (45.45%) | ||
| Decalcification method | No decalcification | 28 (53.85%) | 24 (46.15%) | p value: 0.1352 |
| EDTA | 6 (42.86%) | 8 (57.14%) | ||
| Nitric | 9 (31.03%) | 20 (68.97%) | ||
| Sample type | CNB | 21 (48.84%) | 22 (51.16%) | p value: 0.8240 |
| Surgical specimen with adjuvant treatment | 2 (50.00%) | 2 (50.00%) | ||
| Surgical specimen without adjuvant treatment | 20 (41.67%) | 28 (58.33%) | ||
| FISH | Yes | 13 (65.00%) | 7 (35.00%) | p value: 0.0379 |
| No | 8 (32.00%) | 17 (68.00%) | ||
| NA | 22 | 28 | ||
| FROZENN=8 | ||||
| Diagnosis | Conventional osteosarcoma | 5 | 0 | – |
| Dedifferentiated osteosarcoma | 3 | 0 | ||
Morphological features of cases successfully analysed (D, E, F) and cases that did not meet analysis criteria (A, B, C). A, Parosteal OS decalcified in nitric acid. B, Parosteal OS decalcified in EDTA. C, Conventional OS, non-decalcified, from 2016. D, Pulmonary metastasis of conventional OS. E, Extraskeletal OS, non-decalcified. F, Conventional OS, non-decalcified.
Regarding technical factors, the decalcification method had a clear impact on sample performance, although the difference between the two methods was not statistically significant. As shown in Table 2, samples decalcified with nitric acid exhibited the highest failure rate (68.9%), followed by those decalcified with EDTA (57.1%), whereas non-decalcified samples yielded better results. No relevant differences were observed according to sample type (core needle biopsy versus surgical specimen). The availability of a previous FISH study was associated with a higher probability of successful epigenetic analysis.
Table 3 presents the results of the univariate analysis, expressed as odds ratios with their 95% confidence intervals. In this analysis, the year of diagnosis showed an inverse association with failure of methylation analysis, with an approximate 11% reduction in the probability of failure per year (OR=0.89; p=0.009). Low-grade osteosarcoma histology was significantly associated with a higher risk of failure, approximately 5.6 times greater than that observed in high-grade osteosarcomas (p=0.005). By contrast, samples derived from soft tissue showed a marked reduction in the probability of failure, of around 70% compared with bone tissue (OR=0.29; p=0.017), and tumours of metastatic origin showed an even stronger protective effect, with an approximately 90% reduction in comparison with primary tumours (p=0.031). The use of nitric acid as a decalcifying agent showed a trend towards a higher risk of failure, with an association approaching statistical significance and an approximately 2.6-fold increase in the probability of failure (p=0.051). The remaining variables analysed did not show relevant associations.
Univariate model.
| Level 1 | or | CI_Low | CI_High | p-Value |
|---|---|---|---|---|
| Year | 0.890 | 0.812 | 0.969 | 0.009 |
| Histology (dedifferentiated osteosarcoma) | 0.373 | 0.018 | 3.129 | 0.407 |
| Histology (low-grade osteosarcoma) | 5.600 | 1.826 | 21.272 | 0.005 |
| Histology (extraskeletal osteosarcoma) | 0.840 | 0.246 | 2.747 | 0.774 |
| Tissue (soft tissue) | 0.290 | 0.099 | 0.777 | 0.017 |
| Bone type (flat) | 0.750 | 0.243 | 2.376 | 0.616 |
| Origin (recurrence) | 5.409 | 0.902 | 103.644 | 0.123 |
| Origin (metastasis) | 0.097 | 0.005 | 0.563 | 0.031 |
| Presentation at diagnosis (metastatic) | 0.656 | 0.177 | 2.341 | 0.513 |
| Decalcification method (edta) | 1.555 | 0.475 | 5.331 | 0.467 |
| Decalcification method (nitric ac.) | 2.593 | 1.017 | 6.996 | 0.051 |
| Sample (specimen with adjuvant treatment) | 0.955 | 0.107 | 8.553 | 0.965 |
| Sample (specimen without adjuvant treatment) | 1.336 | 0.584 | 3.082 | 0.493 |
| Fish (yes) | 0.253 | 0.069 | 0.852 | 0.031 |
A multivariate logistic regression model was subsequently fitted, including variables with the greatest clinical and statistical relevance. In the full model, low-grade osteosarcoma histology retained its association with a higher probability of failure, whereas the effects of the remaining variables were attenuated (Table 4).
Multivariate logistic regression model.
| Level 1 | OR | ci_low | CI_High | p-Value |
|---|---|---|---|---|
| Year | 0.909 | 0.801 | 1.030 | 0.134 |
| Histology (low-grade osteosarcoma) | 4.210 | 1.050 | 16.873 | 0.042 |
| Histology (other) | 2.299 | 0.399 | 13.231 | 0.351 |
| Tissue (soft tissue) | 0.209 | 0.034 | 1.297 | 0.093 |
| Origin (non-primary) | 0.820 | 0.180 | 3.730 | 0.797 |
| Decalcification method (edta) | 0.834 | 0.199 | 3.499 | 0.804 |
| Decalcification method (nitric ac.) | 0.913 | 0.278 | 2.998 | 0.881 |
| Fish (yes) | 0.672 | 0.128 | 3.522 | 0.638 |
| Fish (unknown) | 1.043 | 0.292 | 3.721 | 0.949 |
Finally, a parsimonious multivariate model was built, with results shown in Table 5. In this model, each additional year since sample collection was associated with an approximate 11% decrease in the relative probability of failure in epigenetic analysis (OR=0.893; 95% CI=0.810–0.985; p=0.023). Regarding histology, low-grade osteosarcomas showed a significantly higher risk of failure, approximately 3.8 times greater than that observed in high-grade osteosarcomas (OR=3.806; 95% CI=1.086–13.342; p=0.037), whereas the grouped “Other” category showed a non-significant increase.
Finally, soft tissue-derived samples were significantly associated with a reduced probability of failure, of approximately 79% compared with bone-derived samples (OR=0.207; 95% CI=0.047–0.905; p=0.036). The remaining variables included in the model did not show significant associations, indicating that they do not contribute meaningfully once these three variables are accounted for. The final model demonstrated a better balance between fit and parsimony, with an AIC of 120.58, lower than that of the full model (130.09), supporting the relevance of these variables as the main determinants of FFPE sample epigenetic performance.
Overall, the results show that, although FFPE samples are potentially useful for epigenetic studies, their performance is variable and critically dependent on several factors. These include sample age, tumour histology—with poorer performance observed in low-grade osteosarcomas—and, particularly, pre-analytical processing conditions, especially the decalcification method, with the use of strong acids being clearly detrimental. By contrast, fresh-frozen tissue provides consistently superior DNA quality and a 100% success rate, establishing it as the reference standard for epigenetic analyses in osteosarcoma.
DiscussionThis study systematically evaluates the impact of clinical, histological, and pre-analytical variables on the performance of DNA methylation analysis in osteosarcoma samples, a key issue given the increasing incorporation of epigenetic tools into the diagnosis and research of bone tumours.
Our results confirm that fresh-frozen tissue remains the reference standard for epigenetic studies, with a 100% success rate, in clear contrast to the performance observed in FFPE samples. This difference likely reflects the cumulative effect of formalin fixation, prolonged storage, and other pre-analytical factors on DNA integrity, which is particularly critical in methylation microarray-based platforms. In this regard, our findings are consistent with those reported by Okojie et al., 2024, which demonstrate that cryopreserved tissues retain significantly higher DNA quality compared with matched FFPE counterparts, supporting the use of frozen material as the standard for genomic and epigenetic studies.17
Among the factors independently associated with epigenetic assay performance, sample age emerged as a key determinant. Each additional year since tissue collection was associated with an approximately 11% reduction in the probability of success, suggesting a cumulative time-dependent effect on DNA quality in FFPE samples. This progressive impact of prolonged storage on DNA integrity has been described in previous studies. On the one hand, Yi et al. observed that FFPE samples stored for longer periods show reduced amounts of retrievable DNA and increased signs of molecular degradation, which may limit both extraction efficiency and downstream analysis.18 On the other hand, a recent study by Huang et al.19 comparatively analysed FFPE samples of different ages and found that older specimens exhibited more pronounced fragmentation, lower PCR amplification efficiency, and poorer performance in sequencing library preparation, supporting the notion that archival time compromises not only DNA yield but also its suitability for high-throughput analyses. Taken together, these observations support our finding that FFPE block age is an independent predictor of technical failure in epigenetic studies.
Tumour histology also showed a relevant influence. In particular, low-grade osteosarcomas presented a significantly higher risk of failure compared with high-grade osteosarcomas. The poorer performance observed in low-grade osteosarcomas can be explained, at least in part, by their histological characteristics. These tumours are classically described as paucicellular, with abundant fibro-osseous matrix and mature bone, and a limited fraction of viable tumour cells.8 This low tumour cellularity reduces the amount of available tumour DNA, thereby compromising the performance of epigenetic analyses in FFPE samples. In addition, these histological subtypes, both in our series and in routine diagnostic practice, are those most likely to require decalcification. When the type of decalcification and the FISH result are included in the analysis of failures, a higher probability of success is observed in samples decalcified with EDTA and showing a positive FISH result. Therefore, we discourage the use of cases decalcified with strong acids in which FISH has failed, i.e. where probe hybridisation has not been achieved.
Similarly, the type of tissue analysed proved to be a determining factor. Samples derived from soft tissue showed clearly superior performance compared with bone samples, which likely reflects the absence of decalcification and better DNA preservation. Consistently, metastatic samples exhibited a lower failure rate, possibly related to higher tumour cellularity and lower bone content compared with primary lesions, together with shorter formalin fixation times. However, fixation time was not included in the statistical analysis due to the lack of precise data in all cases, particularly in the older specimens.
The decalcification method constituted another pre-analytical factor of particular relevance. Although the use of nitric acid did not reach statistical significance across all analyses, it consistently showed a trend towards a higher risk of failure. This procedure, while necessary for the processing of bone specimens, is an inherently aggressive process that must be carefully controlled, as it can compromise not only histological detail and immunogenicity but also, critically, the integrity of nucleic acids required for methylation analyses. These findings are consistent with the literature, which highlights the efficacy of calcium-chelating agents such as EDTA in preserving tissue morphology and nucleic acid integrity, both essential for molecular analyses and immunohistochemistry, compared with strong acids. In the study by Singh et al. (2013), a statistically significant improvement in both DNA and RNA yield and integrity was observed in samples decalcified using EDTA-based agents (14%) and formic acid compared with those treated with hydrochloric and nitric acid-based decalcifiers.20 These results are also in agreement with the findings of Miquelestorena-Standley et al. (2020)21 and with the recommendations of the Spanish Group of Osteoarticular Pathology (2025).22
However, an unusually high failure rate was observed in our study for EDTA-decalcified samples, with core needle biopsies (CNB) of low-grade OS and surgical specimens showing the poorest performance. This may be explained by the large amount of mineralised bone matrix present in these tumours, in contrast to high-grade lesions, which require shorter decalcification times. As discussed above, time is a critical variable not only in fixation but also in decalcification, and it should be systematically recorded and controlled. It was not included in the analysis because this information was unavailable for most cases older than five years, as was the use of a 50°C oven during processing, which is routinely applied in our diagnostic practice to accelerate decalcification. In addition, nucleic acid extraction techniques from FFPE material have improved in recent years; however, limitations remain when analysing post-fixed tissue, which may be overcome by cryopreservation. Therefore, it is advantageous to plan in advance the allocation of tissue for FFPE processing, cryopreservation, and/or the use of fresh tissue.23
The parsimonious multivariate model integrated these findings and identified three main determinants of epigenetic performance in FFPE samples: sample age, low-grade osteosarcoma histology, and soft tissue localisation. The model's ability to explain most of the observed variability, together with its improved balance between fit and simplicity, supports its practical value as a tool for selecting suitable samples for methylation analysis.
From an applied perspective, these results may help optimise the use of FFPE material in epigenetic studies of osteosarcoma and bone tumours in general, allowing prioritisation of recent samples (less than 5 years old), non-decalcified specimens, or those derived from soft tissue. In cases where decalcification is required, calcium-chelating agents such as EDTA should be preferred, as they are more DNA-preserving, and attention should be paid to the absence of technical failure in other ancillary techniques, such as immunohistochemistry or, in our setting, MDM2 FISH analysis.
In this study, the Illumina FFPE quality control kit was not used for the pre-selection of FFPE samples prior to hybridisation, as based on our previous experience, its results do not reliably correlate with the final success of the analysis. Furthermore, given the limited number of available FFPE samples, all specimens were included in the study in order to retrospectively assess the true success rate and its potential association with other clinicopathological parameters of osteosarcoma.
For these reasons, the handling of biopsy material for subsequent histological and molecular analyses requires careful planning, processing, and prioritisation of tissue allocation. In addition, certain biological samples are now a standard requirement for participation in many therapeutic clinical trials, highlighting the importance of preserving fresh-frozen material. Likewise, there is an increasing number of registry studies, biomarker projects, and biobanking initiatives in which patients may wish to participate.23
Conclusions and proposed protocolBased on the results obtained and the available evidence, the following recommendations and proposed protocol are established to optimise the processing of samples intended for epigenetic analyses in osteosarcoma:
- 1.
Frozen tissue should always be prioritised over paraffin-embedded samples (FFPE), as it better preserves DNA integrity and demonstrates the highest success rate in epigenetic studies, while also reducing costs by avoiding the use of restoration kits.
- 2.
In the case of bone samples requiring decalcification, the use of strong acids (hydrochloric or nitric acid) should be avoided, as they compromise histology, immunogenicity, and nucleic acids. Preference should be given to milder agents such as EDTA or EDTA/formic acid combinations, which yield better results than strong acids, although still inferior to non-decalcified samples. It is essential to document the agent used and the duration of the process.
- 3.
Where decalcification is required, the tissue should be fixed in formalin beforehand, and the duration of exposure to the decalcifying agent should be minimised.
- 4.
The most recent samples available should be prioritised, preferably those obtained within 5 years of diagnosis, as prolonged storage in FFPE is associated with increased DNA degradation and poorer performance in methylation studies.
- 5.
Blocks with a high tumour content and minimal necrosis should be selected; microdissection should be employed where necessary to enrich the tumour fraction.
- 6.
In paucicellular cases, such as low-grade osteosarcomas, it should be borne in mind that the probability of success is lower. In such instances, priority should be given to those blocks in which FISH has hybridised appropriately, as this factor is associated with a higher likelihood of success in the epigenetic analysis.
- 7.
All pre-analytical metadata should be recorded comprehensively in the biobank database (including tissue type, sample age, decalcification agent, duration, tumour percentage, FISH status, etc.), as these variables have a determinative impact on the final quality of the results.
- 1.
Retrospective study with heterogeneity in pre-analytical documentation and without control of fixation or decalcification times.
- 2.
Limited number of fresh samples, as the project commenced in 2023 following approval by the CEIM and the inclusion of informed consent for the Biobank.
- 3.
Prospective validation studies with a standardised decalcification and extraction protocol are required.
The study was conducted in accordance with the principles of the Declaration of Helsinki and the applicable regulations governing biomedical research. The protocol was reviewed and approved by the Research Ethics Committee for Medicinal Products (reference number 2023-1219-1). As the study is based on previously archived biological samples and anonymised clinical data, no additional informed consent was required.
FundingThis work was funded by the José Mª Buesa grant from the Spanish Group for Research on Sarcomas (GEIS; reference 2024-0771-1), by the association La Lucha que Nadie Elige (reference 2023-1219-1), and by a predoctoral grant from the Spanish Association Against Cancer (AECC; reference PRDVA258009BERE).
Conflict of interestThe authors declare that they have no conflict of interest.
The authors wish to express their gratitude to the patient association La Lucha que Nadie Elige for its charitable donations, and to the associations FMPJC, VYDA, APSATUR, ASARGA, and Fundación IKER for supporting the José María Buesa Grant of the Spanish Group for Research on Sarcomas (GEIS), the 18th edition of which enabled the conduct of this study. We also thank the biobank for the management and provision of biological samples, as well as the technical staff of the Pathology Department for their valuable collaboration, professionalism, and support in sample processing. We further acknowledge the SEAP-IAP for awarding us the prize for Best Oral Communication at the XXXII National Congress of Pathology. Finally, we extend our deepest gratitude to our patients and their families for transforming their suffering into research for the benefit of life.
The followings are the supplementary data to this article:
The Spanish text of this article is available as supplementary material in Appendix 1.
