Immunohistochemical detection of the transcription factor osterix in canine and feline osteosarcoma

Materials and Methods

This study received ethical approval from the School of Biodiversity, One Health and Veterinary Medicine, University of Glasgow, Research Ethics Committee (EA61/23).

Study Samples

The origin and distribution of the study’s OSAs are summarized in Table 1. The canine study population comprised 40 decalcified OSA samples (27 productive osteoblastic, 3 poorly productive osteoblastic, 6 chondroblastic, 3 fibroblastic, and 1 telangiectatic), of which 37 represented biopsy specimens and 3 represented necropsy specimens, and 40 CTRL samples (8 chondrosarcomas, 8 fibrosarcomas, 8 hemangiosarcomas, 8 histiocytic sarcomas, and 8 (bleached) malignant melanomas), all of which represented biopsy specimens. The feline study population comprised 20 decalcified OSA samples (7 productive osteoblastic, 3 poorly productive osteoblastic, 3 chondroblastic, 5 giant cell-rich, and 2 telangiectatic), of which 19 represented biopsy specimens and 1 represented a necropsy specimen, and 20 CTRL samples (5 fibrosarcomas, 5 hemangiosarcomas, 5 chondrosarcomas, 3 (bleached) malignant melanomas, and 2 histiocytic sarcomas), of which 19 represented biopsy specimens and 1 represented a necropsy specimen.

OPEN IN VIEWER

Table 1.

Osteosarcoma study samples.

Species	Axial	Appendicular	Extraskeletal	Total
Canine	9	26	5	40
	6 skull2 vertebra1 chest a	9 (7 proximal, 1 distal) humerus6 (5 distal) femur5 distal radius3 (1 distal) tibia1 forelimb a 1 distal ulna1 metatarsus	1 jejunum1 ileum2 liver1 subcutis
Feline	7	11	2	20
	5 skull1 chest a 1 pelvis	3 (2 distal) scapula2 distal femur2 (1 proximal and 1 distal) tibia1 forelimb a 1 proximal humerus1 carpus a 1 hindlimb a	1 subcutis (interscapular)1 subcutis (overlying scapula)

a

Exact bone(s) affected unclear.

Results

Osx IHC in Canine and Feline OSAs and CTRLs

The anti-Osx antibody labeled the nuclei of the malignant spindle cell population producing osteoid in OSAs (Figs. 1b (dog) and 2b (cat)). In more strongly labeling specimens with a large proportion of the malignant spindle cell population producing osteoid demonstrating nuclear labeling, concurrent weaker cytoplasmic labeling was commonly observed (Fig. 1b).

OPEN IN VIEWER

Figure 1.

Immunohistochemistry (IHC) for osterix (Osx) in dogs. (a) Scoring and classification of canine osteosarcomas (OSA; n = 40) and control tumors (CTRL; n = 40). Following IHC for Osx, 37 OSAs and 2 CTRLs were classified as OSA (score ≥ 2), and 3 OSAs and 38 CTRLs were classified as non-OSA (score <2). (b) Canine productive osteoblastic OSA. Strong nuclear and weaker cytoplasmic immunoreactivity is observed in neoplastic osteoblasts. (Circle highlights a mitotic figure). (c) Canine osteosarcoma of telangiectatic subtype. Strong nuclear immunoreactivity is observed in a high proportion of neoplastic cells including those delineating vessel-like structures (box and corresponding inset). (d) Samples were classified according to combined (Scorers 1-3, overall) and individual scores, and classification performance parameters were computed with Clopper-Pearson 95% confidence intervals. In addition to digital scans, classification from glass slide scores was performed by Scorer 2. PPV, positive predictive value; NPV, negative predictive value. (e) Canine nasal chondrosarcoma. Strong nuclear immunoreactivity is observed in a substantial proportion of neoplastic chondroblasts. (f) Canine (bleached) oral malignant melanoma. The neoplastic population is not immunoreactive with anti-Osx antibody; however, strong nuclear immunoreactivity is observed in a region of osseous metaplasia (box and corresponding inset). (Circle highlights a mitotic figure in the melanocytic neoplastic population).

OPEN IN VIEWER

Figure 2.

Immunohistochemistry (IHC) for osterix (Osx) in cats. (a) Scoring and classification of feline osteosarcomas (OSA; n = 20) and control tumors (CTRL; n = 20). Following IHC for Osx, 14 OSAs and 5 CTRLs were classified as OSA (score ≥2), and 6 OSAs and 15 CTRLs were classified as non-OSA (score <2). (b) Feline OSA of giant cell-rich subtype. Strong nuclear immunoreactivity is observed in the neoplastic osteoblasts but not in the multinucleated cells. (c) Feline subcutaneous fibrosarcoma. Strong nuclear immunoreactivity is observed in a substantial proportion of neoplastic fibroblasts. (d) Sample classification was identical between scorers, so single overall/individual classification performance values were computed with Clopper-Pearson 95% confidence intervals. In addition to digital scans, classification from glass slide scores was performed by Scorer 2. PPV, positive predictive value; NPV, negative predictive value. (e) Feline subcutaneous hemangiosarcoma. No immunoreactivity is observed. (f) Mature feline bone. Strong nuclear immunoreactivity is observed in osteoblasts and rarely in osteocytes.

Immunohistochemical scoring and classification of all 80 scanned canine samples (40 OSAs, 40 CTRLs) resulted in 39 samples (37 OSAs and 2 CTRLs) being classified as OSA and 41 samples (3 OSAs and 38 CTRLs) being classified as non-OSA (Fig. 1a, b, c, e, f). This yielded an overall sensitivity of 92.5% (95% CI = 79.6% to 98.4%), a specificity of 95.0% (95% CI = 83.1% to 99.4%), a positive predictive value of 94.9% (95% CI = 82.7% to 99.4%), a negative predictive value of 92.7% (95% CI = 80.1% to 98.5%), and a correctly classified value of 93.8% (95% CI = 86.0% to 97.9%, P < .0001) for Osx as an immunomarker in assessing canine OSAs and salient differentials (Fig. 1d). The 3 canine OSAs not classified as OSA were poorly productive (n = 2) and productive (n = 1) subtypes, whilst the 2 CTRLs classified as OSA were both nasal chondrosarcomas (Fig. 1a, e).

In the canine cohort, productive and chondroblastic OSAs were typically allocated higher scores when compared to other OSA subtypes, whilst poorly productive OSAs exhibited the greatest score variability and were the subtype most often classified as non-OSA (score <2). Of the CTRLs, chondrosarcomas exhibited the highest scores and were most often classified as OSA (score ≥2; Fig. 1a, e).

Immunohistochemical scoring and classification of all 40 scanned feline samples (20 OSAs, 20 CTRLs) resulted in 19 samples (14 OSAs and 5 CTRLs) classified as OSA and 21 samples (6 OSAs and 15 CTRLs) classified as non-OSA (Fig. 2a, b, c, e). This gave an overall sensitivity of 70.0% (95% CI = 45.7% to 88.1%), a specificity of 75% (95% CI = 50.9% to 91.3%), a positive predictive value of 73.7% (95% CI = 48.8% to 90.9%), a negative predictive value of 71.4% (95% CI = 47.8% to 88.7%), and a correctly classified value of 72.5% (95% CI = 56.1% to 85.4%, P = .0032) for Osx as an immunomarker when assessing feline OSAs and salient differentials (Fig. 2d). The 6 feline OSAs not classified as OSA were productive (n = 2), giant cell-rich (n = 2), and telangiectatic (n = 2) subtypes, whilst the 5 feline CTRLs classified as OSA were chondrosarcomas (n = 4; locations: nose, hindlimb, axilla, and subcutis of flank) and a fibrosarcoma (n = 1; location: subcutis of flank; Fig. 2a, c).

In the feline cohort, productive and chondroblastic OSAs typically scored higher than other OSA subtypes, but productive OSAs also demonstrated the greatest variability of scores. Multinucleated cells in OSAs of the giant cell-rich subtype lacked labeling (Fig. 2b), and both telangiectatic OSAs under investigation were classified as non-OSA (score <2). Of the CTRLs, chondrosarcomas exhibited the highest scores and were most often classified as OSA (score ≥2; Fig. 2a).

Observations in Additional Samples

IHC for Osx also resulted in strong nuclear labeling, which was often combined with moderate cytoplasmic labeling of quiescent mature feline (Fig. 2f), canine (Fig. 3a), and murine (mouse and rat; Fig. 3b) osteoblasts, and osteoblasts in the context of (canine and feline) reactive and woven bone formation (Fig. 3c, d). Furthermore, canine and bovine, and to a lesser extent feline cells in the context of osseous and chondroid metaplasia exhibited moderate to strong nuclear Osx immunoreactivity and weaker cytoplasmic immunoreactivity (Figs. 1f and 3e). In canine osteoma, the neoplastic population exhibited strong nuclear Osx immunoreactivity, rarely combined with weak cytoplasmic immunoreactivity (Fig. 3f), whilst in canine multilobular tumor of bone, plump neoplastic spindle cells directly lining the osseous and cartilaginous islands and, to a lesser extent, cells within islands with chondrocyte/chondroblast morphology demonstrated moderate to strong nuclear labeling, occasionally combined with weaker cytoplasmic labeling following IHC for Osx (Fig. 3g).

OPEN IN VIEWER

Figure 3.

Immunohistochemistry for osterix (Osx) in various species and conditions. (a) Mature canine bone. Strong nuclear immunoreactivity is observed in osteoblasts; only small numbers of osteocytes exhibit nuclear immunoreactivity, and osteoclasts (circled) lack immunoreactivity. (b) Mature mouse bone. Strong nuclear and weaker cytoplasmic immunoreactivity is observed in osteoblasts, and rarely in osteocytes. (c) Feline periosteal bone formation. Strong nuclear immunoreactivity is observed. (d) Canine bone remodeling. Strong nuclear immunoreactivity is observed in osteoblasts and osteocytes associated with woven bone. (e) Feline metaplastic bone formation. Occasional strong nuclear immunoreactivity is observed delineating regions of metaplastic bone formation. (f) Canine osteoma. Strong nuclear immunoreactivity is observed in osteoblasts and less often in osteocytes. (g) Canine multilobular tumor of bone. Strong nuclear immunoreactivity is observed in a small proportion of neoplastic cells (box and corresponding inset). (h) Mature chicken bone. No immunoreactivity is observed. (i) Mature zebrafish. No immunoreactivity is observed in bones and cartilage associated with and surrounding the gills. (The pigment represents melanin, box and corresponding inset). (l) Immature feline femur. Strong nuclear immunoreactivity is observed in osteoblasts and chondroblasts of the growth plate and in the cambium layer (box and corresponding inset). (m–n) Mouse embryo. Strong nuclear immunoreactivity is observed in cells with osteoblast and chondroblast morphologies. (n) Higher magnification of the boxed region in (m).

Canine, feline, and murine mature osteocytes and, to a lesser extent, quiescent chondrocytes/chondroblasts rarely demonstrated convincing Osx immunoreactivity (Figs. 2f and 3a, b). No labeling of canine osteoclasts (Fig. 3a) and parrotlet, chicken (Fig. 3h), or zebrafish (Fig. 3i) osteoblasts, osteocytes, and chondrocytes/chondroblasts was observed. In immature feline and mouse tissues, cells with osteoblast or chondroblast morphologies exhibited Osx immunoreactivity (Fig. 3l–n).

Discussion

Definitive diagnosis of OSA in domestic animals can be highly challenging, especially in the event of very small amounts of and/or heterogeneously distributed osteoid. This is reflected in a number of recent studies that have attempted to characterize reliable markers, particularly those targeting malignant neoplastic osteoblasts in dogs.^1,4,35,42 Whilst several markers, including alkaline phosphatase, osteocalcin, osteonectin, osteopontin, and RUNX2, have demonstrated very good sensitivity in identifying canine (or human) OSA, these markers typically have lower specificities in dogs.^4,11,35,42 As a result, Barger et al suggested the application of antialkaline phosphatase and anti-RUNX2 antibodies in series as a sensitive (87%) and specific (85%) approach to the diagnosis of canine OSA. When used as part of an IHC panel, the anti-special AT-rich sequence-binding protein 2 antibody has a sensitivity of 86.0% and a specificity of 98.8% for OSA in dogs. ³³ In addition, Amaral et al ¹ suggested that a panel of antibodies against vimentin, osteonectin, and osteocalcin may represent a useful diagnostic and prognostic aid for canine OSA, although they did not indicate the sensitivity or specificity of this approach. When viewed within the context of these previous investigations, the present study demonstrates that IHC for Osx, applied as a single antibody, is a highly sensitive (92.5%) and specific (95.0%) marker for canine OSAs. Furthermore, combining this IHC marker with an established scoring and classification system was predictive and associated with good interobserver agreement (91.3% classification agreement). Therefore, the approach outlined in this study is proposed as a useful adjunctive tool in the diagnosis of canine OSA.

Reliable markers for feline OSAs have not been reported. In this study, IHC for Osx was also applied to feline samples and demonstrated an overall sensitivity of 70.0%, a specificity of 75.0%, and a correctly classified value of 72.5%, which was lower than for canine samples. Different scoring thresholds may improve the utility of Osx IHC for feline bone tumors; however, this is outside of the scope of this study and would likely require a larger study population.

Beyond the incidence, the age of onset, predilection sites, metastatic rates, and prognostic characteristics of OSAs vary between cats and dogs. ⁹ These variations may suggest differences in the pathobiology of OSA between these 2 species, which could potentially include dissimilarities in binding epitopes and/or transcription factor expression and could therefore underlie the reduced efficacy of Osx IHC in cats compared to dogs. Further investigation of this hypothesis would require molecular characterization of feline Osx, which again lies outside the scope of this study.

The difference in sensitivity of Osx IHC in identifying canine and feline OSAs may also have been influenced by the relative proportions of each subtype within each OSA population, notwithstanding inherent challenges in subtyping OSAs. OSA subtypes were not equally represented across both species; for example, giant cell-rich OSA accounted for 25% of the total feline OSAs but was not included in the canine OSA study population. In dogs, productive osteoblastic OSAs are reportedly the most common OSA subtype. In one study, these accounted for 55% of the OSAs examined (67.5% of the OSAs in the current study), whilst poorly productive osteoblastic OSAs accounted for 14% (8% in the current study), chondroblastic OSAs accounted for 13% (7% in the current study), and giant cell-rich OSAs accounted for 2% (0% in the current study). ²⁴ Reports in cats vary. In 1 study, osteoblastic tumors comprised 42% of feline OSA; mixed osteoblastic and fibroblastic, chondroblastic, and/or telangiectatic subtypes combined accounted for a further 50%; and exclusively fibroblastic OSAs made up the remaining 8%. ⁹ In a different study investigating 10 feline OSAs, 40% were chondroblastic, 30% were fibroblastic, and 30% were osteoblastic subtypes. ⁷ The feline OSA study population in this study comprised roughly comparable proportions of productive (35%) and poorly productive (15%) osteoblastic OSAs, but fewer (15%) chondroblastic OSAs. Therefore, and whilst the exact proportions vary between studies, osteoblastic, in particular productive, OSAs, which would be expected to be more likely to express Osx, seem to be most common in both species, and the variation in incidence of productive osteoblastic subtypes between canine (67.5%) and feline (35%) cases in this study could have contributed to differences in the sensitivity of Osx immunoreactivity between dogs and cats.

In dogs, poorly productive osteoblastic OSAs were the most common subtype classified as non-OSA following IHC for Osx. In cats, productive, giant cell-rich, and telangiectatic subtypes were most commonly classified as non-OSAs following IHC for Osx; feline productive osteoblastic OSAs also had the greatest variability in scores. This most likely reflects the degree of divergence of the neoplastic population from functional osteoblasts, as suggested both by their limited production of osteoid and histological appearance. The neoplastic population in OSA is notoriously diverse, and therefore, lack of or aberrant protein expression is not unexpected.^{11,12,16,29,42}

Productive osteoblastic OSAs in both cats and dogs typically demonstrated the strongest Osx immunoreactivity, likely reflecting a closer relationship of these neoplastic cells to functional osteoblasts, which is also suggested by their histological appearance and consistent production of osteoid. At the same time, both canine and feline chondroblastic OSAs frequently received high IHC scores, and chondrosarcomas were the CTRLs most commonly classified (score ≥ 2) as OSA in this study in both cats and dogs. Indeed, 4/5 feline chondrosarcomas were classified as OSAs following IHC for Osx. The most likely explanation for this is that chondrocytes and osteoblasts arise from a common bipotential progenitor known as a preosteoblast, ³⁰ and although Osx is thought to be critical for the differentiation of preosteoblasts into functional osteoblasts, Osx transcripts have been identified transiently in differentiating chondrocytes in murine embryos. ¹⁰ Furthermore, Osx expression detected by IHC has been reported in chondrocytes in the mandibular cartilage of newborn mice. ⁴⁸ Therefore, the Osx immunoreactivity of chondrosarcomas in this study may be the result of recapitulation of (transient) Osx expression by neoplastic cells. An alternative explanation is that these tumors could indeed be chondroblastic OSAs that were misdiagnosed due to sampling bias and/or sectioning artifact, which precluded identification of osteoid. While this scenario cannot categorically be ruled out, the 2 canine chondrosarcomas classified as OSAs were both nasal, and a misdiagnosis of nasal chondroblastic osteosarcoma as nasal chondrosarcomas seems less likely given their relative frequencies.^25,31,44,46

In comparison to the dogs, the location of 4/5 feline chondrosarcomas classified as OSA following IHC for Osx was more heterogeneous and included 1 nasal mass and 3 masses from the trunk and limbs, of which 1 was subcutaneous and the exact locations of the other 2 were unclear. Similarly, the 1 feline fibrosarcoma classified as an OSA following IHC for Osx originated from the truncal subcutis. Whilst again misdiagnosis cannot categorically be ruled out, especially as fibroblastic osteosarcomas can be challenging to differentiate from fibrosarcomas, the possibility of the immunoreactivity of the non-osteoblastic mesenchymal cells reflecting a common precursor is plausible, especially as osteoblastic and chondroblastic injection site sarcomas have been reported.^{1,15,17,19,29} On a further potentially associated note, 5/20 feline OSAs (25%) were present in the region of the scapula, of which 3 were associated with the bone and 2 were present in the overlying subcutis; no canine periscapular masses, including subcutaneous extraskeletal OSAs, were identified. Therefore, the possibility that these feline tumors could be vaccine-associated may need to be considered.

The neoplastic population in canine osteoma labeled positively for Osx, which is to be expected given the role of Osx in the differentiation of osteoblasts. The neoplastic population in canine multilobular tumor of bone is less well characterized; however, these tumors are thought to arise from precursors within the periosteum. This tumor also demonstrated some Osx immunoreactivity, which is expected given the production of cartilage and bone in such tumors.^24,40

Osx expression was not only observed in cells with osteoblastic but also in cells with chondroblastic morphology in embryonic mouse and immature kitten bone, which is consistent with a role of Osx in bone development. Non-neoplastic mature canine, feline, and murine (mouse and rat) osteoblasts also demonstrated Osx immunoreactivity, as did osteoblasts in canine and feline reactive and woven bone formation. Therefore, IHC for Osx does not distinguish between normal, reactive, and neoplastic osteoblasts and would be expected to label reactive osteoblasts in the context of callus formation, considering its role in the physiological maturation of osteoblasts and homeostasis of adult bone.^3,30 Importantly, Osx immunoreactivity would also be expected in metaplastic bone and possibly also cartilage formation, which must be considered when applying anti-Osx antibody (and this scoring system) to samples with such changes, such as malignant melanomas with intralesional metaplastic bone and/or cartilage formation. In accordance with the role of Osx in the differentiation of osteoblasts, osteocytes and quiescent chondrocytes/chondroblasts were considerably less immunoreactive to anti-Osx antibodies than osteoblasts. ³⁰ Osteoclasts lacked immunoreactivity.

The lack of Osx immunoreactivity observed in osteoblasts (and osteocytes) of birds (parrotlet and chicken) and a zebra-fish in this study may reflect their evolutionary divergence from mammals. While Osx is evolutionarily conserved across most vertebrates, its role and amino acid sequence seem to differ between species.^13,30 Osx-null mice demonstrate embryonic failure of bone formation. ³⁰ The same is true in some fish, e.g. Medaka (Oryzias latipes). ⁴⁷ However, Osx-mutant carp (Cyprinus carpio) and, in other studies, zebrafish (reports vary) are able to mature to juveniles or young adults before bone lesions become appreciable.^{8,23,30,47,49} The absence of immunoreactivity in a zebrafish of unknown age, but presumed to be subadult or adult, in this study is difficult to interpret in the context of conflicting literature; however, a possible reduced reliance on Osx for early bone development in fish may need to be considered. Similarly, the importance of Osx in birds is uncertain; Yu et al ⁴⁷ evaluated 78 avian genomes, found no Osx homologue, and identified an Osx-like gene in just 7 avian species for which there was no clear phylogenetic basis. This peculiarity of the avian genome may explain the absence of labeling of the avian tissues assessed in this study.

Prolonged decalcification (>8 days) is reported to reduce nuclear immunoreactivity in some cases. ²⁶ We compared Osx immunoreactivity between paired formic acid decalcified and non-decalcified sections in a small subset of OSAs (n = 4). The non-decalcified tissues were given identical or slightly higher scores than decalcified tissues, but this did not change the classification (OSA vs non-OSA) in any case. Furthermore, should decalcification significantly impede immunoreactivity, lesions with a greater proportion of mineralized bone, such as productive OSAs, would be expected to require longer decalcification and therefore demonstrate reduced Osx immunoreactivity, whereas in this study, canine productive OSA labeled more strongly than less productive subtypes, although this may also be the consequence of dedifferentiated neoplastic cells as discussed above. Therefore, based on this study, IHC for Osx appears to be suitable for use on tissues decalcified with formic acid. Variable effects of different methods of decalcification on IHC have been reported; in particular, HCl decalcification is reported to decrease immunoreactivity of melanomas to anti-melanA and anti-PNL2 antibodies, and HCl-decalcified canine OSAs demonstrated weaker osteocalcin and osteonectin immunoreactivity compared to tissues decalcified with formic acid or EDTA.^32,42 Therefore, the efficacy of anti-Osx antibody in the context of non-formic acid-based decalcification remains unclear, and the decalcification method ought to be considered when interpreting Osx immunoreactivity.

Similarly, bleaching did not affect the scoring of Osx expression in canine melanoma in this study. However, considering that melanomas would not be expected to label with anti-Osx antibody and that no bleached OSAs were evaluated, the possibility that bleaching OSAs may reduce Osx immunoreactivity cannot be excluded. Bleaching is reported to reduce immunoreactivity of some IHC markers. For example, melanA and HMB-45 immunoreactivity in canine melanoma was abolished after bleaching, but PNL2 was not. ³² Regardless, this is unlikely to be of major concern, as predominantly only very poorly pigmented melanomas would be likely to be subject to IHC for Osx, and these are unlikely to require bleaching.

The retrospective nature of this study meant that there were several variables associated with sample handling that could have impacted the results of this study. Firstly, the age of the blocks; although no obvious association between block age and immunoreactivity score was observed. Secondly, the duration of fixation of these samples was unknown, and prolonged fixation is known to reduce immunoreactivity in some cases. ³² This potential confounding factor is partially mitigated by the fact that the vast majority of study samples were biopsies submitted to one diagnostic facility; therefore, reasonable standardization of fixation and processing could be expected. Osteosarcomas sampled at necropsy (4/60) rather than biopsy would likely have more variable fixation times and prolonged sampling intervals; however, they did not exhibit appreciably decreased intensity of immunoreactivity. Finally, the time spent exposed to decalcification agents also varied between samples, reaching up to 14 days; however, as previously discussed, decalcification did not appear to affect Osx immunoreactivity in this study.

As discussed above, some subtypes of OSA were more common than others, and the number of samples and the distribution of subtypes differed between cats and dogs, which may compromise direct comparison between Osx expression in canine and feline OSAs. Overall, however, the number of samples in this study is greater than those reported in other studies examining IHC markers for OSAs. Considering that certain subtypes of OSA tended to exhibit a lower expression of Osx, it is advised that additional IHC tests, such as for endothelial markers in suspected telangiectatic OSAs, are performed alongside IHC for Osx where necessary.

The inherent difficulty in diagnosing OSA on hematoxylin and eosin-stained slides was the initial impetus for this investigation and is a significant potential limitation of the study. This was mitigated as far as possible by choosing study OSAs based on the definitive presence of osteoid formed by a malignant neoplastic population and further validating this selection via consensus diagnosis between 2, or if necessary 3, board-certified veterinary anatomic pathologists. Equivalent stringency was applied for the selection of non-OSA samples and other tissues examined. Nevertheless, the possibility that some of the CTRLs classified as OSA following IHC for Osx truly represented OSAs in which the production of osteoid by a malignant neoplastic population had not been detected due to sampling bias (including the submission of incisional biopsies) or sectioning artifact cannot categorically be ruled out.

Finally, the potential for recognition of study samples by the scorer involved in initial sample evaluation was mitigated by the blinded assessment of samples and concurrent assessment by a second scorer who had not been involved in initial sample evaluation. These 2 scorers assigned comparative scores (interscorer classification agreement 93.3%). There was minimal overlap (2 of 7 cases) between samples, both re-evaluated on hematoxylin and eosin and scored by the third reviewer/scorer. In general, standardized and unbiased scoring was achieved by annotation of the regions of interest by an individual not involved in scoring and statistical examinations conducted by an individual who was completely independent from sample evaluation.

In conclusion, the results of this study suggest that IHC for Osx, in combination with the scoring system described herein, is a highly sensitive and specific marker for canine OSA and a moderately sensitive and specific marker for feline OSA that can be used on samples decalcified with formic acid, warranting further, even larger-scale studies assessing this marker. Some OSA subtypes are less reliably immunoreactive than others, and particular care is advised when interpreting Osx immunoreactivity of suspected poorly productive canine and feline OSAs, as well as productive, giant cell-rich, and telangiectatic feline OSAs; supplemental relevant IHC antibodies could also be considered. On a similar note, immunolabeling of neoplastic chondrocytes with anti-Osx antibody was also observed; therefore, careful assessment for the presence of osteoid produced by a malignant neoplastic population, combined with taking into account the clinical features, remains key for discrimination between chondroblastic OSA and chondrosarcoma. In addition, since the anti-Osx antibody immunolabels all (canine, feline, and murine) osteoblasts, to a lesser extent osteocytes, and in some cases also chondroblasts and chondrocytes, care must be taken when evaluating Osx immunoreactivity in the context of callus formation, reactive bone formation, and osseous and chondroid metaplasia. Nonetheless, IHC for Osx is anticipated to be a useful tool to support the diagnosis of canine and feline OSA.

Supplemental Material