Canine mast cell tumors: utility of stereologic tools in cytology

Abstract

Quantitative morphologic parameters assessed in cytologic samples of canine cutaneous mast cell tumors (ccMCTs) may assist with surgical planning and prognostication. Robust cutoffs can be defined, with high reproducibility, for parameters such as the nuclear area (NA). The NA may be determined by morphometry (image analysis, NAI) or by stereology, such as the 2D-nucleator method (NAN); stereologic techniques have not been applied to cytologic specimens of ccMCT, to our knowledge. We retrospectively selected routine cytology smears from 51 ccMCT cases and screened them to determine the percentage of neoplastic mast cells with indistinct nuclear borders; this was repeated after the slides were restained with H&E. The NAI and the NAN were estimated in 100 mast cells per animal in H&E-stained slides. All nuclei were visible in H&E smears, and unbiased quantification was feasible. The NAN was similar to NAI, but less time-consuming. Both the NAN and NAI determined by cytology differed in histologic low- and high-grade ccMCTs, and in histologic grade I plus II versus grade III ccMCTs. Stereologic parameters such as the NAN could be considered as complementary techniques for the cytologic evaluation of ccMCTs.

Canine cutaneous mast cell tumors (ccMCTs) are common neoplasms, accounting for up to 21% of all canine skin tumors. ¹⁶ Cytologic evaluation is a quick, relatively noninvasive, and inexpensive way to diagnose ccMCTs. Although cytologic diagnosis of ccMCTs is rarely challenging, differentiating ccMCTs with a benign clinical course from aggressive forms can be puzzling. In most cases, ccMCTs are surgically resected, and grading schemes are applied to histologic sections.^5,13 The efficacy of surgical excision is well correlated with histologic grade and proliferation indices. ¹⁶ However, these validated prognostic factors for ccMCTs are determined during histologic, rather than cytologic, examination. ¹⁶

Quantification procedures in cytologic samples may fill the gap between histopathology grading assessment and surgical planning. ¹⁷ The nuclear area (NA) of mast cells has been shown to be correlated with 3-tier grades (I/II different from III),^12,17 and with prognosis. ¹⁸ Quantitative parameters are objective, have the advantage of allowing clear definition of cutoffs, and have high reproducibility. ¹ Image analysis applied to cytologic smears has proved useful in assessment of neoplastic lesions, from ccMCTs ¹⁷ to mammary tumors ¹⁴ or perianal adenocarcinomas. ¹⁵ However, morphometric estimation of NA can be time-consuming because the nuclear limits are usually outlined manually by an operator.^6,12,17 Moreover, nuclear features are often difficult to assess in routine cytologic preparations, especially in highly granulated ccMCTs. ⁹

Stereologic methods have not been applied to the study of ccMCTs in cytologic samples, to our knowledge. Contrary to morphometric approaches that use rulers or caliper micrometers for direct measurements of cells or convert pixels of digital images into micrometers, stereology methods use a toolbox of probes or test systems, based on mathematical and statistical principles.^2,7,10 These geometric probes are superimposed directly on cells,^7,10 and are used to recover 3-dimensional (3D) properties of cells (such as number, length, area, or volume) from 2D planes.^2,7,10 These 2D planes are applied usually in tissue sections, but stereologic methods can also be applied to cytologic samples. ⁴

We compared the assessment of the NA by a morphometric approach and by a stereologic technique (2D-nucleator) in cytologic samples and evaluated their correlation with histologic grading systems^5,13 to define a threshold value that could distinguish low- and high-grade ccMCTs.

A retrospective analysis of the Cytology Diagnostic Services archive was performed (Jan 2012–June 2016); we retrieved 68 ccMCT cases. Slides of fine-needle aspirates were reviewed to assess cell preservation and number of mast cells; 17 cases were excluded because of poor cell preservation or low cellularity. The slide with more or better-preserved cells was selected in each case. We enrolled slides from 51 cases of ccMCT. The NA was determined by the stereologic method (NAN; 2D-nucleator) and by a morphometric approach (NAI; outlining nuclei using ImageJ, https://imagej.net). Corresponding histopathology slides were retrieved from 32 cases from the Laboratory of Pathology of the Institute of Biomedical Sciences Abel Salazar (University of Porto, Portugal). All procedures were approved by local ethics committees (ORBEA 237/2017).

All cytologic slides had been stained with Hemacolor (Merck) using the same protocol (1, 2, and 3 min in each solution). A minimum of 100 mast cells were screened per slide to determine the percentage of mast cells with obscured nuclear limits, and their number recorded. Coverslips were removed by immersion in xylene for 1–4 d, and slides were rehydrated and destained in an alcohol–acid solution, controlled by microscopic observation. After rinsing in tap water, slides were immersed in Gill hematoxylin overnight, ⁹ rinsed again in tap water, and stained with eosin (5 min).

The NAI was assessed by taking 10–12 photos of each case (final magnification of 4,050×). The operator observed the images on a flat computer screen and outlined manually the mast cell nuclei using the computer mouse.^6,17 ImageJ software then provided the area of each nucleus (Fig. 1A).

Figure 1.

Determination of nuclear area (NA) in cytologic samples from canine cutaneous mast cell tumors. A. Estimation of the NA in the H&E-restained slides with ImageJ, by outlining the nuclei of mast cells with the computer mouse. B. Estimation of the NA in the H&E-restained slides with the 2D-nucleator, in which the operator selected the nucleolus of each mast cell and the software generated 2 perpendicular lines (blue lines; when no nucleolus exists, the center of the nucleus is used); then, the intersections between these lines and the nuclear borders were marked (white crosses). C. Boxplots of the NA of cytologic smears in low- and high-grade canine cutaneous mast cell tumors, determined by the image analysis (ImageJ) and stereologic (2D-nucleator) methods. The boxes show first quartile, second quartile (median), and third quartile. The whiskers represent the range (minimum and maximum). Significant differences between grades existed using both methods. D. Bland–Altman plot of the comparison of NAs between image analysis (ImageJ) and stereologic (2D-nucleator) methods; mean bias −2.85 (95% CI: −0.88, −4.82).

Stereology software was used (CAST-Grid v.1.60; Olympus) for the 2D-nucleator method. ³ All measurements were performed at the monitor (final magnification 4,050×), using a 100× oil-immersion objective. The first field was selected randomly and thereafter the motorized stage (Prior; Cambridge, UK) was moved to quantify the NANs in a minimum of 100 cells (average 101; range 100–120). ² According to the 2D-nucleator rules, ³ the operator selected the nucleolus with the computer mouse and the software generated 2 perpendicular lines; then the operator marked the intersection between these lines and the nuclear outlines (Fig. 1B). The NAN was computed by NA = πl², in which l is the average distance between the nucleolus and nuclear limits. ³ Mast cells with ≥2 nuclei were counted separately. Cells other than mast cells were excluded.

Histologic grading (3-tier [Patnaik ¹³ ] and 2-tier [Kiupel ⁵ ] systems) was performed on 32 ccMCTs, by 2 observers (M. Santos, P. Dias-Pereira) who reached consensus while simultaneously examining histologic slides on a multi-head microscope. Histologic grade was correlated with the NAN and NAI. Researchers assessing the NAN and NAI were anonymized to the histologic grade of a case.

Statistical analysis was performed (R v.3.6.1; https://www.r-project.org/). Data were tested for normality, and correlations between NAN and NAI were evaluated using the Spearman test. Methods were compared by Passing–Bablok regression analysis and Bland–Altman plots; bias and 95% limits of agreement were calculated. The NAs estimated cytologically and histologic grades were compared using the Kruskal–Wallis test (NAI) or one-way ANOVA (NAN). Receiver operating characteristic (ROC) curve analysis determined the best cutoffs for NAN and NAI. The significance level was established at p ≤ 0.05, and data were presented as mean ± SD.

The 51 ccMCTs were from 29 female and 22 male dogs, 5–16-y-old, of various breeds; most common breeds were Labrador Retriever (17) and Boxer (9). The most common locations for ccMCT were limbs (16) and abdomen or thorax (16).

In Romanowsky-stained smears, nuclear limits of 43.6 ± 25.5% mast cells were obscured by granules, whereas the nuclear limits of all mast cells were clearly observed in restained H&E smears. We measured 5,418 mast cells (5,272 mononucleate and 146 binucleate cells, respectively) using the nucleator. The NA of mononucleate cells (NAN 57.3 ± 14.0 μm²) was different from that of binucleate cells (NAN 49.2 ± 12.3 μm²; p = 0.006), independent of the method used. Given that few binucleate mast cells were observed per case, we excluded these from the remaining statistical analysis.

The time needed to obtain the NA estimations differed between methods: 40 min per smear to obtain the NAN, and 60 min to obtain the NAI. However, the mean NAN and NAI appeared similar (Fig. 1C), and a significant correlation existed between NAN and NAI (r = 0.87; p < 0.0001). Passing–Bablok regression analysis showed that 2D-nucleator and image analysis techniques were highly comparable, even if Bland–Altman plots highlighted a constant negative bias for NAN (Fig. 1D; Suppl. Fig. 1).

Among the 32 cases of ccMCT with histologic grading, 22 were low-grade and 10 were high-grade, in the 2-tier system. Their NA differed, either using the 2D-nucleator [NAN (p = 0.01)] or image analysis [NAI (p = 0.02)]. ROC curve analysis ⁸ determined the best cutoff values for distinguishing low- versus high-grade tumors (Table 1). Among those 32 cases, 4, 24, and 4 cases were graded I, II, and III, respectively (Suppl. Fig. 2). Differences existed for the NAN between grades I and III (p = 0.02) and between grades II and III (p = 0.005). By contrast, NAI differences existed only between grades II and III (p = 0.004). Using the 3-tier grading, the NAI cutoff points were higher than those of the 2-tier grading (Table 2).

Table 1.

Cutoffs for the nuclear area in cytologic smears for distinguishing low- versus high-grade (2-tier, Kiupel ⁵ ) canine cutaneous mast cell tumors, determined by the 2D-nucleator and image analysis (ImageJ) methods.

Nuclear area	Cutoff (µm2)	AUC	Se	Sp	PPV	NPV
Nucleator	50.1	0.786 (0.625–0.947)	100 (65.6–100)	50 (30.7–69.3)	47.6 (26.4–69.7)	100 (67.9–100)
Image analysis	55.5	0.768 (0.598–0.938)	100 (65.6–100)	50 (30.7–69.3)	47.6 (26.4–69.7)	100 (67.9–100)

AUC = the area under the curve; NPV = negative predictive value; PPV = positive predictive value; Se = sensitivity; Sp = specificity.

Table 2.

Cutoffs for the nuclear area in cytologic smears for distinguishing grade I and II versus grade III (Patnaik ¹³ ) canine cutaneous mast cell tumors, determined by the 2D-nucleator and image analysis (ImageJ) methods.

Nuclear area	Cutoff (µm2)	AUC	Se	Sp	PPV	NPV
Nucleator	65.7	0.874 (0.619–1.00)	66.7 (12.5–98.2)	100 (85.4–100)	100 (19.8–100)	96.7 (80.9–99.8)
Image analysis	63.5	0.920 (0.808–1.00)	100 (30.9–100)	86.2 (67.4–95.5)	42.9 (11.8–79.8)	100 (83.4–100)

AUC = the area under the curve; NPV = negative predictive value; PPV = positive predictive value; Se = sensitivity; Sp = specificity.

We found that diagnostic cytologic specimens can be used for assessing relevant quantitative parameters of ccMCT, which can assist in defining the prognosis. The NA has been correlated with the 3-tier grading^12,17 and prognosis. ¹⁸ We corroborated these findings by showing that the NAI of grade I/II was different from grade III. Moreover, NAI and NAN were significantly different in low- versus high-grade ccMCTs in the 2-tier grading system.

Classical studies with nuclear profiles of epithelium have shown that point counting and image analysis produced similar NA values, with higher efficiency of the stereologic method, with a third of time needed. ² The 2D-nucleator took less time (40 min) than image analysis nuclear outlining (60 min), which is a time-consuming procedure that must be performed manually. Automatic outlining by ImageJ was not used because cytologic samples have lower contrast compared with tissue sections, and direct observation of smears is always needed to avoid quantifying cells other than mast cells.

The NAI determined herein was comparable ¹⁷ or lower ¹² than that reported by others. The first study on this subject ¹⁷ either determined the NAI in smears fixed in alcohol for 24 h and then stained with H&E or in air-dried smears stained with Romanowsky-type solutions (panoptic stains). The estimations obtained in alcohol-fixed preparations were lower, as a result of the cell retraction inherent to immersion fixation. ¹⁷ In contrast, the estimations in air-dried smears were higher ¹⁷ (69.3 ± 6.4 µm²) than ours (51.9 ± 6.7 µm² for smears of 3-tier, grade II). These discrepancies can be because of granules that sometimes obscure the nuclear limits, making outlining difficult. The presence of granules may also justify the differences to the NAI reported by other studies (72.3 ± 13.9 µm² for the NA of 3-tier, grade II). ¹²

The NAI and NAN had acceptable performance for differentiating high- versus low-grade ccMCTs, and excellent performance for differentiating grade III from I/II. By determining the NAN in cytologic samples, we found a 79% chance of distinguishing low- from high-grade ccMCTs (Table 1), which is advantageous for surgical planning. The cutoffs can be viewed from different angles; a ccMCT with a NAN <50.1 μm² is definitely low-grade, a ccMCT with NAN ≥50.1 μm² has a 50% chance of being high-grade (Table 1). A ccMCT with a NAN >62.8 μm² is definitely high-grade (30% sensitivity, 100% specificity), which is valuable information when more radical treatments, such as limb amputations, are being considered.

We used destaining/restaining to unmask the mast cell nuclei, and this process can be viewed as a limitation. Without this procedure, nuclear contours would probably be larger, and excluding these cells from quantification would also introduce significant bias. We cannot ignore that the destaining/restaining procedure may have changed the NA, although this seems unlikely from previous reports, in which direct comparisons between Diff-Quik and H&E smears were performed. ⁹ The availability of stereologic software may be another limitation to conducting stereology. This may be overcome by using manual methods or free software. For instance, the NA can be determined by overlaying an acetate with a grid of points over a computer screen or by using free web resources, such as the STEPanizer ¹⁹ or ImageJ stereology plugins. ¹¹ Alternatively, cytologic smears with adequate cellularity can be sent to specialized laboratories to have the quantification performed.

The histologic grade has a major role in ccMCT prognosis, and previous studies have reported that the NAI also has a prognostic role. ¹⁸ In our study, as in others, ¹² we were unable to distinguish Patnaik ¹³ tier I from II. Further studies, with a larger number of cases and with rigorous follow-up, are needed to define the prognostic value of the NAN and to relate it to other clinical outcome predictors, such as grading and proliferation scores. ¹⁶ Stereologic parameters such as the NAN should be considered as complementary techniques for the clinical evaluation of ccMCTs, pinpointing aggressive tumors for which tailored surgical and adjuvant therapies are needed.