Human Pathology (2015) 46, 304–312

www.elsevier.com/locate/humpath

Original contribution

Microvessel density, lymphovascular density, and lymphovascular invasion in primary cutaneous melanoma—correlation with histopathologic prognosticators and BRAF status☆ Phyu Phyu Aung MD, PhD a,1 , Dominick Leone MPH, MS b , John Kyle Feller MA a , Shi Yang MD c , Marier Hernandez MD d , Ron Yaar MD e , Rajendra Singh MD f , Thomas Helm MD g , Meera Mahalingam MD, PhD, FRCPath a,⁎ a

Dermatopathology Section, Department of Dermatology, Boston University School of Medicine, Boston, MA 02118 Department of Biostatistics, Boston University School of Public Health, Boston, MA 02118 c Department of Pathology, Boston University School of Medicine, Boston, MA 02118 d Department of Dermatology, UT Southwestern Medical Center, Dallas, TX 75390 e Aurora Diagnostics, Greensboro, NC 27408 f Departments of Dermatology and Pathology, Mt Sinai School of Medicine, New York, NY 10029 g Department of Dermatology, State University of New York at Buffalo, Buffalo, NY 14203 b

Received 11 September 2014; revised 10 November 2014; accepted 12 November 2014

Keywords: Primary cutaneous melanoma; Microvessel density; Lymphovascular density; Lymphovascular invasion; Histopathologic prognosticators; BRAF

Summary The relationship between microvessel density (MVD), lymphovascular density (LVD), and lymphovascular invasion (LVI) in primary cutaneous melanoma (PCM) remains unclear. Given this, a total of 102 PCMs were assessed for MVD (vascular endothelial growth factor receptor 2 and Endocan), LVD (D2-40), and LVI (immunostaining with D2-40/S-100 and hematoxylin and eosin); tumoral S-100A13, vascular endothelial growth factor receptor 2, and Endocan; and BRAF status. LVD was associated with MVD (P = .01). MVD was higher in PCMs with depth greater than or equal to 2 mm and ulceration (P = .04, .05), whereas LVD was higher in PCMs with depth greater than or equal to 2 mm and mitoses (P = .03, .02). After adjusting for MVD and LVD, only ulceration was associated with LVI (P b .02). A BRAF mutation was seen in 30.4% cases, and when present, both LVD and host response (P = .0008 and .04, respectively) were significantly associated with MVD. Immunostaining with S-100A13 was noted in 99% of cases and a significant association noted only with ulceration (P = .05). Immunostaining increased LVI positivity (46.5% versus 4.9% by hematoxylin and eosin, P b .0001). MVD and LVD are not associated with LVI, appear to be closely related with each other, and are associated with select markers of poor prognosticative value. The association between a host response and LVD and MVD in PCMs with a BRAF mutation suggests that they exhibit potential for strategizing immunotherapies. © 2015 Elsevier Inc. All rights reserved.



Disclosures: None to disclose. ⁎ Corresponding author. Dermatopathology Section, Department of Dermatology, Boston University School of Medicine, 609 Albany St, J-401, Boston, MA 02118, USA. E-mail addresses: [email protected] (P. P. Aung), [email protected] (D. Leone), [email protected] (J. K. Feller), [email protected] (S. Yang), [email protected] (M. Hernandez), [email protected] (R. Yaar), [email protected] (R. Singh), [email protected] (T. Helm), [email protected] (M. Mahalingam). 1 Current address: Departments of Pathology and Dermatology, Dermatopathology Section, U. T. - M. D. Anderson Cancer Center, Houston, TX 77030. http://dx.doi.org/10.1016/j.humpath.2014.11.006 0046-8177/© 2015 Elsevier Inc. All rights reserved.

MVD, LVD and LVI in melanoma

1. Introduction In cutaneous malignancies, microvessel density (MVD) has been shown to correlate with tumor grade in canine mast cell tumors [1]. In a study of human cutaneous nonmelanoma skin cancers, although MVD was found to be higher in squamous cell carcinoma than basal cell carcinoma and Bowen disease, a correlation with histopathologic prognosticators was not performed [2]. Although this correlation has indeed been demonstrated in malignant melanoma, results have been inconclusive and conflicting [3-10]. Using a plethora of markers, studies reporting an association between increased vascularity and an unfavorable outcome [3-6] have been refuted by others showing that increased MVD is significantly associated with improved patient survival [7-10]. Utilization of the lymphatic endothelial marker, D2-40, for measurement of lymphovascular density (LVD) in cutaneous malignancy has been reported in squamous cell carcinoma of the head and neck region with conflicting results regarding its prognosticative value [11]. This appears to be true for LVD in melanomas as well. In support of this, in 1 study, metastatic melanomas had significantly more LVD, which was associated with poor disease-free and overall survival, whereas in another, decreased LVD was present in thicker and more proliferative tumors (Ki-67) [6,7]. The incidence of lymphovascular invasion (LVI), based on hematoxylin and eosin (H&E) staining alone in melanoma has been shown to range from 0% to 6%. Select studies have shown that this detection rate increases with the use of select immunohistochemical stains targeting endothelial cells [12-14]. The relevance of detection of LVI in primary cutaneous melanoma (PCM) lies in that it has been shown to be significantly associated with time to regional nodal metastatic disease as well as first metastasis and disease-related death [15-17]. The premise of the current study was to ascertain precisely the relationship between MVD, LVD, and LVI in PCM. MVD was assessed using vascular markers including vascular endothelial growth factor receptor 2 (VEGFR-2) and Endocan; LVD was assessed by D2-40; and LVI was assessed by both H&E and double staining of S-100 and D240. We also assessed for tumoral expression of the proangiogenic marker (S-100A13), VEGFR-2, and Endocan. Markers used in the current study were selected for the following reasons: VEGFR-2, an autocrine growth factor receptor for VEGF, shown to be the dominant effector of VEGF function in the metastatic melanoma microenvironment [18,19]; Endocan or endothelial cell specific molecule 1, a soluble proteoglycan secreted by endothelial cells, overexpression of which has been shown in vitro to be a poor prognosticator in melanoma [20]; and S-100A13, a proangiogenic molecule and a calcium-binding protein involved in the release of fibroblast growth factor family [21]. An additional aim was to ascertain the correlation between MVD, LVD, and established histopathologic prognosticators

305 as well as the BRAF status and S-100A13 expression in PCM.

2. Materials and methods 2.1. Sample selection This study was approved by the Boston University School of Medicine Institutional Review Board (docket no. H31284). Archival tissues with a diagnosis of PCM (n = 102) were retrieved from the pathology files of the Skin Pathology Laboratory, Boston University School of Medicine (Boston, MA), between January 2010 and December 2012. Inclusion criteria were randomly selected cases of invasive PCM with a depth of at least 1 mm (a cut-off selected to facilitate quantification of intratumoral MVD). Histopathologic sections of all cases were reviewed by 2 board-certified dermatopathologists (initial sign out on all by a dermatopathologist; cases were then rereviewed and diagnosis confirmed by the senior author). All patient data were deidentified. The median age of the patients was 67 years (range, 19103 years) of which 70% (n = 72) were men. Mitosis was present in 90 of 102 and absent in 12 of 102. Host response was present in 48 of 102 and absent in 54 of 102. Ulceration was present in 28 of 102 and absent in 74 of 102. Regression (including partial or active regression and defined by the presence of fibrosis or a heavy lymphocytic infiltrate with loss or degeneration of tumor cells) was present in 81 of 102 and absent in 21 of 102. LVI detected by H&E stain was noted in 5 of 102 cases. All were of the American Joint Committee on Cancer (AJCC) clinical grade T2a and above at the time of initial diagnosis including 44-T2a, 9-T2b, 19-T3a, 16-T3b, 7-T4a, and 7-T4b. The median thickness of all tumors assessed was 2.4 mm (range, 1.1-8.3 mm).

2.2. Immunohistochemistry Immunohistochemistry was performed on 4-μm thickness formalin-fixed, paraffin-embedded sections using commercially available markers D2-40 at 1:200 (Dako, Carpinteria, CA), VEGFR-2 at 1:500 (clone 55B11; Cell Signaling Technology, Danvers, MA), S-100A13 at 1:500 (SigmaAldrich, St Louis, MO), S-100 at 1:1000 (Dako), and Endocan MEP14 at 1:500 (Lunginnov, Lille, France). Double immunolabeling was performed using EnVision DuoFLEX Doublestain System (Dako) in combination with Vector Blue AP Substrate Kit (Vector, Burlingame, CA) and a Methyl Green counterstain (Vector). All stained slides were reviewed and scored by the first author (P. A.) and the senior author (M. M.) in a blinded fashion, and any disagreements reviewed together to achieve a consensus score.

306 2.2.1. VEGFR-2 Staining of endothelial cells of dermal vessels served as the positive internal controls in each case, where they could be visualized. Within lesional tumoral cells, positive staining was noted by ascertaining cytoplasmic and/or nuclear staining. Immunohistochemical expression was evaluated with a semiquantitative scoring system by percentage of positive cells and intensity using the following scale for percentage positivity: 1, 0% to 5%; 2, 6% to 25%; 3, 26% to 50%; 4, 51% to 75%; and 5, 76% to 100% and the following scale for intensity: 1, weak staining; 2, moderate staining; and 3, strong staining. For purposes of statistical analysis, cases with a total score of 0 or less than 5% staining were counted as negative. 2.2.2. Endocan Staining in endothelial cells served as the positive external control in each case, where they could be visualized. Within lesional tumoral cells, positive staining was noted by ascertaining cytoplasmic and/or nuclear staining. Immunohistochemical expression was evaluated with a semiquantitative scoring system by percentage of positive cells and intensity using the same scoring scale as for VEGFR-2. 2.2.3. S-100A13 Staining in basal epidermal keratinocytes, outer and inner root sheath of the hair follicles, eccrine glands, sebaceous glands, dermal antigen-presenting cells, lymphocytes, and endothelial cells served as positive internal controls in each case, where they could be visualized. Within lesional tumoral cells, positive staining was noted by ascertaining cytoplasmic and/or nuclear staining. Staining in endothelial cells within the tumor could not be appreciated and therefore quantified, as tumoral cells were strongly and diffusely positive for S-100A13. Tumoral immunohistochemical expression was evaluated with a semiquantitative scoring system by percentage of positive cells and intensity using the following scale for percentage positivity: 0, 0% to 25%; 1+, 26% to 50%; 2+, 51% to 75%; and 3+, greater than 76% to 100% and the following scale for intensity: 1+, weak staining; 2+, moderate staining; and 3+, strong staining with a total score obtained by a sum of both. For purposes of statistical analysis, cases with a total score of 0 or less than 25% staining were counted as negative.

2.3. Assessment of mean MVD and LVD Intratumoral MVD (confined within the tumor border) was assessed as follows: the sections were scanned at low magnifications (×40 and ×100) to identify the most vascular areas of the tumor (designated “hotspots”) per a previously described technique, and 4 fields (×400 magnification, highpower field, 0.16 mm2 per field) were examined and the mean value calculated [22].

P. P. Aung et al.

2.4. LVI Slides from the same block were stained with H&E and double stained with D2-40 and S-100 to evaluate LVI, defined by the presence of tumor cells within the endotheliallined lumen of vessels. Only unequivocal cases of LVI were recorded as positive.

2.5. Genetic analyses In all PCMs included in the study, the lesional tissue comprised greater than 95% of the sample precluding potentially confounding issues relating to sensitivity as a function of a minimal tumoral component in sections analyzed [23]. DNA was extracted by proteinase K digestion. Briefly, 5-μm-thick sections were scratched off glass slides, and the formalin-fixed, paraffin-embedded archival tissue was deparaffinized and rehydrated, tissue pellets were then dried, and proteinase K solution was added and digested at 55°C for overnight. Crude DNA was boiled for 10 minutes and quantified by spectrophotometer and diluted to appropriate concentration. Sanger sequencing was performed on the BRAF gene exon 15 (forward gene coding strand only) spanning codon 600, using an ABI BigDye TerV3.1 cycle sequencing terminator ready reaction kit (Applied Biosystems, Foster City, CA) and performed on Genetic Analyzer 3130XL. The sequencing results were analyzed with ABI DNA Sequencing Analysis Software version 3.7. A positive and negative control was included in each batch of sequencing analysis.

2.6. Statistical analysis To determine which method of staining performed better, Kruskal-Wallis was used to assess the association between tumor staining score and MVD, whereas a sign test (M statistic) was used to differentiate as to staining intensity. The sign test is a nonparametric statistical procedure to detect a median difference (μd) when comparing 2 related (“paired”) samples, is conservative, and preferred when the underlying distribution may not be symmetric. Tumor scores of zero and one were collapsed into a single category so as to increase power. The distribution of LVD, MVD, and the percentage of samples showing LVI was examined by prognosticator to determine potential confounders. In addition, each prognosticator was examined as to its affect on the effect estimate for the association of LVI with both LVD and MVD using logistic regression. Before selection for multivariat analysis, the relationship between the covariates was examined. To help prevent collinearity in the final model, BRAF-mitosis and ulcerationthickness associations were checked using χ2 analysis. Likewise, the crude association between MVD and LVD was assessed using Spearman correlation. Furthermore,

MVD, LVD and LVI in melanoma statistical interaction of BRAF-mitosis and AJCC stage– ulceration was assessed using logistic regression. To minimize any residual confounding, a backward covariate selection method was used to determine the association between LVI and MVD (multiple logistic regression) and between MVD and LVD (multiple linear regression), to control for clinical prognosticators. This was then compared with a second analysis of the association between LVI and LVD, controlling for clinical prognosticators. These results were then used to guide examination of the association between LVD and MVD, using multiple linear regression to control for clinical prognosticators. To test if BRAF mutational status was an effect modifier, data were stratified by genotype into mutant and wild type. Separate analyses were run using multiple linear and logistic regression according to the methodology detailed above for each genotype group. Overall fit of multivariat models was done using the coefficient of determination (R2) and the F statistic for assessing MVD and LVD and Akaike Information Criterion (AIC) for logistic regression of MVD and LVD on LVI. All analyses were done using SAS software version 9.3 (Cary, NC).

3. Results 3.1. Immunostaining enhances the rate of detection of LVI in PCM A total of 101 PCMs were assessed for LVI using immunostaining for D2-40 and S-100 and H&E stain. The use of immunostaining increased the rate of LVI positivity by approximately 10-fold (47/101 cases, 46.5% using immunostaining versus 5/101 cases, 4.9% by H&E; P b .0001) (Fig. 1A-D). Statistically significant differences in LVI were noted only between the following groups: BRAF status and ulceration (P = .05 and P = .004 respectively) (Table 1).

3.2. VEGFR-2 is superior to Endocan in the assessment of MVD Using VEGFR-2, the overall mean MVD was 6 in the intratumoral area (Fig. 2A). Using Endocan, the overall mean intratumoral MVD was 0.15 (Fig. 2B). Using Kruskal-Wallis to assess for an association between immunohistochemical stain and presence of MVD, both VEGFR-2 (χ2 = 6.57; degrees of freedom [df] = 2; P = .037) and Endocan-stained (χ2 = 25.24; df = 3; P b .0001) tumors demonstrated at least 1 difference in the distribution of mean MVD scores. Although both immunohistochemical stains are theoretically associated with MVD, the VEGFR-2 staining score was almost 1 U higher compared with Endocan (μd = 0.86; M = 17; P b .0001) and identified an

307 average of almost 6 additional vascularities (μd = 5.86; M = 50.5; P b .0001). Therefore, to ascertain the correlation between MVD, LVD, and established histopathologic prognosticators and/or the BRAF status, we chose VEGFR2 to measure MVD.

3.3. MVD correlates with select histopathologic prognosticators but not the BRAF status In a bivariate analysis, MVD was higher in melanomas with depth greater than or equal to 2 mm and in melanomas with ulceration (P = .04 and 0.05, respectively) (Table 1). After adjusting for mean MVD, only ulceration was significantly associated with LVI (P = .02) with an almost 4fold increase of LVI in the presence of ulceration (odds ratio [OR] = 3.63) (Table 2). LVD was initially removed from the multivariat model, as it was not significantly associated with LVI after adjusting for all the other potential covariates. After controlling for BRAF mutation, thickness, host response, and ulceration, mean MVD was not significantly associated with LVI (OR = 1.00; 95% confidence interval [CI], 0.82-1.26).

3.4. LVD correlates with select histopathologic prognosticators but not the BRAF status In a bivariate analysis, LVD was significantly higher in melanomas with depth greater than or equal to 2 mm and in melanomas with mitoses (P = .03 and .02, respectively) (Table 1, Fig. 2C). Although not achieving statistical significance, a trend toward the same was noted with tumor AJCC stage of T3a and above (P = .08) (Table 1). After adjusting for mean LVD, only ulceration was significantly associated with LVI (P = .009), with an almost 4-fold increase in the likelihood of LVI, when ulceration is present (OR = 3.63; 95% CI, 1.38-9.53) (Table 3). This paralleled the multivariate analysis results for MVD and LVI. Here, MVD was initially removed, as it was not significantly associated with LVI after adjusting for all the other potential covariates. After controlling for potential confounding prognosticators, mean LVD was not significantly associated with LVI (OR = 1.02; 95% CI, 0.82-1.26) (Table 3). As neither MVD nor LVD was associated with LVI in our samples, we chose to exclude LVI from further consideration in the relationship between MVD and LVD.

3.5. LVD is associated with MVD after controlling for histopathologic prognosticators as well as BRAF status In the unadjusted analysis, MVD appeared to be associated with LVD (r = 0.37, P = .0002), and this relationship persisted after controlling for other clinical and histologic features. After adjusting for other prognosticators, LVD and MVD were independently associated (β = .28, P = .01). None of the

308

P. P. Aung et al.

Fig. 1 Representative cases demonstrating LVI, case 102 (A and B) and case 47 (C and D).Original magnification, H&E ×100 (A), ×40 (C); double immunostaining ×400 (B and D).

Table 1

LVD by D2-40 and MVD by VEGFR; correlation with histopathologic prognosticators and BRAF status (bivariat analysis)

Prognosticator

LVD b

n

P

MVD†

n

P

LVI % (n)

P

Present

Absent

a

BRAF Mutant Wild type Thickness b2 mm ≥2 mm AJCC staging T1-T2 T3a-T3b T4a-T4b Mitosis Present Absent Host response Present Absent Ulceration Present Absent Regression Present Absent

4.57 4.17

30 70

0.56

6.29 5.9

31 70

0.41

39.6 (19) 60.4 (29)

21.8 (12) 78.2 (42)

0.05

3.9 4.76

47 51

0.03

5.67 6.44

48 52

0.04

41.7 (20) 58.3 (28)

54.7 (29) 45.3 (24)

0.19

4.04 4.98 3.77

50 35 14

0.08

5.66 6.59 5.93

52 35 14

0.11

50.0 (24) 37.5 (18) 12.5 (6)

52.8 (28) 32.1 (17) 15.1 (8)

0.85

4.43 3.12

90 10

0.02

6.11 5.3

90 11

0.24

91.7 (44) 8.3 (4)

87.0 (47) 13.0 (7)

0.53

5.47 4.22

8 91

0.33

6.75 6.01

8 92

0.61

8.3 (4) 91.7 (44)

7.6 (4) 92.4 (49)

1.00

4.61 4.2

28 71

0.28

6.71 5.82

28 72

0.05

41.7 (20) 58.3 (28)

15.1 (8) 84.9 (45)

0.004

3.94 4.42

21 78

0.42

5.96 6.1

21 79

0.52

20.8 (10) 79.2 (38)

20.7 (11) 79.3 (42)

1.00

Abbreviations: LVD, lymphovascular density; MVD, microvessel density; LVI, lymphovascular invasion. NOTE. Missing cases for parameters assessed (either due to absence of tissue, lesional tumoral area, or inadequate DNA) are denoted as follows: a One case. b Two cases.

MVD, LVD and LVI in melanoma

309

Fig. 2 Representative images of immunohistochemical staining with VEGFR-2, Endocan, and S-100A13. MVD with VEGFR-2 (A) and Endocan (B). C, LVD with D2-40. D, Tumoral expression with S-100A13. Original magnification ×200 (A-C), ×100 (D).

histopathologic prognosticators or BRAF “status” was statistically significant (Table 4).

3.6. BRAF mutation is an effect measure modifier A BRAF mutation was seen in 31 (30.4%) of 102 of cases. These included mutations in BRAF V600E (n = 20), V600K (n = 6), D594A (n = 10), L597Q (n = 1), L597S (n = 1), K601E (n = 1), and V600R (n = 1). The association between LVD and MVD as well as MVD and a host response may be fundamentally different based on the presence of a BRAF mutation. Samples with wild-type BRAF showed no significant differences in mean MVD based on clinical prognosticators, nor showed an association between MVD and LVD (β = .14; 95% CI, −0.11 to 0.39;

Table 2 Correlation of MVD and LVI controlling for other histopathologic prognosticators and BRAF status (multivariat analysis) Predictor

OR

95% CI

P

Mean MVD BRAF mutant Thickness N2mm Host response Ulceration

1.003 2.09 1.26 1.23 3.63

0.82-1.26 0.82-5.32 0.53-2.98 0.27-5.70 1.38-9.53

.98 .12 .60 .79 .015

Abbreviations: CI, confidence interval; LVD, lymphovascular density; MVD, microvessel density; OR, odds ratio. NOTE. Overall model fit—AIC = 138.8, χ2 = 11.7; P = .039.

P = .28). In contrast, samples with any mutation in BRAF showed a significant increase in the mean MVD, as mean LVD increased (β = .59; 95% CI, 0.29-0.88; P = .0008) and a 3.7 fold increase in mean MVD among samples with a host response (β = 3.74; t = 2.24; df = 21; 95% CI, 0.46 -7.02; P = .04) (Table 5).

3.7. Tumoral expression of the VEGFR-2, Endocan, and S-100A13 Among all PCM samples, positive staining of VEGFR-2 was noted in 49 (48.5%) of 101 with a mean proportional positivity of 1.34 (Fig. 2A). Endocan/endothelial cell–specific molecule was positive in 17 (16.8%) of 101 cases with a mean proportional positivity of 0.48 (Fig. 2B).

Table 3 Correlation of LVD and LVI controlling for other histopathologic prognosticators and BRAF status (multivariat analysis) Predictor

OR

95% CI

P

Mean LVD BRAF mutant Host response Ulceration

1.02 1.92 1.27 3.63

0.82-1.26 0.75-4.92 0.27-5.91 1.38-9.53

.88 .18 .76 .009

Abbreviations: CI, confidence interval; LVD, lymphovascular density; LVI, lymphovascular invasion; OR, odds ratio. NOTE. Overall model fit—AIC = 134.9, χ2 = 10.8; P = .029.

310

P. P. Aung et al.

Positive staining with S-100A13 was noted in 101 (99%) of 102 cases with a mean proportional positivity of 4.7 (Fig. 2D). A significant association was noted only between S-100A13 expression and ulceration (P = .05).

4. Discussion Although a correlation between MVD and select histopathologic prognosticators has been demonstrated in malignant melanoma, results have been somewhat conflicting. Using a plethora of markers that include UEA-1, CD31, CD34, D2-40, and maspin increased vascularity at the base of deeper melanomas; and an unfavorable outcome has been reported [3-6,24-27]. Contrasting with these are studies refuting the same. Briefly, these include evidence showing that decreased MVD is present in thicker and more proliferative melanomas and that increased MVD is significantly associated with improved patient survival [7-10]. Arguing in favor of a role for MVD as a poor prognosticator is our finding of significantly increased MVD in tumors with depth greater than or equal to 2 mm and in tumors with ulceration and a significantly higher LVD in melanomas with depth greater than or equal to 2 mm and in melanomas with mitoses. Further in support of this, we found LVD to be associated with MVD after controlling for histopathologic prognosticators as well as BRAF status. To date, there has been a discrepancy in results regarding expression of VEGFR-2 in melanoma in different studies. Using the VEGFR-2–A3 antibody, expression has been noted in 79% to 89% of metastatic melanomas [28-31]. In contrast, using the 55B11 antibody, expression was noted in only 7% of metastatic melanomas. Of note in the same melanoma samples, using the VEGFR-2–A3 antibody expression was noted in 79% of the same. Using the 55B11 antibody, Miettinen et al [32] also noted no expression in any of their metastatic melanoma (0/107) or

Table 4 Correlation of MVD and LVD after controlling for other histopathologic prognosticators and BRAF status Predictor

Effect

SE

P

Mean LVD BRAF mutant Depth Ref: T1-T2 T3 T4 Mitosis Host response Ulceration Regression

0.275 0.364

0.105 0.468

.0104 .4383

– 0.401 0.667 0.421 0.547 0.606 0.092

– 0.477 0.643 0.703 0.758 0.467 0.507

– .4026 .3022 .5509 .4728 .1972 .8567

Abbreviations: LVD, lymphovascular density; MVD, microvessel density; Ref, reference. NOTE. Effect, difference of means (β estimate). Overall model fit—R2 = 16.5%, F = 2.19; P = .0353.

primary desmoplastic melanoma (0/17) samples. The differences are likely a consequence of epitopes against which the antibodies are directed. The VEGFR-2–A3 epitope raised against amino acids 1158 to 1345 maps at the C terminus of the Flk-1 of mouse origin, whereas the 55B11 was produced by immunizing rabbits with a recombinant protein containing the carboxyterminal 150 amino acid residues of human VEGFR-2. To date, VEGF has been the main target for an angiogenic ligand in melanoma. Preclinical studies have shown that the administration of bevacizumab, a humanized monoclonal IgG antibody directed against VEGF, alone did not appear to reduce tumor burden in patients with metastatic melanoma [33]. Using the 55B11 clone, tumoral cells in most samples in our study (96%) showed only focal and/or weak positivity with relatively low intensity. Our findings indicate that therapies targeting VEGFR are perhaps not efficacious for the simple reason that melanomas express VEGFR-2 in less than 25% of cells and that too weakly. The latter suggests that the density of expression of this particular antigen on melanoma cells is low. The incidence of LVI (0%-6%) in PCM is disproportionate to that of sentinel lymph node involvement, which ranges from 19% to 47% irrespective of tumor depth [12,15]. This discrepancy may be due to difficulties associated with ascertaining LVI by H&E alone, a factor of particular relevance in thick melanomas in which tumoral occlusion of vascular lumina can preclude detection. Findings from the

Table 5 BRAF status; an effect measure modifier of the association between MVD and LVD Predictor BRAF mutant a Mean LVD Depth Ref: T1-T2 T3 T4 Mitosis Host response Ulceration Regression Wild-type BRAF b Mean LVD Depth Ref: T1-T2 T3 T4 Mitosis Host response Ulceration Regression

Effect (β)

SE

P

0.589

0.151

.0008

– −0.387 1.022 1.285 3.744 1.052 1.098 0.138

– 0.703 1.032 1.241 1.674 0.684 0.698 0.127

– .5879 .3334 .3122 .0362 .1386 .1305 .283

– 0.969 0.438 0.073 −0.19 0.754 −0.649

– 0.593 0.757 0.798 0.846 0.593 0.669

– .1073 .5646 .9274 .8233 .2082 .3358

Abbreviations: LVD, lymphovascular density; MVD, microvessel density. a Overall model—R2 = 62.0%, F = 4.89; P = .0021. b Overall model—R2 = 12.7%, F = 1.27; P = .2827.

MVD, LVD and LVI in melanoma

311

Fig. 3 Schematic representation of study results. Results from bivariate analyses are represented by jagged lines; results from multivariate analyses are represented by solid lines; orange solid line represents multivariate analyses with effect measure modification.

current study confirm this, as the average depth of melanomas in which LVI was detected by immunohistochemical stain but not by H&E was 2.40 mm. Our results, in keeping with previous observations, appear to confirm that immunostaining increases detection rate of LVI [12-15,34]. In a multivariate analysis, after adjusting for both MVD and LVD, only ulceration was significantly associated with LVI. From a scientific perspective, this may in part be explained by the fact that ulceration is typically associated with enhanced expression of proangiogenic cytokines, such as platelet-derived growth factor and tumor growth factor [35]. Massi et al [21] first reported that expression of S-100A13 may represent a new angiogenic and prognostic marker for cutaneous melanoma, as S-100A13 protein expression was correlated positively with the intensity of VEGF-A staining both by immunohistochemical and real-time PCR and MVD by CD105. Like Massi et al [21], in our study, positive S100A13 diffuse and strong immunostaining was noted in 99% of our cases. However unlike Massi et al [21], we noted a significant correlation between tumoral S-100A13 protein expression and ulceration but not with MVD. A possible reason for the discrepancy may be due to differences in the specificity of these antibodies. In 1 study, the specificity of CD105 was 69% versus 81% of VEGFR-2 [36]. In conclusion, findings from the current study indicate that MVD and LVD are not associated with LVI, appear to closely relate with each other, and are associated with select markers of poor prognosticative value (Fig. 3). The association between host response and both MVD and LVD, present solely in the context of BRAF mutation, suggests that LVD/MVD exhibits potential for strategizing immunotherapies in PCM. Given that this association is independent of LVI, the existence of 2 separate mechanisms for metastasis cannot be entirely excluded. However, correlation with follow-up data is required to support this presumption.

References [1] Preziosi R, Sarli G, Paltrinieri M. Prognostic value of intratumoral vessel density in cutaneous mast cell tumors of the dog. J Comp Pathol 2004;130:143-51. [2] O'Grady A, O'Kelly P, Murphy GM, Leader M, Kay E. COX-2 expression correlates with microvessel density in non-melanoma skin cancer from renal transplant recipients and immunocompetent individuals. HUM PATHOL 2004;35:1549-55. [3] Srivastava A, Laidler P, Davies RP, Horgan K, Hughes LE. The prognostic significance of tumor vascularity in intermediate-thickness (0.76-4.0 mm thick) skin melanoma. A quantitative histologic study. Am J Pathol 1988;133:419-23. [4] Graham CH, Rivers J, Kerbel RS, Stankiewicz KS, White WL. Extent of vascularization as a prognostic indicator in thin (b0.76 mm) malignant melanomas. Am J Pathol 1994;145:510-4. [5] Rongioletti F, Miracco C, Gambini C, Pastorino A, Tosi P, Rebora A. Tumor vascularity as a prognostic indicator in intermediate-thickness (0.76-4 mm) cutaneous melanoma. A quantitative assay. Am J Dermatopathol 1996;18:474-7. [6] Dadras SS, Paul T, Bertoncini J, et al. Tumor lymphangiogenesis: a novel prognostic indicator for cutaneous melanoma metastasis and survival. Am J Pathol 2003;162:1951-60. [7] Straume O, Jackson DG, Akslen LA. Independent prognostic impact of lymphatic vessel density and presence of low-grade lymphangiogenesis in cutaneous melanoma. Clin Cancer Res 2003;9:250-6. [8] Kerr EH, Wang D, Lewis JSJR, Said-Al-Naief N, Hameed O. Lack of correlation between microvascular density and pathological features and outcomes in sinonasal and oral mucosal melanomas. Head Neck Pathol 2011;5:199-204. [9] Carnochan P, Briggs JC, Westbury G, Davies AJ. The vascularity of cutaneous melanoma: a quantitative histological study of lesions 0.851.25 mm in thickness. Br J Cancer 1991;64:102-7. [10] Busam KJ, Berwick M, Blessing K, et al. Tumor vascularity is not a prognostic factor for malignant melanoma of the skin. Am J Pathol 1995;147:1049-56. [11] Watanabe S, Kato M, Kotani I, Ryoke K, Hayashi K. Lymphatic vessel density and vascular endothelial growth factor expression in squamous cell carcinomas of lip and oral cavity: a clinicopathological analysis with immunohistochemistry using antibodies to D2-40, VEGF-C and VEGF-D Yonago. Acta Med 2013;56:29-37.

312 [12] Rose AE, Christos PJ, Lackaye D, et al. Clinical relevance of detection of lymphovascular invasion in primary melanoma using endothelial markers D2-40 and CD34. Am J Surg Pathol 2011;35:1441-9. [13] Niakosari F, Kahn HJ, Marks A, Form L. Detection of lymphatic invasion in primary melanoma with monoclonal antibody D2-40: a new selective immunohistochemical marker of lymphatic endothelium. Arch Dermatol 2005;141:440-4. [14] Petitt M, Allison A, Shimoni T, Uchida T, Raimer S, Kelly B. Lymphatic invasion detected by D2-40/S-100 dual immunohistochemistry does not predict sentinel lymph node status in melanoma. J Am Acad Dermatol 2009;61:819-28. [15] Doeden K, Ma Z, Narasimhan B, Swetter SM, Detmar M, Dadras SS. Lymphatic invasion in cutaneous melanoma is associated with sentinel lymph node metastasis. J Cutan Pathol 2009;36:772-80. [16] Zettersten E, Sagebiel RW, Miller III JR, Tallaperaeddy S, Leong SP, Kashani-Sabet M. Prognostic factors in patients with thick cutaneous melanoma (N 4 mm). Cancer 2002;94:1049-56. [17] Egger ME, Gilbert JE, Burton AL, et al. Lymphovascular invasion as a prognostic factor in melanoma. Am Surg 2011;77:992-7. [18] Molhoek KR, Erdag G, Rasamny JK, et al. VEGFR-2 expression in human melanoma: revised assessment. Int J Cancer 2011;129:2807-15. [19] Straume O, Akslen LA. Expression of vascular endothelial growth factor, its receptors (FLT-1, KDR) and TSP-1 related to microvessel density and patient outcome in vertical growth phase melanomas. Am J Pathol 2001;159:223-35. [20] Hendrix MJC, Seftor EA, Hess AR, Seftor RE. Molecular plasticity of human melanoma cells. Oncogene 2003;22:3070-5. [21] Massi D, Landriscina M, Piscazzi A, et al. S100A13 is a new angiogenic marker in human melanoma. Mod Pathol 2010;23:804-13. [22] Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis and metastasis—correlation in invasive breast carcinoma. N Engl J Med 1991;324:1-8. [23] Feller K, Yang S, Mahalingam M. Immunohistochemistry with a mutation-specific monoclonal antibody as a screening tool for the BRAFV600E mutational status in primary cutaneous malignant melanoma. Mod Pathol 2013;26:414-20. [24] Barnhill RL, Piepkorn MW, Cochran AJ, Flunn E, Karaoli T, Folkman J. Tumor vascularity, proliferation, and apoptosis in human melanoma micrometastases and macrometastases. Arch Dermatol 1998;134:991-4.

P. P. Aung et al. [25] Chua R, Setzer S, Govindarajan B, Sexton D, Cohen C, Arbiser JL. Maspin expression, angiogenesis, prognostic parameters, and outcome in malignant melanoma. J Am Acad Dermatol 2009;60:758-66. [26] Ribatti D, Nico B, Floris C, et al. Microvascular density, vascular endothelial growth factor immunoreactivity in tumor cells, vessel diameter and intussusceptive microvascular growth in primary melanoma. Oncol Rep 2005;14:81-4. [27] Depasquale I, Thompson WD. Microvessel density for melanoma prognosis. Histopathology 2005;47:186-94. [28] Mehnert JM, McCarthy MM, Jilaveanu L, et al. Quantitative expression of VEGF, VEGF-R1, VEGF-R2, and VEGF-R3 in melanoma tissue microarrays. HUM PATHOL 2010;41:375-84. [29] Straume O, Akslen LA. Importance of vascular phenotype by basic fibroblast growth factor, and influence of the angiogenic factors basic fibroblast growth factor/fibroblast growth factor receptor-1 and ephrinA1/EphA2 on melanoma progression. Am J Pathol 2002;160:1009-19. [30] Salven P, Heikkilä P, Joensuu H. Enhanced expression of vascular endothelial growth factor in metastatic melanoma. Br J Cancer 1997;76:930-4. [31] Pisacane AM, Risio M. VEGF and VEGFR-2 immunohistochemistry in human melanocytic naevi and cutaneous melanomas. Melanoma Res 2005;15:39-43. [32] Miettinen M, Rikala MS, Rys J, Lasota J, Wang ZF. Vascular endothelial growth factor receptor 2 as a marker for malignant vascular tumors and mesothelioma: an immunohistochemical study of 262 vascular endothelial and 1640 nonvascular tumors. Am J Surg Pathol 2012;36:629-39. [33] Varker KA, Biber JE, Kefauver C, et al. A randomized phase 2 trial of bevacizumab with or without daily low-dose interferon alfa-2b in metastatic malignant melanoma. Ann Surg Oncol 2007;14:2367-76. [34] Petersson F, Diwan AH, Ivan D, et al. Immunohistochemical detection of lymphovascular invasion with D2-40 in melanoma correlates with sentinel lymph node status, metastasis and survival. J Cutan Pathol 2009;36:1157-63. [35] Murphy MO, Ghosh J, Fulford P, et al. Expression of growth factors and growth factor receptor in non-healing and healing ischaemic ulceration. Eur J Vasc Endovasc Surg 2006;3:516-22. [36] Marioni G, Giacomelli L, D'Alessandro E, et al. Laryngeal carcinoma recurrence rate and disease-free interval are related to CD105 expression but not to vascular endothelial growth factor 2 (Flk-1/Kdr) expression. Anticancer Res 2008;28:551-7.

Microvessel density, lymphovascular density, and lymphovascular invasion in primary cutaneous melanoma-correlation with histopathologic prognosticators and BRAF status.

The relationship between microvessel density (MVD), lymphovascular density (LVD), and lymphovascular invasion (LVI) in primary cutaneous melanoma (PCM...
1MB Sizes 0 Downloads 5 Views