CD44v3 levels in primary cutaneous melanoma are predictive of prognosis: assessment by the use of tissue microarray.
ABSTRACT Despite the use of sentinel node biopsy techniques, the search continues for other strategies to improve the accuracy of estimating prognosis in melanoma patients. Various biomarkers have previously been studied for use in this role, but none has yet achieved acceptance in routine practice. We have applied the novel technology of tissue microarray for the high throughput screening of a cohort of 120 primary cutaneous melanoma specimens for expression of the transmembrane glycoprotein CD44, splice variant 3 (v3), which has previously been implicated in tumor progression. A highly significant correlation between CD44v3 expression and Breslow thickness, Clark's level and patient age was demonstrated (Spearman correlation p < 0.001). Regarding clinical outcome, CD44v3 expression was shown to be significantly associated with better outcome (chi(2) = 7.2219, p = 0.0072). Furthermore, subgroup analysis revealed a sequentially improved survival probability associated with the intensity of CD44v3 staining (chi(2) = 12.5162, p = 0.0058). Analysis in a Cox multivariate model, however, did not show CD44v3 to be independently predictive of prognosis. The implications of these findings are considered, and the use of CD44v3 as a potential prognostic marker or a target for therapeutic manipulation are discussed.
- SourceAvailable from: PubMed Central[Show abstract] [Hide abstract]
ABSTRACT: Background and purpose: CD44 variants have been associated with tumor invasion and metastasis, but CD44 expression patterns have not been systematically investigated in pancreatic carcinoma. This study systematically investigated whether CD44 expression patterns are involved in pancreatic carcinoma metastasis and prognosis. We applied primers specific for all CD44 variants and CD44s to analyze the expression patterns of CD44 (CD44v2-CD44v10 and CD44s) using quantitative real-time PCR (qRT-PCR). We then further evaluated their roles in pancreatic carcinoma metastasis and prognosis using clinical survival analysis. Increased CD44v expression and decreased CD44s expression were found in metastatic pancreatic carcinoma in three different cell lines and in human tumor tissue. Clinical analysis showed that CD44v6+ and CD44v9+ were correlated with lymph node metastasis, liver metastasis and TNM stage. However, CD44s- was associated with liver metastasis, tumor differentiation and TNM stage. Survival analysis showed that patients with CD44v6+/CD44s- or CD44v6+/CD44s- had lower overall survival (OS) rates, although the individual expression of CD44v6, CD44v9 and CD44s was also related to decreased OS rates. Univariate analysis showed that lymph node metastasis; vessel invasion; hepatic metastases; TNM stage; and individual or co-expression of CD44v6, CD44v9 and CD44s were risk factors affecting survival. Multivariate analysis showed that CD44v6+/CD44s- was an independent predictor of survival. We found that CD44v6+, CD44v9+ and CD44s- were associated with pancreatic carcinoma metastasis and progression and that CD44v6+/CD44s- was an independent risk factor affecting survival in pancreatic carcinoma. Therefore, the different expression patterns of CD44v/CD44s may determine pancreatic carcinoma prognosis.Virtual slides: The virtual slide(s) for this article can be found here: http://www.diagnosticpathology.diagnomx.eu/vs/1579257224116287.Diagnostic Pathology 04/2014; 9(1):79. · 2.41 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Cutaneous melanoma (CM) causes the greatest number of skin cancer-related deaths worldwide. Predicting CM prognosis is important to determine the need for further investigation, counseling of patients, to guide appropriate management (particularly the need for postoperative adjuvant therapy), and for assignment of risk status in groups of patients entering clinical trials. Since recurrence rate is largely independent from stages defined by morphological and morphometric criteria, there is a strong need for identification of additional robust prognostic factors to support decision-making processes. Most data on prognostic biomarkers in melanoma have been evaluated in tumor tissue samples by conventional morphology and immunohistochemistry (IHC) as well as DNA and RNA analyses. In the present review, we critically summarize main high-quality studies investigating IHC-based protein biomarkers of melanoma outcome according to Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK)-derived criteria. Pathways have been classified and conveyed in the "biologic road" previously described by Hanahan and Weinberg. Data derived from genomic and transcriptomic technologies have been critically reviewed to better understand if any of investigated proteins or gene signatures should be incorporated into clinical practice or still remain a field of melanoma research. Despite a wide body of research, no molecular prognostic biomarker has yet been translated into clinical practice. Conventional tissue biomarkers, such as Breslow thickness, ulceration, mitotic rate and lymph node positivity, remain the backbone prognostic indicators in melanoma.Archiv für Pathologische Anatomie und Physiologie und für Klinische Medicin 02/2014; · 2.68 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Context.-The distribution of the standard melanoma antibodies S100, HMB-45, and Melan-A has been extensively studied. Yet, the overlap in their expression is less well characterized. Objectives.-To determine the joint distributions of the classic melanoma markers and to determine if classification according to joint antigen expression has prognostic relevance. Design.-S100, HMB-45, and Melan-A were assayed by immunofluorescence-based immunohistochemistry on a large tissue microarray of 212 cutaneous melanoma primary tumors and 341 metastases. Positive expression for each antigen required display of immunoreactivity for at least 25% of melanoma cells. Marginal and joint distributions were determined across all markers. Bivariate associations with established clinicopathologic covariates and melanoma-specific survival analyses were conducted. Results.-Of 322 assayable melanomas, 295 (91.6%), 203 (63.0%), and 236 (73.3%) stained with S100, HMB-45, and Melan-A, respectively. Twenty-seven melanomas, representing a diverse set of histopathologic profiles, were S100 negative. Coexpression of all 3 antibodies was observed in 160 melanomas (49.7%). Intensity of endogenous melanin pigment did not confound immunolabeling. Among primary tumors, associations with clinicopathologic parameters revealed a significant relationship only between HMB-45 and microsatellitosis (P = .02). No significant differences among clinicopathologic criteria were observed across the HMB-45/Melan-A joint distribution categories. Neither marginal HMB-45 (P = .56) nor Melan-A (P = .81), or their joint distributions (P = .88), was associated with melanoma-specific survival. Conclusions.-Comprehensive characterization of the marginal and joint distributions for S100, HMB-45, and Melan-A across a large series of cutaneous melanomas revealed diversity of expression across this group of antigens. However, these immunohistochemically defined subclasses of melanomas do not significantly differ according to clinicopathologic correlates or outcome.Archives of pathology & laboratory medicine 08/2013; 137(8):1063-73. · 2.78 Impact Factor
CD44v3 levels in primary cutaneous melanoma are predictive of prognosis:
Assessment by the use of tissue microarray
Marc D. Pacifico1*, Rajiv Grover1, Paul I. Richman1, Frances M. Daley2, Francesca Buffa2and George D. Wilson3
1The RAFT institute of Plastic Surgery, Mount Vernon Hospital, Northwood, Middlesex HA6 2RN, United Kingdom
2The Gray Cancer Institute, Mount Vernon Hospital, Northwood, Middlesex HA6 2RN, United Kingdom
3Karmanos Cancer Institute, Wayne State University, Hudson Webber Building, Detroit, MI, USA
Despite the use of sentinel node biopsy techniques, the search con-
tinues for other strategies to improve the accuracy of estimating
prognosis in melanoma patients. Various biomarkers have previ-
ously been studied for use in this role, but none has yet achieved
acceptance in routine practice. We have applied the novel technol-
ogy of tissue microarray for the high throughput screening of a
cohort of 120 primary cutaneous melanoma specimens for expres-
sion of the transmembrane glycoprotein CD44, splice variant 3
(v3), which has previously been implicated in tumor progression.
A highly significant correlation between CD44v3 expression and
Breslow thickness, Clark’s level and patient age was demonstrated
(Spearman correlation p < 0.001). Regarding clinical outcome,
CD44v3 expression was shown to be significantly associated with
better outcome (v25 7.2219, p 5 0.0072). Furthermore, subgroup
analysis revealed a sequentially improved survival probability
associated with the intensity of CD44v3 staining (v25 12.5162,
p 5 0.0058). Analysis in a Cox multivariate model, however, did
not show CD44v3 to be independently predictive of prognosis. The
implications of these findings are considered, and the use of
CD44v3 as a potential prognostic marker or a target for therapeu-
tic manipulation are discussed.
' 2005 Wiley-Liss, Inc.
Key words: CD44v3; prognosis; melanoma; tissue microarray
Melanoma continues to present an increasing clinical problem:
its incidence continues to rise1whilst there is still no effective
standard adjuvant therapy available following surgical excision.
Novel techniques have been developed for establishing accurate
estimation of prognosis, most importantly sentinel lymph node
biopsy (SLNB). However, SLNB and completion lymphadenec-
tomy following a positive result have been implicated in an
increased risk of local and in-transit recurrence.2–4Furthermore,
studies have shown that up to 50% of patients who develop meta-
stases do so directly to distant sites or directly to ‘‘in-transit’’
sites.5Hence, in these patients, SLNB is of little value in predict-
ing outcome. A method of accurately predicting the natural history
of disease progression and time-course in an individual patient
still remains to be established.
Currently, prognosis is estimated by using established clinico-
pathological variables such as Breslow thickness, presence of
ulceration and Clark’s level amongst others.6SLNB is used as an
adjunct but has not yet gained universal acceptance for the reasons
discussed earlier. If it were possible to increase the armoury of
laboratory tests available to more precisely predict patient out-
come, it would enable more accurate planning of individual fol-
low-up regimes. This would be of incalculable psychological ben-
efit to the patient and, with the promise of novel adjuvant thera-
pies, may identify patient subgroups that could benefit from early
Metastatic spread is a complex process thought to occur at the
onset of the vertical growth phase of the melanoma.7It consists of
a series of carefully orchestrated events involving cell–cell and
cell–matrix interactions. CD44, a widely distributed transmem-
brane glycoprotein, is an adhesion molecule that has been impli-
cated in such interactions.8Multiple isoforms of CD44 may be
generated by alternative splicing of transcripts of this gene.9Those
containing the v3 domain are able to bind growth factors and
might play a role in the control of cell proliferation.10There is
increasing evidence that CD44 may contribute to malignant pro-
gression and metastasis formation.11In support of this concept, it
has been found that the extent of CD44 variant 6 (v6) correlates
with the stage of tumor progression in human colorectal adenocar-
cinomas,12bladder carcinoma13and non-Hodgkin’s lymphoma.14
The data suggest that the mechanism involved may be largely
dependent on the ability of CD44 to mediate cell attachment to
hyaluronate, an extracellular polysaccharide,15which contributes
to tumor cell proliferation, migration, invasion and formation of
metastatic tumor emboli.16In the case of melanoma cells, CD44
expression has been shown to relate to both tumorigenicity and
the potential for experimental metastasis formation.17Previous
studies examining the expression of several CD44 variants have
demonstrated upregulation of CD44H in benign lesions, dysplastic
lesions and melanoma in situ, but reduced expression within mela-
noma with increasing invasiveness (i.e. Clark’s levels), rather than
Breslow thickness.18,19Although various splice variants, including
v5, v7, v8 and v10, have been reported as being involved in mela-
noma progression,20,21the data regarding the v3 splice variant is
as yet unconvincing.18,21
The advent of tissue microarray technology has enabled high-
throughput screening to ascertain patterns of protein expression in
a large number of tissue specimens simultaneously,22overcoming
some of the time consuming aspects of conventional immunohis-
tochemistry. Using this strategy, small cores are retrieved from
selected regions of archival tissue blocks and are then precisely
arranged in a new paraffin block, which is sectioned and stained.
The aim of this study was to use tissue microarray technology to
screen archival samples of primary cutaneous melanoma from
patients on whom long-term follow-up data was available for
expression of CD44v3 and investigate any correlations with clini-
copathological variables and survival.
Patients and methods
The records of melanoma patients with a minimum 7-year fol-
low up (range 7–10 years) were sought from the pathology records
of Mount Vernon Hospital. A database of 222 primary melanoma
patients was created and relevant paraffin-embedded primary mel-
anoma specimens retrieved from the pathology archives at Mount
Vernon Hospital. Patients included were of stage IA–III at the
time of initial presentation, further details of patient characteristics
are shown in Table I. The database contained information on
patient demographics, melanoma disease course and patient out-
come. Details of established clinicopathological predictors of
prognosis including Breslow depth, Clark’s level, site of mela-
noma, patient age, patient sex and presence of ulceration were
Grant sponsors: The Restoration of Appearance and Function Trust
(RAFT) and The Royal College of Surgeons of England.
*Correspondence to: Department of Plastic Surgery, Mount Vernon
Hospital, Northwood, Middlesex HA6 2RN, UK.
Fax: 144 (0)1923-844-031. E-mail: email@example.com
Received 6 March 2005; Accepted after revision 29 July 2005
Int. J. Cancer: 118, 1460–1464 (2006)
' 2005 Wiley-Liss, Inc.
Publication of the International Union Against Cancer
All blocks were sectioned and stained with Haematoxylin and
Eosin (H&E). A consultant histopathologist [PIR] reviewed the
slides for quality and presence of melanoma in addition to reas-
sessing the depth recorded in the original pathology report. Poor
quality specimens with insufficient melanoma were not included
in the array. Areas of melanoma, distinct from surrounding normal
skin, were identified and marked for subsequent sampling.
Construction of the tissue microarray
Four core biopsies were randomly obtained from the paraffin-
embedded tissue within the previously identified and marked area.
Cores were transferred to a recipient paraffin block using the tech-
nique described by Kononen et al.,22using a Tissue Micro-
arrayerTM(Beecher Instruments, Silver Spring, Maryland, USA).
We have previously described this technique in detail.23Spacing
between adjacent samples on the array was 0.8 mm, and a sam-
pling needle with a caliber of 0.6 mm was used. The 4 cores from
each specimen were arranged adjacently. An Excel file (Microsoft,
USA) was simultaneously constructed with the coordinates during
the construction of the array. Following the construction of the tis-
sue microarray, the first slide produced was stained with H&E to
ensure only melanoma was sampled, distinct from the surrounding
epithelial cells, or as near to this as possible.
The master block was sectioned and slides of 5 lm thickness
produced. Slides were stained using the Dako Envision method
(DAKO Ltd., Cambridgeshire, UK). Formalin fixed paraffin
embedded sections were dewaxed in xylene for 5 min before rehy-
dration through graded alcohols (100, 90 and 70%) to water. Anti-
gen retrieval was performed on the slides by placing them in a
bath of 10 mM citric acid pH 6 and boiling for 15 min (3 3 5 min),
using an 800 W microwave oven (2450 MHz Panasonic NN-
6453BBPQ, Japan). The volume of fluid was topped up to its
original level and the slides were then left to stand for 20 min at
room temperature, before being washed well in running tap water.
Slides were then transferred to the Dako Autostainer (DAKO,
UK) containing Peroxidase block (DAKO, S2023) the detection
reagents (Chemate Envision HRP, DAKO K5007, Cambridge-
shire, United Kingdom) and anti-human CD44v3 antibody (Nova-
castra Labs Ltd. Newcastle, UK, NCL-CD44v3 clone VFF-327v3)
diluted 1/25 in antibody diluent (DAKO, UK, S2023). The Auto-
stainer programme included 5 min in peroxidase block, 1-hr incu-
bation in primary antibody, 30-min incubation in Envision poly-
mer and 5 min in Vector Labs SG HRP substrate (Vector Labs,
Peterborough, UK SK-4700); which appeared grey at the end of
staining. When the programme was complete, stained slides were
removed from the machine and counterstained in Nuclear fast red
(Vector Labs UK H3403) for 60 sec. Slides were then washed in
water, dehydrated in graded alcohols (70, 90 and 100%) cleared in
xylene and mounted in DPX (Surgipath Europe Ltd 08600E,
Peterborough, United Kingdom). Each staining run incorporated a
positive control slide of human tonsil as recommended as a posi-
tive control by the primary antibody manufactures of which the
membrane of the tonsil epithelial cells were positive. A negative
control was also incorporated and involved the substitution of the
anti-human CD44v3 antibody for an isotypic control antibody at
the same protein concentration.
Assessment of staining
The slides were scored by 2 independent observers without
prior knowledge of the clinical outcome for staining positivity and
stain intensity. Cores were scored as positive if any degree of
CD44v3 staining was evident within the melanoma cells. The tis-
sue array was then scanned and a semi-quantitative scoring system
was constructed for staining intensity based on the variation across
the specimens. This consisted of being classified as either nega-
tive, weak, moderate or strongly staining, as compared with a pos-
itive control of human tonsil tissue. The stain intensity score was
taken as the strongest staining sample from the 4 biopsies that rep-
resented each tumor. Examples are shown in Figure 1.
Assessment of association between CD44v3 and clinicopatho-
logical features was carried out using Chi-squared analysis. Uni-
variate analysis was undertaken using the Kaplan Meier method
with the JMP1statistical package (Version 4.0, SAS Institute,
USA) with stratification according to CD44v3 intensity. Log rank
tests were performed on the data, which were then analyzed using
FIGURE 1 – Figure illustrating examples of tissue array cores and
the positive control: (clockwise from top left) negative core, weakly
positive core (with some membrane staining), strongly staining core
and the positive control of human tonsil.
TABLE I – DETAILS OF PATIENTS AND MELANOMAS
INCLUDED IN THE MICROARRAY
Number of patients in microarray
Mean age at diagnosis
Site of melanoma
Head and neck
AJCC stage of patient at initial presentation
Tumour subtypes represented
Lentigo maligna melanoma
Mean Clark’s level
Mean tumour thickness (mm)
Median tumour thickness (mm)
Number of ulcerated melanoma
Number developing regional recurrence
Number of patients surviving in cohort
59.86 (range 21–97)
3 (range 2–5)
2.93 (range 0.1–35)
1Unless indicated otherwise, values in parentheses indicate percen-
CD44V3 AS A PROGNOSTIC MARKER FOR MELANOMA
v2tests to obtain p values. Cox’s multivariate analysis was per-
formed using SPSS statistical package (SPSS, Chicago, Illinois,
USA). CD44v3 was scored as negative and positive in the expres-
sion analysis, and as negative (0), weak, (1) moderate (2) and
strong (3) in the intensity analysis. Site was scored using 4 catego-
ries: head and neck (1), trunk (2), upper limbs (3) and lower limbs
(4). Ulceration was scored as ulcerated (0) or not ulcerated (1).
Gender was scored as female (0) or male (1). Age, Clark’s level
and Breslow thickness were considered as continuous variables.
Clinicopathological features of array specimens
Although all 222 paraffin-embedded specimens were success-
fully sectioned and stained with H&E, only 120 had enough
archival material available for the present study and were, there-
fore, implanted into the recipient block. Suitable areas of tumor
were identified and the tissue array was successfully constructed.
H&E staining of the array revealed that further specimens were
unsuitable for analysis because of either lack of melanoma tissue
(sampling error) or loss from the block. This meant that 84 speci-
mens were assessed from 4 cores, 26 from 3 cores and 8 from 2
specimens and in 2 specimens no cores were quantifiable. Detailed
analysis was performed on the melanomas from which 4 cores
were available, this number having been shown to be accurately
representative of the whole melanoma specimen.23Table I illus-
trates details of the patients and the specimens included in the tis-
sue microarray. Notably, the median tumor thickness is consider-
ably lower than the mean tumor thickness, which was skewed by 6
very thick tumors (ranging between 10 and 35 mm in thickness).
Without those tumors, the mean tumor thickness was calculated as
1.96 mm. The majority of tumors included in the array were less
than 1 mm in thickness. The inter-observer disagreement when
analyzing the stained microarray was 5.6%.
CD44v3 expression and correlation with clinicopathological
Fifty seven of the 84 tumor specimens were positive for
CD44v3 (67.9%). No stromal cell positivity was incorporated into
the results when evaluating the melanoma cores. Melanoma cell
positivity was confined to the cell membrane. The expression of
CD44v3 was highly correlated with Clark’s level, Breslow thick-
ness and patient age (Table II). The strong correlation between
CD44v3 expression intensity and both Clark’s level and Breslow
thickness (Spearman correlation p < 0.001 for both) is illustrated
in Figures 2 and 3. Melanomas with the strongest CD44v3 staining
intensity were all less than 4.0 mm in depth. Similarly, the propor-
tion of Clark’s level 5 tumors was highest in the negatively stained
group, and decreased in proportion with increasing intensity of
staining. The observed staining pattern is consistent with a role of
CD44v3 expression early in melanoma progression.
CD44v3 expression and clinical outcome
Figure 4 shows univariate survival curves in relation to CD44v3
expression. The log rank test of significance between Kaplan-
Meier estimates of survival showed that those patients whose
tumors were CD44v3 negative had a significantly higher probabil-
TABLE II – CORRELATIONS BETWEEN CD44V3 STAINING INTENSITY AND
ESTABLISHED PROGNOSTIC VARIABLES FOR MELANOMA
Site of tumour2
1Spearman test.–2Pearson v2test.
FIGURE 2 – Figure illustrating the association between the intensity
of CD44v3 expression and Breslow depth (Spearman’s correlation
p < 0.001). Breslow depth was grouped according to the ‘‘T’’ staging
of the AJCC staging classification of melanoma. The staining intensity
was scored as negative (0), weak (1), moderate (2) or strong (3).
FIGURE 4 – Survival curves comparing CD44v3 positive and CD44v3
negative melanomas. Log rank v25 7.2219, p 5 0.0072. —— CD44v3
negative, - - - - CD44v3 positive.
FIGURE 3 – Figure illustrating the association of the intensity of
CD44v3 expression with Clark’s levels (Spearman’s correlation p <
0.001). The staining intensity was scored as negative (0), weak (1),
moderate (2) or strong (3).
PACIFICO ET AL.
ity of poor outcome than those whose tumors were CD44v3 posi-
tive (v25 7.2219, p 5 0.0072). The 5-year survival for the
CD44v3 negative melanoma patients (n 5 27) was 43% compared
with a 5-year survival of 75% for the CD44v3 positive melanoma
patients (n 5 57). Subgroup stratification of the CD44v3 positive
tumors according to the intensity of expression (Figure 5) demon-
strated that the stronger the expression of CD44v3 in the tumor,
the better the patient outcome. The correlation with outcome was
again highly statistically significant (v25 12.5162, p 5 0.0058).
When CD44v3 positive tumors were analyzed with other estab-
lished clinicopathological variables of melanoma prognosis in a
Cox multivariate model, it was shown not to be independently pre-
dictive of prognosis (Table III).
This study was prompted by the need to discover more accurate
strategies for the prediction of prognosis in melanoma. The ability
to construct a tissue microarray has considerably advanced the
screening of prognostic markers.22Having previously validated
the use of tissue microarray technology for use in melanoma,23we
used this technique to examine the relationship of CD44v3 and
patient outcome in cases of primary melanoma with long-term fol-
low-up. On the basis of the validation study, the analysis of
marker expression and clinical outcome was restricted to only
those specimens (120 patients) where sufficient material was
available to array 4 cores per specimen. After construction of the
array, further specimens were excluded due to lack of tumor mate-
rial in the core or loss of the core. This highlights the problems of
working with mainly thin (<1 mm) melanoma specimens where
just over half of the specimens identified were suitable for the
array and a further 30% were lost because of technical problems
associated with arraying and sectioning.
A further point of discussion regarding the use of the tissue
microarray in evaluating immunohistochemical markers is the
method of assessing outcome. Traditional use of whole specimen
sections allows for an estimation of the percentage of specimen
staining positive for a particular marker. However, evaluation of
the biopsy cores in the array does not permit this. Whilst the 4
cores are deemed to be representative of the whole specimen, they
do not provide enough material for the estimation of percentage of
specimen staining positive, but, in contrast, they allow for an
assessment of positivity of staining (i.e. positive or negative) and
an assessment of the maximum strength of positivity observed
amongst the 4 cores studied. We feel that this is a valid investiga-
tive strategy, applicable to the clinical situation: if the marker
were to be used by a clinical histopathologist in the hospital set-
ting, his or her subjective interpretation of the degree of positivity
of a specimen would be used for outcome results.
The aim of this study was to examine the possibility of using
CD44v3 as a prognostic marker in cases of primary cutaneous
melanoma. The results demonstrated a significant relationship
between both expression and intensity of expression and patient
outcome (p 5 0.0072 and 0.0058, respectively). A strong correla-
tion between CD44v3 expression and both Breslow depth, Clark’s
level and patient age (p < 0.001) was also found, which is consis-
tent with the multivariate analysis revealing that CD44v3 was not
independently predictive of survival.
CD44 was originally described as a homing receptor for lym-
phocytes24and has since been shown to have a role in such diverse
physiological processes as lymphocyte activation, cell–cell adhe-
sion, cell–matrix interaction and cell motility.8Interest has
focused on its possible role as an adhesion molecule involved in
the malignant progression and metastatic potential of a variety of
tumors.18,25–28Experimental studies have demonstrated that the
metastatic capacity of B16 mouse melanoma cells29and human
melanoma cell lines17is enhanced by CD44H expression and this
capacity can be blocked by anti-CD44 antibodies.30Characteriza-
tion of the human CD44 gene demonstrated that it has at least 20
exons and is located on chromosome 11. Multiple isoforms are
generated by alternative splicing of transcripts of this gene.9The
larger splice variants (CD44v1–10) contain additional amino acids
positioned in the extracellular domain of the molecule. They have
a more restricted pattern of expression and appear to be preferen-
tially associated with epithelial cell subpopulations. The functions
of these isoforms remain relatively unclear15; however, CD44
molecules containing the v3 domain are able to bind growth fac-
tors and might play a role in the control of cellular proliferation.10
The evidence relating to splice variants is conflicting; with dif-
ferent studies detecting widely varying expression levels of splice
variants.20,21Little information relating to splice variant v3 and
melanoma is available mainly because of limited detection of
expression.18,21,31D€ ome et al. analyzed 46 primary melanomas
with a minimum follow-up of 61 months and were the first to dem-
onstrate the expression of the v3 splice variant by immunohisto-
chemistry in melanoma.32CD44v3 expression was only observed
in intermediate and thick tumors and furthermore, was not
detected in a nonmetastatic group of patients. The 5-year survival
of patients with CD44v3 positive tumors was lower than those
who did not express this marker. These results are wholly at var-
iance with those found in the present study possibly because of
smaller numbers in the study by D€ ome, of which only 15 tumors
showed any positivity of staining. Another explanation between
the differences in outcome between our study and that by D€ ome is
the differences in patient population. A significant proportion of
tumors used in our study were less than 1 mm thick, resulting in a
bias toward thin tumors, whilst the study by D€ ome did not use
FIGURE 5 – Survival curves comparing staining intensity of
CD44v3 in primary melanomas. Log rank v25 12.5162, p 5 0.0058.
CD44v3 negative, ——– CD44v3 11 positive, ÆÆÆÆÆÆÆÆ CD44v3 21 posi-
tive,- - - CD44v3 31 positive.
TABLE III – COX MULTIVARIATE ANALYSIS INCLUDING CD44V3
EXPRESSION AND OTHER ESTABLISHED MARKERS OF PROGNOSIS1
95% CI for the hazard
1CD44v3 expression is shown not to be independently predictive of
CD44V3 AS A PROGNOSTIC MARKER FOR MELANOMA
such a patient population since their antibody did not react with
The present study showed a clear correlation between CD44v3
staining intensity and outcome, with the strongest staining being
correlated with the most favorable outcome. Although loss of
CD44v3 expression has not previously been shown to correlate
with poor outcome in melanoma, it has been shown in several
other cancers, including oral squamous cell carcinoma,33lung
adenocarcinoma28and ovarian cancer.26
The role of CD44v3 in tumor progression remains to be deter-
mined. The v3 splice variant is the only CD44 isoform able to bind
heparin-binding growth factors and cytokines through its heparin
sulphate side chains.10,34We hypothesize that cells expressing the
v3 variant may therefore accumulate matrix-bound growth factors.
This could improve the survival of cells remaining in their original
environment and relieve the selective pressure to progress toward
a metastasis-forming phenotype. Further work on this theory
clearly needs to be done, which may include the evaluation of
melanoma metastases for CD44v3 expression. This would explain
the findings of this study, in which the stronger the expression of
CD44v3, the better the prognosis of the patient. On the basis of
the findings of this study, we feel that there may be a role for using
CD44v3 staining in the assessment and prediction of prognosis of
primary cutaneous melanoma.
The authors have no commercial interest in the outcome of this
1. MacKie RM, Bray CA, Hole DJ, Morris A, Nicolson M, Evans A,
Doherty V, Vestey J. Incidence of and survival from malignant mela-
noma in Scotland: an epidemiological study. Lancet 2002;360:587–
Clary BM, Mann B, Brady MS, Lewis JJ, Coit DG. Early recurrence
after lymphatic mapping and sentinel node biopsy in patients with pri-
mary extremity melanoma: a comparison with elective lymph node
dissection. Ann Surg Oncol 2001;8:328–37.
Statius Muller MG, van Leeuwen PA, van Diest PJ, Pijpers R, Nij-
veldt RJ, Vuylsteke RJ, Meijer S. Pattern and incidence of first site
recurrences following sentinel node procedure in melanoma patients.
World J Surg 2002;26:1405–11.
Estourgie SH, Nieweg OE, Valdes Olmos RA, Hoefnagel CA, Kroon
BB. Review and evaluation of sentinel node procedures in 250 mela-
noma patients with a median follow-up of 6 years. Ann Surg Oncol
Meier F, Will S, Ellwanger U, Schlagenhauff B, Schittek B, Rassner
G, Garbe C. Metastatic pathways and time courses in the orderly pro-
gression of cutaneous melanoma. Br J Dermatol 2002;147:62–70.
Balch CM, Soong SJ, Gershenwald JE, Thompson JF, Reintgen DS,
Cascinelli N, Urist M, McMasters KM, Ross MI, Kirkwood JM,
Atkins MB, Thompson JA, et al. Prognostic factors analysis of 17,600
melanoma patients: validation of the American Joint Committee
on cancer melanoma staging system. J Clin Oncol 2001;19:3622–
Clark WH, Jr, Elder DE, Guerry D, IV, Braitman LE, Trock BJ,
Schultz D, Synnestvedt M, Halpern AC. Model predicting survival in
stage I melanoma based on tumor progression. J Natl Cancer Inst
Underhill C. CD44: the hyaluronan receptor. J Cell Sci 1992;103(Pt
Screaton GR, Bell MV, Jackson DG, Cornelis FB, Gerth U, Bell JI.
Genomic structure of DNA encoding the lymphocyte homing receptor
CD44 reveals at least 12 alternatively spliced exons. Proc Natl Acad
Sci USA 1992;89:12160–4.
10. Bennett KL, Jackson DG, Simon JC, Tanczos E, Peach R, Modrell B,
Stamenkovic I, Plowman G, Aruffo A. CD44 isoforms containing
exon V3 are responsible for the presentation of heparin-binding
growth factor. J Cell Biol 1995;128:687–98.
11. Bartolazzi A, Peach R, Aruffo A, Stamenkovic I. Interaction between
CD44 and hyaluronate is directly implicated in the regulation of
tumor development. J Exp Med 1994;180:53–66.
12. Heider KH, Dammrich J, Skroch-Angel P, Muller-Hermelink HK,
Vollmers HP, Herrlich P, Ponta H. Differential expression of CD44
splice variants in intestinal- and diffuse-type human gastric carcino-
mas and normal gastric mucosa. Cancer Res 1993;53:4197–203.
13. Nemec RE, Toole BP, Knudson W. The cell surface hyaluronate bind-
ing sites of invasive human bladder carcinoma cells. Biochem Bio-
phys Res Commun 1987;149:249–57.
14. Horst E, Meijer CJ, Radaszkiewicz T, Ossekoppele GJ, Van Krieken
JH, Pals ST. Adhesion molecules in the prognosis of diffuse large-cell
lymphoma: expression of a lymphocyte homing receptor (CD44),
LFA-1 (CD11a/18), and ICAM-1 (CD54). Leukemia 1990;4:595–9.
15. Stamenkovic I, Aruffo A, Amiot M, Seed B. The hematopoietic and
epithelial forms of CD44 are distinct polypeptides with different adhe-
sion potentials for hyaluronate-bearing cells. EMBO J 1991;10:
16. Catterall JB, Jones LM, Turner GA. Membrane protein glycosylation
and CD44 content in the adhesion of human ovarian cancer cells to
hyaluronan. Clin Exp Metastasis 1999;17:583–91.
17. Birch M, Mitchell S, Hart IR. Isolation and characterization of human
melanoma cell variants expressing high and low levels of CD44. Can-
cer Res 1991;51:6660–7.
18. Harwood CA, Green MA, Cook MG. CD44 expression in melanocytic
lesions: a marker of malignant progression? Br J Dermatol 1996;135:
19. Karjalainen JM, Tammi RH, Tammi MI, Eskelinen MJ, Agren UM,
Parkkinen JJ, Alhava EM, Kosma VM. Reduced level of CD44 and
hyaluronan associated with unfavorable prognosis in clinical stage I
cutaneous melanoma. Am J Pathol 2000;157:957–65.
20. Seiter S, Tilgen W, Herrmann K, Schadendorf D, Patzelt E, Moller P,
Zoller M. Expression of CD44 splice variants in human skin and epi-
dermal tumours. Virchows Arch 1996;428:141–9.
21. Ranuncolo SM, Ladeda V, Gorostidy S, Morandi A, Varela M, Lastiri
J, Loria D, Del Aguila R, Joffe EB, Pallotta G, Puricelli L. Expression
of CD44s and CD44 splice variants in human melanoma. Oncol Rep
22. Kononen J, Bubendorf L, Kallioniemi A, Barlund M, Schraml P,
Leighton S, Torhorst J, Mihatsch MJ, Sauter G, Kallioniemi OP. Tis-
sue microarrays for high-throughput molecular profiling of tumor
specimens. Nat Med 1998;4:844–7.
23. Pacifico MD, Grover R, Richman P, Daley F, Wilson GD. Validation
of tissue microarray for the immunohistochemical profiling of mela-
noma. Melanoma Res 2004;14:39–42.
24. Stoolman LM. Adhesion molecules controlling lymphocyte migration.
25. Gunthert U, Stauder R, Mayer B, Terpe HJ, Finke L, Friedrichs K.
Are CD44 variant isoforms involved in human tumour progression?
Cancer Surv 1995;24:19–42.
26. Saegusa M, Machida D, Hashimura M, Okayasu I. CD44 expression
in benign, premalignant, and malignant ovarian neoplasms: relation to
tumour development and progression. J Pathol 1999;189:326–37.
27. Setala L, Lipponen P, Tammi R, Tammi M, Eskelinen M, Alhava E,
Kosma VM. Expression of CD44 and its variant isoform v3 has no
prognostic value in gastric cancer. Histopathology 2001;38:13–20.
28. Suzuki H, Yamashiro K. Reduced expression of CD44 v3 and v6 is related
to invasion in lung adenocarcinoma. Lung Cancer 2002;38:137–41.
29. Hart IR, Birch M, Marshall JF. Cell adhesion receptor expression dur-
ing melanoma progression and metastasis. Cancer Metastasis Rev
30. Guo Y, Ma J, Wang J, Che X, Narula J, Bigby M, Wu M, Sy MS.
Inhibition of human melanoma growth and metastasis in vivo by anti-
CD44 monoclonal antibody. Cancer Res 1994;54:1561–5.
31. Seelentag WK, Boni R, Gunthert U, Futo E, Burg G, Heitz PU, Roth
J. Expression of CD44 isoforms and b 1,6-branched oligosaccharides
in human malignant melanoma is correlated with tumor progression
but not with metastatic potential. J Cutan Pathol 1997;24:206–11.
32. D€ ome B, Somlai B, Lad? anyi A, Fazekas K, Z€ oller M, T? ım? ar J.
Expression of CD44v3 splice variant is associated with the visceral
metastatic phenotype of human melanoma. Virchows Arch 2001;
33. Bankfalvi A, Krassort M, Buchwalow IB, Vegh A, Felszeghy E,
Piffko J. Gains and losses of adhesion molecules (CD44, E-cadherin,
and b-catenin) during oral carcinogenesis and tumour progression. J
34. van der Voort R, Taher TE, Wielenga VJ, Spaargaren M, Prevo R,
Smit L, David G, Hartmann G, Gherardi E, Pals ST. Heparan sulfate-
modified CD44 promotes hepatocyte growth factor/scatter factor-
induced signal transduction through the receptor tyrosine kinase c-
Met. J Biol Chem 1999;274:6499–506.
PACIFICO ET AL.