Main

Autophagy is a self-degradation mechanism by which cells recycle their long-lived proteins and defective organelles under conditions of microenvironmental stress, such as oxygen depletion and nutrient deprivation.1 By disposing excess or defective cytoplasmic components into the lysosomes, cells renew themselves acquiring, at the same time, energy through metabolic consumption of the digested material. Although autophagy is a major cell survival pathway, its excessive activation leads to massive degradation of cellular components, shifting the balance to self-destruction and autophagic cell death. Several studies suggest that autophagy is intensified in neoplastic cells, but its role in growth and progression of human malignancies remains poorly understood and even more complex is its significance in the response of neoplastic cells to various cytotoxic agents and radiation.2

Autophagy is characterized by the formation of double membrane vacuoles containing cytoplasmic constituents, the autophagosomes; these are fused with lysosomes to form the autolysosomes and which subsequently degrade and recycle the sequestered material. An essential component of the autophagic vacuoles is the microtubule-associated protein 1 light chain 3, that is MAP1LC3A (LC3A)—a homolog of yeast autophagy-related (Atg) protein 8. MAP1LC3A exists in two forms, the MAP1LC3A-I (cytosolic) and the MAP1LC3A-II (membrane bound).3, 4, 5 MAP1LC3A-I (18 kDa) is derived from a proLC3 (30 kDa) protein after cleavage by Atg4. Following activation by Atg7 and Atg3, MAP1LC3A-I is converted into the membrane-bound form MAP1LC3A-II (16 kDa).5 The latter binds tightly to preautophagosomal, autophagosomal and autolysosomal membranes forming a suitable marker of autophagic activity.3, 4, 5

BECN1 (Beclin 1) is the mammalian ortholog of yeast Atg6 gene, which is essential for autophagosome formation.6, 7 The human Beclin 1 gene resides on chromosome 17q21 and is monoallelicaly deleted in many human breast, ovarian and prostatic cancer cell lines.6, 7, 8, 9 Furthermore, Beclin 1+/− mutant mice suffer from high incidence of spontaneous tumors, suggesting that BECN1 has a tumor suppressor function6, 7, 9, 10, 11 with a regulatory role in autophagy.11

Hypoxia is a prevailing feature in malignant tissues mainly because of an insufficient, and often immature, vascular network that prevents adequate blood and oxygen perfusion.12, 13 Experimental studies suggest that hypoxia is a potent stress stimulus triggering autophagy in normal and malignant cells.14 The hypoxia-inducible factor HIF1A, a key transcription factor regulating the expression of a variety of genes involved in angiogenesis and anaerobic metabolism,15 is also involved in the regulation of autophagy in neoplastic cells under hypoxic conditions.16, 17, 18

In the present study, we investigated the autophagic process and its relationship to prognosis in a series of 99 uveal melanomas after immunohistochemical staining for the Atg proteins MAP1LC3A and BECN1. The possible relationship between autophagy and intratumoral hypoxia, assessed by the expression of HIF1A and its downstream protein lactate dehydrogenase 5,15 was also sought. Note that LDH5 is composed of five LDHA subunits.

Materials and methods

Tissue Specimens

A total of 99 uveal melanomas treated with enucleation at the Department of Ophthalmology, Charité Medical University, Campus Benjamin Franklin, Berlin were included in this study. None of the patients had been treated with radiotherapy before or after the operation. The median follow-up was 79 months (range 6–154 months). Informed consent was obtained from the patients. The study was also approved by the Research Committee of the Democritus University of Thrace/University General Hospital of Alexandroupolis, where it was conducted.

The specimens had been fixed in 10% formalin and processed routinely to paraffin wax. Following histopathological examination, the melanoma cells were assigned to cell types (spindle cell, epithelioid, and mixed-cell type with spindle or epithelioid cell predominance) and assessed for localization, pigmentation, and necrosis, as indicated in Tables 1 and 2. Tumor thickness was assessed by ultrasonography.

Table 1 Clinical characteristics and histopathological features in patients with uveal melanomas
Full size table
Table 2 Association of MAP1LC3A expression with pathological parameters in 99 uveal melanomas
Full size table

The Primary Antibodies and the Immunohistochemical Techniques

Autophagy was detected by a standard immunohistochemical technique using two different primary antibodies: the purified rabbit polyclonal antibody MAP1LC3A (AP1805a; Abgent, San Diego, CA, USA) at 1:20 dilution;19, 20, 21 and the rabbit monoclonal antibody EPR1733Y to BECN1 (ab51031; Abcam, Cambridge, UK), used at a dilution 1:50.22

The detection of hypoxia-related proteins HIF1A and LDHA (LDH5 isoenzyme) was achieved after employing a further two markers: the ESEE122 monoclonal antibody (Oxford University, UK) at dilution 1:20 and overnight incubation, for identification of the HIF1A protein;23 and the polyclonal antibody ab9002 (Abcam) at dilution 25 μg/ml and 75 min incubation, for revealing LDHA (LDH5 isoenzyme).24, 25

Tissue sections from uveal melanomas were cut at 3 μm and stained immunohistochemically as previously described.19, 20, 21, 22, 23, 24, 25

Evaluation of Autophagic Activity and Hypoxia

The patterns of MAP1LC3A expression in uveal melanomas were diffuse cytoplasmic and cytoplasmic/juxta-nuclear (Figure 1a and b). These were evaluated as follows. The proportion of tumor cells showing a diffuse cytoplasmic reactivity per section was determined at × 200 magnification. Using the 50% as a cutoff point, cases were divided into groups of low (≤50%) and high (>50%) cytoplasmic MAP1LC3A reactivity. The proportion of tumor cells expressing the perinuclear MAP1LC3A pattern was assessed in a similar manner. These tumors were meant to be separated into groups of low, intermediate, and high reactivity using the 33rd and 67th percentiles. However, as the 67th percentile was 0%, two groups of perinuclear reactivity were eventually formed, the low and the high (see Results).

Figure 1
figure 1

(a) Diffuse cytoplasmic pattern of MAP1LC3A in a uveal melanoma. (b) Cytoplasmic/perinuclear pattern of MAP1LC3A in a uveal melanoma. (c) Weak cytoplasmic reactivity of BECN1 in a uveal melanoma. (d) Strong cytoplasmic reactivity of BECN1 in a uveal melanoma.

Full size image

The reactivity of BECN1 in uveal melanomas was in all cases diffuse cytoplasmic and was evaluated, on the basis of the extent and intensity of staining, at × 200 magnification, as previously indicated:22

  1. 1

    Low level: when cytoplasmic reactivity expressed strongly in <10% of tumor cells, and/or weakly in >90% of tumor cells (Figure 1c);

  2. 2

    Overexpression: when >10% of tumor cells showed a strong cytoplasmic reactivity, and no >50% were negative; overexpression was considered as being (2a) extensive, when >50% of neoplastic cells expressed strong BECN1 reactivity or (2b) limited, when 10–50% of tumor cells showed a similar reaction (Figure 1d);

  3. 3

    Underexpression: if >50% of tumor cells in a section were negative.

HIF1A and LDHA (LDH5 isoenzyme) reactivity was mixed cytoplasmic and nuclear. The percentage of melanoma cells with nuclear and/or strong cytoplasmic HIF1A expression was assessed separately in all optical fields at × 200 magnification and the mean value of all fields was used to score each case. Melanomas with nuclear HIF1A expression in >10% of tumor cells and/or strong cytoplasmic expression in >50% of tumor cells were considered as being of high reactivity.26 LDHA (LDH5 isoenzyme) was evaluated in a similar way. The assessment was performed blindly by two independent observers (AG and ES).

Western Blotting

In order to validate the specificity of MAP1LC3A antibody, a western blot analysis of mouse liver supernatant and pellet extracts was performed using the AP1805a antibody.

Statistical Analysis

Statistical analysis was performed using the GraphPad Prism 5.0 package (GraphPad, San Diego, CA, USA; http://www.graphpad.com). The χ2, the Fisher's exact t-test, or the unpaired two-tailed t-test was used for testing relationships between categorical variables. Linear regression analysis was used to compare groups of continues variables. A P-value ≤0.05 was considered significant.

Results

Table 1 shows the patient characteristics and the histopathological features of all uveal melanomas in the series. The diameter of the tumors, assessed by ultrasonography, ranged from 4 to 23 mm (median 15 mm) at the base of the lesions, and was from 1 to 22 mm (median 9 mm) height.

MAP1LC3A Antibody Validation

Validation of MAP1LC3A antibody's specificity (AP1805a; Abgent), performed by western blot analysis of mouse liver extracts, gave two bands corresponding to the MAP1LC3A-I and II proteins (Figure 2).

Figure 2
figure 2

Western blotting showing (a) the band of recombinant MAP1LC3A protein with tag, (b) the two bands of MAP1LC3A, the MAP1LC3A-I (soluble), and the MAP1LC3A-II (membrane bound) form, in the pellet from mouse liver, and (c) MAP1LC3A-II, but not MAP1LC3A-I, in the supernate of the tissue.

Full size image

MAP1LC3A and BECN1 Expression Patterns

After immunohistochemical staining for MAP1LC3A, two distinct patterns of cytoplasmic expression were consistently recognized in uveal melanomas: the diffuse cytoplasmic, and the cytoplasmic/juxta-nuclear occurring in a crescentic or ring-like (perinuclear) manner. A third pattern, the so-called ‘stone-like structures’, reported earlier in a number of epithelial tumors by our group, was not detected in uveal melanomas.19, 27

The diffuse cytoplasmic MAP1LC3A pattern varied considerably between cases, ranging from 0 to 100% of melanoma cells (median 40%). Using the 50th (50%) percentile, the cases were divided into groups of low (0–50%, 59 cases), and high (51–100%, 40 cases) cytoplasmic reactivity.

The perinuclear MAP1LC3A pattern also varied between cases, ranging from 0 to 90% of tumor cells (median 0%). The 33rd and 67th percentiles were 0%, so cases were grouped in two categories of low (0%, 68 cases) and high (10–90%, 31 cases) perinuclear reactivity.

BECN1 was diffusely expressed in the cytoplasm of melanoma cells. Overexpression was evident in 48/99 cases (48.4%), of which 22/99 (22.2%) were extensive (>50% of cells), and 26/99 (26.2%) were limited. Underexpression was noted in 29/99 (29.3%) cases, while a basal level of BECN1 activity was seen in 22/99 (22.2%) cases.

Both high cytoplasmic and perinuclear MAP1LC3A expression were related with extensive beclin 1 overexpression (P=0.02 and P=0.03, respectively) (Table 4).

Intra- and Inter-observer Variability

The immunohistochemical expression patterns for MAP1LC3A (cytoplasmic and perinuclear) and BECN1 (cytoplasmic) were examined for intra- and inter-observer variability. Two experienced pathologists assessed the slides separately and repeated the assessment 30 days later. The second assessment was highly correlated with the first for all observers (r>0.97, P<0.0001 for all three parameters). Similarly, the two investigator's scoring correlated to each other for all three variables (r>0.92, P<0.0001). The final decision was taken on the conference microscope.

MAP1LC3A and BECN1 in Relation to Pathological Features

Both high cytoplasmic and perinuclear MAP1LC3A patterns were linked with intense pigmentation of uveal melanomas (P=0.01 and P=0.0002, respectively).

Uveal melanomas composed exclusively of spindle cells were declining to express BECN1 (P=0.005) (Table 3). Overexpression of BECN1, whether limited or extensive, was linked with intense pigmentation (P=0.01).

Table 3 Association of BECN1 expression with pathological parameters in 99 uveal melanomas
Full size table

There was a significant association between high cytoplasmic MAP1LC3A expression and extensive overexpression of BECN1 (P=0.03), and between high perinuclear MAP1LC3A expression and the overall BECN1 overexpression (P=0.03) (Table 4).

Table 4 Association between MAP1LC3A and BECN1 expression
Full size table

MAP1LC3A and Hypoxia-Related Proteins

Linear regression analysis showed that perinuclear, but not the cytoplasmic, MAP1LC3A pattern expression was directly linked with the hypoxia-related proteins HIF1A and LDHA (LDH5 isoenzyme) (Table 5). The same applied to BECN1. No association was noted between cytoplasmic MAP1LC3A pattern and the hypoxia-linked proteins.

Table 5 Linear regression analysis of the percentage of melanoma cells expressing strongly BECN1 and cytoplasmic or perinuclear MAP1LC3A patterns according to cytoplasmic or nuclear expression of HIF1 A and LDHA (LDH5 isoenzyme)
Full size table

In group analysis, a high perinuclear MAP1LC3A expression was associated with a high LDHA (LDH5 isoenzyme) expression, both cytoplasmic (P<0.0001) and nuclear (P<0.0001). Similarly, an increased perinuclear MAP1LC3A expression was connected with HIF1A expression, whether cytoplasmic or nuclear, but the difference was not significant (P=0.15 and 0.12, respectively) (data not shown).

Survival Analysis

Table 6 shows the univariate and multivariate analysis of metastatic and death events. The presence of an epithelioid component in uveal melanomas was linked with a poor metastasis-free and overall survival (P=0.02 and 0.02, respectively). At univariate analysis, an extensive overexpression of BECN1 as well as an underexpression of this protein were linked with poor metastasis-free interval and disease-specific overall survival (P=0.01).

Table 6 Univariate and multivariate analysis in metastatic and death events in 99 uveal melanomas
Full size table

Figure 3 shows that patients being almost at the two extremes of BECN1 expression, that is, patients with extensive overexpression of BECN1 and those with underexpression of this protein had the worse prognosis. Nevertheless, the former manifested metastases much earlier than the latter. In fact, in the presence of extensive overexpression of BECN1, only 58% of the patients were alive at 4 years. By contrast, underexpression patterns correlated with delayed appearance of metastases and death events, with over 80% of patients being alive at 4 years (P=0.02 for metastasis and P=0.05 for death events).

Figure 3
figure 3

Kaplan–Meier survival curves of (a) metastasis-free survival, and (b) overall disease-specific survival, stratified by BECN1 expression.

Full size image

In multivariate analysis, the patterns of BECN1 expression and the presence of an epithelioid component were the only prognostic variables that maintained an independent significance (Table 6).

Discussion

Hypoxia is a major microenvironmental condition in solid tumors that defines tumor aggressiveness either by selecting resistant sub-populations of cancer cells or by triggering specific molecular pathways that facilitate tumor growth, invasiveness, and metastasis.28, 29 Hypoxic cells have been detected in human melanoma xenografts by pimonidazole immunohistochemistry30 and, indeed, were shown to promote metastatic spread.31, 32 HIF1A, in particular, is largely expressed in human neoplasia, including melanomas, and is linked with metastasis33, 34 and melanocyte transformation, in conjunction with the Akt gene.35

Furthermore, HIF1A is actively involved in triggering autophagy, a major intracellular pathway promoting the survival of tumor cells under oxygen and nutrient deprivation.16, 17, 18 This autophagic process has been documented in human melanomas as early as 1982 by Horikoshi et al,36 and melanoma cells are known to induce autophagic activity with perinuclear expression after being irradiated with UV light.37 Yet, the role of autophagy and its links with hypoxia and the autophagy/hypoxia interaction in human melanomas remains obscure.

In this study, the autophagy-related proteins Atg8 (MAP1LC3A) and Atg6 gene (BECN1) were used as immunohistochemical targets to detect autophagic activity in a series of 99 uveal melanomas.

Miracco et al38 investigating the expression of autophagic genes in cutaneous melanocytic lesions found a gradual decrease in these proteins with tumor progression; the proportion of cases with high cytoplasmic expression of LC3 declined from benign through dysplastic to malignant melanocytic lesions, being at lowest in metastatic melanomas. While a similar decrease in MAP1LC3A expression was reported in cerebral and ovarian cancer,39, 40 in other malignancies, such as esophageal and gastrointestinal carcinomas, an increased LC3 expression was prevailing.41 There were also reports connecting an increased LC3 activity with poor outcome in pancreatic tumors and with a better survival in glioblastoma patients with poor performance score.40, 42

In most of these studies, however, the expression of LC3 protein was reported as being consistently cytoplasmic, a staining pattern that was also seen with a frequency of 40% in our series of uveal melanomas, and which, presumably, corresponds to the soluble non-membrane-bound form of MAP1LC3A (MAP1LC3A-I). It should be mentioned, however, that a substantial proportion of cases amounting to 31% in our material, also showed a distinct perinuclear MAP1LC3A expression. This is not entirely unexpected, since an analogous ‘paranuclear pinpoint stain’ was described by Miracco et al38 in their series of cutaneous malignant melanomas. Rieber and Rieber37 investigating human C8161 melanoma cells, also noted that 30% of UV-irradiated melanomas showed this juxta- or perinuclear pattern of expression. This is interesting as lysosomes usually accumulate to the perinuclear region where they fuse with cytoplasmic autophagosomes (a process also known as autophagocytosis) to form the autolysosomes. It appears, therefore, that the perinuclear MAP1LC3A expression may reflect an active autophagic function, as indicated by the accumulation of MAP1LC3A-II-positive autophagosomes to the perinuclear region of the lysosomal domain.43, 44

The question, of course, remains why melanomas, whether uveal or cutaneous, are lacking ‘stone-like’ structures, despite being aggressive tumors. Obviously, there is no easy answer to that, but it is presumed that malignant melanocytic tumors, being endowed with a rich vascular supply, either need not require an accelerated form of autophagy to survive the adverse conditions of the microenvironment or that these tumors are more resistant to nutrient deprivation—a phenomenon known as austerity.45 Alternatively, there may be another, yet unsuspected, mechanism operative in malignant melanomas.

With regard to BECN1, an autophagy-related gene with tumor suppressor activity, Miracco et al38 reported comparable results to those shown for MAP1LC3A, that is gradual decline of cytoplasmic expression of BECN1 in cutaneous melanocytic lesions from 100% in benign nevi to 86.4% in dysplastic nevi, 54.3% in melanomas, and 26.7% in melanoma metastases. Such decreased cytoplasmic expression of BECN1 has been correlated with tumor progression in breast, ovarian, and cerebral neoplasms39, 46, 47 and with a reduced survival rate in neoplasms of esophageal, colorectal, hepatocellular, and cerebral origin.48, 49, 50, 51 Liang et al46 explained this immunohistochemical decline in cytoplasmic expression of the protein in terms of BECN1 inactivation and loss of its tumor suppressor functions. There have been, however, other tumors, including colorectal and gastric adenocarcinomas, in which progression appeared to be dependent upon an increased BECN1 expression.52

The above conflicting results, although may suggest different roles of BECN1 expression depending on cancer cell types,47, 52 they may also imply the existence of various patterns of BECN1 expression up to date unrecognized. In this context, we hypothesized that there are four patterns of BECN1 expression, the low level of BECN1 expression, which may reflect a basal level of autophagic function; the overexpression of BECN1 (two grades), which apparently represent an intensification of the autophagic machinery under the adverse intratumoral conditions of hypoxia and acidity, as indicated in this and in other studies;50, 53 and, of course, there may be a reduction or loss of BECN1 expression, a probable result of allelic gene deletions8 or a consequence of deregulated microRNA function, such as miT-30a.54 As BECN1 interacts with members of the bcl-2 protein acting as a tumor suppressor,55 potentiation of the anti-apoptotic machinery may account, at least in part, for the poor outcome of patients with reduced Beclin 1 expression.

Indeed, in the present study, both the underexpression and the excessive overexpression of BECN1 were linked with metastases and an overall poor disease-specific survival. However, the temporal patterns of metastases and death were distinct, as overexpression was correlated with a rapid onset of metastases, while the underexpression of BECN1 was linked with delayed onset of metastatic disease and delayed death rates after 4–5 years from surgery. As only overexpression was linked with hypoxia/acidity features (HIF1A and LDHA, LDH5 isoenzyme), it is postulated that the reasons of disease progression largely differ between the two groups. MAP1LC3A expression, on the other hand, did not show any association of clinical importance, given that the association between high cytoplasmic or perinuclear expression and intense pigmentation does not seem to affect the outcome.

It is concluded that autophagy is upregulated in over 40–50% of uveal melanomas, and these cases were correlated with tumor hypoxia; the latter was inferred by the intimate relationship of MAP1LC3A and BECN1 with HIF1A and LDHA (LDH5 isoenzyme), the major LDH isoenzyme involved in anaerobic metabolism. The overexpression of BECN1, being linked with tumor hypoxia and acidity, was associated with poor prognosis. A low BECN1 expression, probably representing gene silencing mechanism in a subset of tumor, was also linked with death events but these were of delayed metastasis. Uveal melanomas with the presence of an epithelioid component showed a clear tendency for metastases and a reduced overall survival, but this has been known for long56 and needs no further comment. It will be of interest to investigate in the future the relation of autophagy with tumor-infiltrating macrophages in uveal melanomas.