Cell motility as a prognostic factor in Stage I nonsmall cell lung carcinoma

The role of gelsolin expression


  • Dar-Bin Shieh D.D.S., D.M.Sc.,

    1. Division of Experimental Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • John Godleski M.D.,

    1. Harvard School of Public Health; Department of Pathology, Brigham and Women's Hospital, Harvard Medical School; Boston, Massachusetts
    Search for more papers by this author
  • James E. Herndon II Ph.D.,

    1. Division of Biometry, Duke University Medical Center, Durham, North Carolina
    Search for more papers by this author
  • Toshifumi Azuma M.D.,

    1. Division of Experimental Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Harriet Mercer B.Sc.,

    1. Division of Thoracic Surgery, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • David J. Sugarbaker M.D.,

    1. Division of Thoracic Surgery, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • David J. Kwiatkowski M.D., Ph.D.

    Corresponding author
    1. Division of Experimental Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts
    • Division of Experimental Medicine, Brigham and Women's Hospital, 221 Longwood Avenue, Boston, MA 02115.
    Search for more papers by this author



Tumor cell motility is an important characteristic that facilitates the multistep process of tumor metastasis. Rac, ABP-280, and gelsolin are proteins that interact with actin and are important in cell motility.


The authors studied a cohort of 229 Stage I nonsmall cell lung carcinoma (NSCLC) patients who had a minimum of 3 years follow-up and had been previously analyzed for 22 clinical, pathologic, and molecular features, of which 9 had been found to provide significant prognostic information in a Cox proportional hazards model. Tumor sections were stained by the avidin-biotin complex method using monoclonal antibodies against rac, ABP-280, and gelsolin.


In a pilot analysis of over 50 patients each, rac and ABP-280 were found to be moderately-to-highly expressed in the majority of tumors and to provide no prognostic information. Gelsolin expression was more variable and appeared to be negatively correlated with survival in the pilot population. In the larger 229-patient population, high focal gelsolin expression was seen in 32 tumors (14%) and conferred the highest relative risk (4.04) of cancer recurrence among all factors tested, compared with tumors that had no or low gelsolin expression. Moderate focal gelsolin expression, seen in 46 patients (20%), also conferred a significant risk of cancer recurrence, with a relative risk of 2.26 compared with tumors that had no or low gelsolin expression. Consideration of average gelsolin expression and of overall survival yielded similar results.


Gelsolin expression appears to be a significant prognostic factor for cancer recurrence in cases of Stage I NSCLC. Cancer 1999;85:47–57. © 1999 American Cancer Society.

Lung carcinoma is the leading cause of cancer death in the U.S., with a predicted 180,000 deaths in 1997.1 Nonsmall cell lung carcinoma (NSCLC) comprises 70–80% of all lung carcinomas and has a 5-year survival rate of about 15%. Stage I disease (American Joint Committee on Cancer, International Staging System2) is associated with the best prognosis among all stages of NSCLC, but 30–40% of patients in Stage I experience recurrent disease despite complete surgical resection.2–4 Recent trials of Stage III NSCLC have suggested that chemotherapy and radiation added to surgery improves the survival of those patients.5–9 These preliminary studies also suggest that adjuvant chemotherapy and radiotherapy may be effective in reducing the recurrence risk for patients with Stage I disease who are undergoing a complete surgical resection.

Because chemotherapy and radiation add significant toxicity, many investigators are eager to identify prognostic factors for recurrent disease in the Stage I NSCLC population, so that these adjuvant therapies can be selectively applied. Many studies from numerous centers have been published, often with conflicting results, so that it is difficult to distinguish tumor features with true adverse prognostic significance.10, 11 Tumor features of Stage I NSCLC that have been identified in 2 or more large studies (>100 patients) with significant adverse prognostic influence include tumor size, histology types and subtypes, K-ras codon 12 mutation, abnormalities in p53, the presence of nerve cell adhesion molecule expression, DNA ploidy and proliferative activity of the tumors, the presence of neuroendocrine differentiation, HER-2/erb-B expression, blood vessel/lymphatic invasion, angiogenic index, blood group antigen expression, and mitotic index.12–17

Tumor cell metastasis occurs in a multistage process. Local invasion and destruction of extracellular matrix precedes entry of tumor cells into blood vessels, lymphatics, or other transport channels. The tumor cells must survive in the circulation, extravasate out of vascular channels into metastatic sites, and grow in the new location.18–20 Necessary functional attributes of tumor cells include production of angiogenic factors that lead to generation of fragile capillary structures, and elaboration of metalloproteinases and other compounds capable of degrading extracellular matrices.21–23 Moreover, tumor cell shape changes and locomotion are required in several stages of the metastatic process, and increased cell motility has been associated with metastatic potential in animal and human tumors.18, 24–26 However, few studies have investigated the correlation between the expression of intracellular proteins that are important for tumor cell motility and metastatic potential.

This article describes an analysis of the prognostic significance of the expression of three widely expressed proteins involved in the organization and regulation of the actin cytoskeleton in Stage I NSCLC: rac, ABP-280, and gelsolin. Rac is a 21 kDa GTPase and a member of the ras superfamily; it is thought to be a crucial intermediary in signalling events for ruffling and translocational motility.27–30 ABP-280 is a 280 kDa homodimeric protein that cross-links actin filaments into orthogonal networks.31 It is thought to be essential to the actin filament organization in the cell cortex that facilitates efficient locomotion of cultured melanoma cells, and its expression may correlate with invasive behavior of melanocytic tumors.32, 33 Gelsolin is an 80 kDa protein that regulates actin filament organization through complex effects on the dynamics of actin assembly into filaments, including the severing of actin filaments.34, 35 Increased expression of gelsolin in cultured fibroblasts increases motility,36, 37 although its expression is down-regulated in many human carcinomas, including those developing in the breast, colon, and bladder.38–42


Patient Population

We used a previously characterized cohort of 244 Stage I NSCLC patients treated between January 1984 and December 1992 at Brigham and Women's Hospital.43 These patients had a minimum clinical follow-up of 3 years and were subjected to a comprehensive analysis of multiple prognostic factors: extent of surgical resection, gender, age, greatest dimension of tumor, histologic classification according to World Health Organization (WHO) criteria,44 differentiation (well, moderately, or poorly differentiated), WHO adenocarcinoma subtype (acinar, papillary, bronchioloalveolar, or solid tumor with mucin formation), vascular invasion of pulmonary arteries or veins, lymphatic vessel invasion, visceral pleural invasion, tumor giant cells, plasma cell infiltration, lymphocyte infiltration, atelectasis, and mitotic index (number of mitotic figures per 10 high-power fields). Immunohistochemical staining to analyze expression of erbB-2 (HER-2/neu,) H-ras, rb, bcl-2, p53, and blood group A antigen was also performed, as was analysis for mutations at K-ras codon 12. Two hundred twenty-nine blocks from the original 244 cases were available for analysis of rac, ABP-280, and gelsolin expression. This study was approved by the Brigham and Women's Hospital Human Research Committee. Informed consent was obtained from all patients prior to surgery.


Monoclonal antibodies against rac, ABP-280, and gelsolin were available for these studies and had been previously prepared or used in our laboratory.37, 39 The specificity of each antibody was confirmed by Western blot analysis of extracts prepared from NSCLC cell lines45 (data not shown), which indicated that each antibody reacted with a single band of the appropriate size.

The staining procedure was a modified avidin-biotin complex method. Five-micron sections of the paraffin embedded tumors were mounted on gelatin-coated slides, baked at 60°C for 30 minutes, and deparaffinized in xylene, followed by rehydration in a graded ethanol series. Intrinsic peroxidase was inactivated in 0.3% hydrogen peroxide for 30 minutes. Antigen retrieval by microwaving in 10 mM citric acid for 20 minutes was used for the anti-ABP-280 antibody. Slides were incubated in 10mM Tris-HCl pH7.4, 0.14M NaCl (TBS) with 4% normal goat serum for 1 hour, the specific antibody and in parallel control normal mouse serum at 10× antibody concentration was then added in a similar solution overnight at 4°C. After rinsing in TBS, biotin-conjugated goat antimouse antibody was applied for 30 minutes, additional TBS rinsing was performed, the slides were incubated with horseradish peroxidase–conjugated streptavidin for 30 minutes, additional rinsing was performed, and the slides were then developed in aminoethylcarbazole to generate an orange-red reaction product. The slides were counterstained with hematoxylin. Monoclonal antibody preparations were all hybridoma supernatants and were used at dilutions of 1:200 for rac, 1:100 for ABP-280, and 1:50 for gelsolin.

Slides were read in a blinded manner with respect to the clinical outcomes of the corresponding patients. The level of expression of each protein was graded absent, low, moderate, or high by two observers who double-scoped for all cases (D.B.S. and D.J.K.). Pulmonary macrophages, which were almost universally present in these tumor sections, served as positive controls and stained strongly with all three antibodies. Fibroblasts in tumor stroma also stained with all three antibodies, though more weakly, and served as an additional internal controls. Tumor staining was graded absent when staining was completely absent or less than that of fibroblasts, low when staining was equivalent to fibroblasts, moderate when staining was intermediate between that of fibroblasts and macrophages, and high when staining was similar to or exceeded that of pulmonary macrophages. Seventeen percent of tumors displayed variable gelsolin staining in tumor cells. In these cases, two grades of staining were assigned, one corresponding to the low level of staining seen in most tumor cells (hereafter termed the average gelsolin expression), and the other, a focal grade that was based upon the more intense staining of focal regions (hereafter termed the focal gelsolin expression), which consisted of two or more clusters of multiple tumor cells at minimum. The focal and average gelsolin expression grades were the same for the 83% of tumors with uniform staining.

For confirmation, after the initial sequential reading of all slides, the slides were grouped according to grade of staining and read again. In this reanalysis, about 10% of slides were reassigned to a different grade. In all cases, this resulted in a shift up or down of one level of staining. These latter readings were used in all subsequent analyses.

Statistical Analyses

A pilot study of 50 patients was initially conducted to determine whether further examination of each of 3 markers (rac, ABP-280, and gelsolin) was merited. In that pilot, a Wilcoxon two-sample test was used to compare the distribution of the markers across outcome groups (cancer free or cancer relapse).

An initial analysis of the entire cohort for gelsolin staining indicated that the survival of patients with absent and low gelsolin staining was equivalent. Therefore, a Cox proportional hazards model was used to determine whether gelsolin expression, categorized as absent or low, moderate, or high, provided any additional prognostic information for cancer free or overall survival beyond that provided by known prognostic factors. Factors that were shown to have independent prognostic significance in our previous analysis of this population43 were extent of resection (wedge vs. lobectomy/pneumonectomy), solid tumor adenocarcinoma subtype, maximum tumor dimension >4 cm, female gender, lymphatic invasion, K-ras codon 12 mutation, p53 immunostaining, H-ras immunostaining, and age >60 years.

Cancer free survival was defined as the time between surgery and the first indication of disease recurrence (relapse or cancer death). If a patient died without cancer recurrence, the patient's cancer free survival time was censored at the time of death. Overall survival was defined as the time between surgery and death.


In preliminary studies, we confirmed the specificity of our monoclonal antibodies (data not shown) and examined the expression pattern of the three proteins—rac, ABP-280, and gelsolin—in normal lung (Fig. 1). Pulmonary macrophages and vascular smooth muscle cells showed strong staining for all three proteins. Normal alveolar epithelium showed weak or absent expression of all three proteins, whereas bronchiolar and bronchial epithelium showed high, moderate, and weak expression of rac, ABP-280, and gelsolin, respectively. However, moderate expression of gelsolin was seen in bronchial epithelium in areas of inflammation. In general, tumor cells showed moderate-to-strong expression of ABP-280 and rac (Fig. 2). Gelsolin staining was more variable, with the majority of cases showing absent or weak expression and a minority expressing gelsolin at moderate-to-high levels (Fig. 3). Seventeen percent of blocks showed variable staining for gelsolin, leading to an assignment of an average gelsolin expression grade and a focal gelsolin expression grade (see “Methods” for details).

Figure 1.

Expression of rac, ABP-280, and gelsolin is shown in normal lung. Arrows indicate the structures of interest: (a) alveolae; (b) terminal and respiratory bronchioles; and (c) tertiary, (d) secondary, and (e) primary bronchi. All three proteins are highly expressed in macrophages (M) and smooth muscle cells (SM), but weakly or not at all in alveolae. Rac is expressed highly, ABP-280 moderately, and gelsolin weakly in bronchiolar and bronchial epithelia. Heterogeneous expression of all three is seen in mucous glands (G).

Figure 2.

Expression of rac (A–C) and ABP-280 (D–F) is shown in lung cancer sections. M: macrophage. A and D show moderate expression, whereas B, C, E, and F show high expression.

Figure 3.

Expression of gelsolin is shown in lung cancer sections. M: macrophage. A: absent expression; B: moderate expression; C: low or average expression, high focal expression; D: low expression; E: high expression; F: normal bronchial mucosa.

To explore the prognostic significance of expression of these proteins in NSCLC, we performed a pilot study in which tumors from ≥50 Stage I patients were examined. At least 25 of these patients had experienced recurrent disease, whereas the other 25 had remained cancer free; but they were otherwise chosen randomly. As seen in Table 1, the distribution of rac and ABP-280 expression among those who were and were not long term cancer free survivors was similar. In contrast, focal gelsolin expression was generally lower (P = 0.016) in the subset without recurrent disease compared with the group with cancer recurrence. Therefore, we extended the analysis of gelsolin expression to all tumor blocks available from a cohort of 244 patients with Stage I NSCLC that had been previously studied at our institution.

Table 1. Distribution of Expression of Rac, ABP-280, and Gelsolin in the Pilot Study of 50 or More Tumors
Expression levelAbsentLowModerateHighTotal
 Cancer free0671225
 Cancer relapse01111830
P = 0.162
 Cancer free0352129
 Cancer relapse0181625
P = 0.663
Focal gelsolin
 Cancer free8133125
 Cancer relapse569525
P = 0.016

Blocks from 229 patients from this previously studied cohort were available for gelsolin immunostaining, of which 74 had experienced cancer recurrence. In this 229-patient cohort, the median age was 66 years, there were 129 (56%) males and 100 (44%) females, and there were 57 (25%) squamous cell carcinomas, 132 (58%) adenocarcinomas, 28 (12%) large cell carcinomas, 6 (3%) mixed histologies, and 6 (3%) not specified. Considering focal gelsolin expression (Table 2, top), 28% of tumors expressed no gelsolin, 38% had low levels of gelsolin expression, and 34% expressed gelsolin at moderate or high levels. Considering average gelsolin expression (Table 2, bottom), the distribution of levels of expression were similar, but with the expected shift to lower levels of expression. In this larger population, significantly decreased cancer free survival was again associated with increasing focal gelsolin expression (P < 0.0001, Fig. 4A) and with increasing average gelsolin expression (P < 0.0001, Fig. 4B). Focal and average gelsolin expression were significantly correlated with lymphatic invasion (Table 3, P = 0.012 and 0.0065, respectively). Average gelsolin expression was significantly correlated with H-ras p21 expression and bcl-2 expression (Table 3, P = 0.027 and 0.056, respectively). There was no correlation between focal or average gelsolin expression and histology or any other molecular feature examined.

Table 2. Gelsolin Expression in the Entire 229 Patient Cohort
 Expression level
Focal gelsolin
 Cancer free50 (32%)64 (41%)26 (17%)15 (10%)155
 Cancer recurrence15 (20%)22 (30%)20 (27%)17 (23%)74
  Total65 (28%)86 (38%)46 (20%)32 (14%)229
Average gelsolin
 Cancer free55 (36%)70 (45%)27 (17%)3 (2%)155
 Cancer recurrence15 (20%)32 (43%)17 (23%)10 (14%)74
  Total70 (31%)102 (44%)44 (19%)13 (6%)229
Figure 4.

Kaplan–Meier survival estimates for 229 patients are stratified by focal gelsolin expression (A) and average gelsolin expression (B). Tick marks reflect censored survival times for individual patients without cancer recurrence at last follow-up.

Table 3. Correlation between Gelsolin Expression and Other Tumor Characteristics
 AbsentLowModerateHighP valuea
  • a

    Exact chi-square test.

Focal gelsolin expression
 Lymphatic invasion0.012
  Absent47 (72%)62 (72%)23 (50%)16 (50%)
  Present18 (28%)24 (28%)23 (50%)16 (50%)
Average gelsolin expression
 Lymphatic invasion0.0065
  Absent49 (70%)73 (72%)21 (48%)5 (38%)
  Present21 (30%)29 (28%)23 (52%)8 (62%)
 H-ras p21 expression0.027
  Absent45 (65%)76 (76%)27 (61%)5 (38%)
  Present24 (35%)24 (24%)17 (39%)8 (62%)
bcl-2 expression0.056
  Absent49 (71%)58 (57%)21 (48%)6 (46%)
20 (29%)43 (43%)23 (52%)7 (54%)

Gelsolin was then included with 9 other factors, previously identified as conferring prognostic significance in the total population of 244 patients, in a Cox proportional hazards regression model to define independent risk factors for cancer free survival in the current cohort of 229 patients. The other factors included in this model were age >60 years, male gender, wedge resection versus lobectomy/pneumonectomy, adenocarcinoma subtype solid tumor with mucin, tumor dimension >4 cm, lymphatic invasion, p53 expression, K-ras codon 12 mutation, and lack of H-ras p21 expression. High focal gelsolin expression, seen in 32 patients (14%), compared with low or absent focal gelsolin expression, was found to carry the highest relative risk of cancer recurrence (4.04) among all measures tested (Table 4, top), and was the strongest independent risk factor for cancer recurrence. Moderate focal gelsolin expression, seen in 46 patients (20%), was also an independent risk factor compared with low or absent gelsolin expression, but at a lower risk ratio (2.26). Similar results were obtained considering average gelsolin expression levels (Table 4, bottom), for which the relative risk of cancer recurrence was 6.28 for high or average gelsolin expression compared with low or absent expression, and 1.88 for moderate compared with low or absent expression. Similar results were seen considering overall survival (Table 4, right lanes), and restricting the analysis to the subset of patients who had undergone a lobectomy or pneumonectomy.

Table 4. Multivariate Analysis of Independent Factors Associated with an Increased Risk of Cancer Recurrence in 229 Patients, Including Focal Gelsolin Expression (top half) and Including Average Gelsolin Expression (bottom half) (in Order of Relative Risk)
Cox proportional hazards modelCancer free survival analysisOverall survival analysis
VariableP valueRisk ratioLoweraHighRisk ratio
  • a

    95% confidence intervals for estimates of risk.

Focal gelsolin high vs. absent/low0.00014.042.147.633.49
 Wedge vs. lobectomy/pneumonectomy0.00012.931.685.112.44
 Solid tumor adenocarcinoma0.00222.681.435.022.17
 Low H-ras p21 expression0.00422.321.304.121.98
 Focal gelsolin moderate vs. absent/low0.00742.
 Maximum tumor dimension ≥4 cm0.0351.891.053.431.74
 Lymphatic invasion0.0131.861.143.051.79
 K-ras codon 12 mutation0.0201.861.103.132.04
 Age ≥60 years0.0711.770.953.302.11
 p53 expression0.0291.741.062.851.41
 Male gender0.0671.620.972.722.16
Average gelsolin high vs. absent/low0.00016.282.7614.325.15
 Low H-ras p21 expression0.00202.651.434.932.27
 Wedge vs. lobectomy/pneumonectomy0.00152.481.424.342.16
 Solid tumor adenocarcinoma0.00782.391.264.531.95
 Lymphatic invasion0.00941.921.173.151.81
 Average gelsolin moderate vs. absent/low0.0371.881.043.412.05
 Maximum tumor dimension ≥4 cm0.0511.811.003.281.69
 Male gender0.0291.791.063.022.37
 Age ≥60 years0.0771.750.953.241.93
 K-ras codon 12 mutation0.0451.701.012.851.88
 p53 expression0.0381.681.032.751.38


Tumor cell motility has long been recognized as an important malignant characteristic that facilitates metastatic progression of tumors both in vivo and in animal model systems.18, 20, 24–26, 46 Cell motility is a complex process involving multiple proteins that interact with and modulate the cortical actin network to induce its coordinated disassembly and reformation under tight spatiotemporal control.47, 48

We examined the expression of three proteins in early stage NSCLC that modulate the organization of the actin cytoskeleton, participating in the motility of multiple cell types. Rac and ABP-280 were expressed at appreciable levels in all tumor blocks examined, perhaps indicative of their important role in cell organization in all lung tumors. There was no correlation between low, moderate, or high level expression of these proteins and clinical outcome in a small pilot study. Gelsolin expression, on the other hand, was found to vary widely among the lung tumors examined and did not correlate with either rac or ABP-280 expression (data not shown). About one-third of lung tumor blocks showed no gelsolin expression, and somewhat more showed low gelsolin expression. The remainder showed significant gelsolin expression that was clearly increased compared with that seen in normal lung mucosa or normal stroma cells, and a fraction of those (14%) showed expression at relatively high levels.

We found that high level expression of gelsolin conferred a significantly worse prognosis on the patients from whom such tumors were removed, and that high level gelsolin expression conferred the highest risk of cancer relapse among all molecular, clinical, and pathologic features evaluated in this population (22 distinct features).43 Although the number of patients studied here is substantial, these findings should be confirmed by other independent studies in well-characterized Stage I NSCLC populations.

Gelsolin is a multifunctional actin-modulating protein, which is activated by micromolar Ca2+ or lowered pH (<6.5) and inactivated by polyphosphoinositides, and can nucleate (accelerate) actin filament assembly, sever actin filaments, and cap the barbed end of filaments.35, 49, 50 The importance of high gelsolin expression as an indicator of recurrence risk in Stage I NSCLC is consistent with in vitro studies in which gelsolin expression in fibroblasts has been positively correlated with translocational motility, ruffling activity, and pinocytosis.36, 37, 51 Gelsolin has also been identified as a substrate for caspace-3, an aspartate specific protease activated during apoptosis.52 The gelsolin cleavage product is constitutively activated for severing actin filaments and mediates, in part, the morphologic changes of apoptosis in some cell types. However, in Jurkat cells gelsolin also appears to function as an inhibitor of apoptosis by an undefined mechanism.53 Thus, the role of gelsolin in apoptosis may also contribute to its prognostic significance in NSCLC. However, we suspect that it is gelsolin's role in motility that causes NSCLC tumors with high gelsolin expression to be associated with a poorer prognosis. The finding that both focal and average gelsolin expression correlated with lymphatic invasion is consistent with enhanced motility in the gelsolin-expressing lung carcinoma cell.

Loss of gelsolin expression has been observed during bladder, breast, lung, and colon carcinogenesis, both in cell lines and through immunohistochemical analysis of human tumors.54–57 Although frequent, loss of gelsolin is not universal in cancers from these tissues. Gelsolin transfection into transformed NIH 3T3 cell lines and bladder carcinoma cell lines leads to a reversion of transformation in assays in vitro and in limited studies in vivo.54, 58 However, multiple other cytoskeletal and other proteins, when expressed in the NIH3T3 model system, also have tumor suppression properties.59, 60 Although the majority of the lung tumors examined here had no or low gelsolin expression, consistent with these previous studies in other epithelial cell malignancies54–56 and lung carcinoma,57 a significant fraction expressed gelsolin at high levels and had a generally worse prognosis. It is difficult to reconcile all of these observations into a simple model for the action of gelsolin during malignant epithelial cell transformation. We hypothesize that gelsolin expression is generally extinguished during transformation, but in some tumors later events lead to and permit its expression, and in those tumors it facilitates metastasis by increasing tumor cell motility.

Although tumor cell motility is recognized as an important contributor to the multistep metastatic process, relatively little investigation has been carried out regarding its importance as a prognostic factor. We propose that surrogate measures of tumor cell motility, such as high gelsolin expression, may have broad prognostic significance to human cancer. This concept is supported by recent studies indicating the importance of thymosin β15 expression in prostate carcinoma metastasis.61