SEARCH

SEARCH BY CITATION

Keywords:

  • ATF5;
  • cancer;
  • microarray;
  • cancer tissue microarray;
  • breast cancer;
  • apoptosis;
  • ductal carcinoma;
  • lobular carcinoma

Abstract

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

ATF5, a transcription factor important in differentiation, proliferation and survival, has been found to be highly expressed in neural progenitor cells and in certain tumors including glioblastomas (GBMs), but its expression in other normal and neoplastic tissues has not been extensively investigated. A tissue microarray immunostained for ATF5 showed diffuse nuclear expression (as defined by the presence in greater than 25% of cells) in 63% (117/186) of neoplastic samples, when compared to only 32% (20/62) in nonneoplastic tissues. When analyzed by histologic subtype, a significantly greater proportion of adenocarcinomas, transitional cell carcinomas, squamous cell carcinomas and metastatic carcinomas of various tissue origins had nuclear staining when compared to nonneoplastic tissues. There was no significant difference in ATF5 expression in renal cell carcinomas, lymphomas and seminomas, when compared to nonneoplastic tissues. An expanded series of nonarray breast resection specimens revealed a significantly greater proportion of ATF5 positivity in ductal and lobular carcinomas, when compared to normal breast tissue. Past work found that loss of ATF5 function triggers death of GBM cells, but not of normal activated astrocytes. Here, we observed that loss of ATF5 function caused significant apoptotic death of neoplastic breast cell lines, but not of nonneoplastic breast cell lines. Our data demonstrate elevated ATF5 expression in a wide variety of neoplasms and that interference with ATF5 function selectively triggers death of breast carcinoma cells. Such findings may have potential therapeutic application. © 2007 Wiley-Liss, Inc.

Activating transcription factor 5 (ATF5; also referred to as ATFx) is a transcription factor in the ATF/CREB family of basic leucine zipper (bZip) proteins, which has not been extensively studied. Although ATF5 expression has been studied in the brain,1, 2, 3, 4, 5 there are only limited reports describing ATF5 expression in other tissues.6, 7, 8, 9, 10 Previous experiments have shown that ATF5 is expressed in neuroprogenitors2, 3 and, when constitutively expressed in these cells, prevents them from differentiating and allows them to continue to proliferate.2, 3, 5 Conversely, interference with ATF5 function with a dominant-negative construct or siRNA causes neuroprogenitors to prematurely exit the cell cycle and to differentiate.2, 3, 4 In contrast to its presence in neuroprogenitors, ATF5 is undetectable in postmitotic neurons or in mature glial cells.2, 3, 5

Because neoplasias may derive from mitotically active pools of stem or progenitor cells, the presence of ATF5 in neuroprogenitors and its capacity to block cell cycle exit and differentiation raised it as a potentially attractive molecule to consider in the context of tumors. Recently, ATF5 was found to be expressed in all series of 29 human glioblastomas (GBMs), in addition to and all 7 human and rodent GBM cell lines surveyed.4 Another group also confirmed the presence of ATF5 in GBM and the levels of expression appeared to correlate with patient survival.8 ATF5 was also shown to be upregulated in follicular carcinomas of the thyroid.7 In addition, it has been shown that a variety of genes, including ATF5, involved with cell proliferation and inhibition of apoptosis, are downregulated upon treatment of prostate cancer cells with phytochemicals (indole-3-carbinol)9 and chemotherapeutic agents (docetaxel and estramustine).10 ATF5 expression was furthermore reported to be suppressed by the tumor suppressor RASSF1A in nasopharyngeal carcinoma cells,11 again implying a potential role in tumor formation and/or maintenance.

Initial studies on the role of ATF5 in tumor cells indicate that it may be required for their survival. Downregulation or interference with ATF5 function evoked massive apoptotic death in all 7 GBM lines tested in vitro and of 1 tumor-forming line in vivo.4 Interference with ATF5 function in several other tumor-derived lines also induced death.12 Such effects appear to be selective in that loss of, or interference with, ATF5 function did not affect survival of various nontransformed cells that express it, including neuroprogenitors,2, 3 activated astrocytes4 and several cell lines. Thus, it appears that targeting ATF5 may be a potential selective approach for cancer therapy. For this reason, it was desirable to obtain additional information about the expression of ATF5 in various neoplastic and nonneoplastic tissues, as well as to compare the role of ATF5 in the survival of additional neoplastic and normal cell types. Here, we performed immunohistochemical staining of a tissue microarray (TMA) and a series of neoplastic and nonneoplastic breast resection specimens. We also used cell culture to evaluate the effect of ATF5 loss-of-function on the survival of transformed and nontransformed mammary cell lines.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Reagents

Cell culture medium DMEM, molecular biology reagents and LipofectAMINE 2000 were from Invitrogen. Fetal bovine serum was from JRH Biosciences. Rabbit anti-Ki67 was from Novacastra. Normal 10% goat-serum was from Zymed. Polyclonal rabbit anti-ATF5 was as previously described by Angelastro et al.2 DAB reagent was from DAKO Envision System kit.

Microarray

Two TMA (HD-CA-1 Lot no. 022603 and TE30 Lot no. 022704) representing a spectrum of neoplastic and nonneoplastic tissues were obtained from Clinomics Biosciences. The mean age of the study population was 61 years (range 0–92 years) with 171 males (54%), 141 females (45%) and 2 with unknown gender. There was a mean age of 63 years (range 20–92 years) in the group of neoplastic specimens and a mean age of 55 years (range 0–86 years) in the group of nonneoplastic specimens (Table I).

Table I. ATF5 Expression Assessed by Immunohistochemistry in Neoplastic and Nonneoplastic Tissues of the TMA (Tissue Microarray)
Tumor typeTotal0123Percent positive (>25% cells) (%)p-value (+ vs. −)p-value (strong diffuse + vs. −)
  • Adeno, adenocarcinoma; TCC, transitional cell carcinoma; SqCC, squamous cell carcinoma; RCC, renal cell carcinoma; Other, all tissue samples on the TMA with less than 5 samples (acinic cell carcinoma, adenosquamous carcinoma, adenoid cystic carcinoma, astrocytoma, basal cell carcinoma, carcinosarcoma, cystadenoma, islet cell tumor, medullary carcinoma, medulloblastoma, meningioma, mesothelioma, benign mixed tumor, sarcoma). 2X2 contingency tables were done to compare Grades 2 and 3 (positive) to Grades 0 and 1 (negative) for the neoplastic and nonneoplastic categories as an example of statistical difference and to compute a likelihood ratio test statistic using the Fisher's exact test.

  • *

    Statistically significant at p < 0.05.

Neoplastic-overall1865316477063<0.001*0.001*
 Adeno87236243467<0.001*0.002*
 TCC1733110650.02*0.003*
 SqCC34104119590.02*0.17
 Lymphoma74012430.681.0
 Seminoma53020401.00.55
 RCC62022330.590.58
 Metastatic112144730.02*0.09
 Other (<5 samples)196229590.060.04*
Nonneoplastic6236651532
Overall2488922528555

Human breast resection specimens

Paraffin-embedded sections (5 μm) of surgically excised human breast tissue were provided by the Department of Pathology of Columbia University Medical Center.

Immunohistochemistry

The paraffin-embedded microarray and human breast resection slides was stained for ATF5 expression. Paraffin was removed by heating sections at 60°C for 1–2 hr, followed by 3 incubations in 100% xylene for 5 min each. Subsequent incubations were in 100, 75, 50 and 0% ethanol for 5 min each. Antigen retrieval was achieved by incubating the slides in 10 mM citrate buffer (pH = 6.0) at 100°C in a Black and Decker HS 800 steamer for 40 min. Endogenous peroxidase was blocked by incubation with 0.3% hydrogen peroxide for 10 min, followed by 3 washes in water. The tissue was then permeabilized by incubation with 0.04% Tween 20 in TBS (3 times, 5 min each) and immunostained with ATF5 antiserum (1:600) in PBS containing 1% BSA for 1 hr at room temperature. Visualization was achieved with DAB reagent following the manufacturer's protocol (DAKO, Envision System kit). The sections were counterstained with light hematoxylin to reveal nuclei and cellular morphology. The experiment was repeated to show reproducibility of the results. Antiserum against the Ki67 antigen (1:1,000) was used as a positive immunostaining control. The same protocol was performed in the absence of the primary antibody, to show the absence of nonspecific staining in each experiment.

For each sample, the extent of immunohistochemical reactivity in the microarray was evaluated for both the staining intensity and the percentage of tumor cells positive for nuclear ATF5 expression. Immunoreactivity was documented semiquantitatively in the following categories: 0, weak staining in <25% nuclei; 1, strong staining in <25% nuclei; 2, weak staining in >25% of nuclei; 3, strong staining in >25% nuclei. The tissue sample was not evaluated if the tissue was folded or lost in the processing. The slides were reviewed by 2 pathologists (SM, MS) independently and then compared with an initial consensus of 95%. When results were discordant, the sample was reviewed again by both pathologists and a consensus was reached. Tumors were then grouped according to histological subtype, and those with 5 or more tissue samples were compared.

The paraffin-embedded sections of human breast tissue from resection specimens were evaluated for the number of ATF5 positive nuclei (per 200 total nuclei in duct epithelium) within the neoplastic cells to show the percent of neoplastic nuclei that were positive for ATF5. In addition, the invasive carcinoma cells were evaluated for estrogen receptor (ER) expression, progesterone receptor (PR) expression and HER2/neu overexpression. Those tumors with greater than 10% of cells positive for estrogen or PR were considered ER or PR positive, respectively. The tumors receiving a grade of 0 or 1+ by immunohistochemistry were considered not to overexpress HER2/neu, whereas those tumors with a grade of 2+ or 3+ by immunohistochemistry were considered to overexpress HER2/neu.

Cell culture

Breast cancer cell lines (HCC1143, BC13/MDA-MB-435s, BC39/SUM159PT, BC5/MDA-MB-468 and Sum1315) and nonneoplastic cell lines (HMEC, MCF10A) were grown in DMEM medium supplemented with 10% fetal bovine serum. Cells were passaged into 24-well culture dishes for transfections. Mouse embryonic stem (ES) cells were grown as described.13

Transient transfections

Transfections were prepared as previously described by Angelastro et al.,4 pLeGFP mock and pLeGFPfusionFlag-Tagged-NTAzip-ATF5 were transfected into cell monolayers in 24-well dishes using 1 μg of plasmid/well and 2 μg/well of LipofectAMINE 2000 for 6 hr, after which the cells were re-fed with fresh culture medium. Each transfection was performed in triplicate in order to test the reproducibility of the results.

Quantitative assessment of cell death

Transfected cell cultures were fixed and immunostained for the expression of eGFP and ATF5 as previously described by Angelastro et al.,2 and then incubated with Hoechst dye 33342 at 1 μg/ml in PBS and 0.3% Triton X-100 for 5 min at room temperature to detect apoptotic nuclei. eGFP-positive cells that coexpressed ATF5 and possessed condensed nuclei and/or fragmented chromatin were scored as apoptotic.

Immunofluorescent staining

Paraffin-embedded sections (5 μm) were deparaffinized by heating sections at 60°C for 2 hr, followed by 3 incubations in 100% xylene for 5 min each. Subsequent incubations were in 100, 75, 50 and 0% ethanol for 5 min each. Antigen retrieval was achieved by incubating the slides in 10 mM citrate buffer (pH = 6.0) at 100°C in a Black and Decker HS 800 steamer for 40 min. The slides were then blocked using 0.3% Triton X-100 in 10% nonimmune goat serum. The slides were stained for 8–10 hr with 2 primary antibodies, 837 (ATF5, 1:500 dilution) and low-molecular-weight cytokeratin (CK8/18, 1:100 dilution). A third primary antibody CK5/6 (CK5/6, 1:100 dilution) was then added for 2 hr. After 3 washes with PBS, the slides were stained with the secondary antibodies for 2 hr (1:1,000 dilution antirabbit, 1:100 anti-IgG1 and anti-IgM). After 3 washes with PBS, avidin D AMPCA was added for 1 hr. The slides were then washed again, coverslipped and placed in a dark place until visualization.

Statistical analysis

In the TMA, we tested the null hypothesis that there is no difference in the expression of ATF5 between neoplastic and nonneoplastic tissues. For statistical analysis, the tissue was considered positive for ATF5 if >25% tumor cells showed distinct nuclear staining (weak or strong; categories 2 and 3). All nonneoplastic tissues were grouped together because of the small sample size of normal tissue in each organ. In addition, tumors with less than 5 representative samples on the TMA were grouped into 1 category. This “other” group included the following tumors: acinic cell carcinoma, adenosquamous carcinoma, adenoid cystic carcinoma, astrocytoma, basal cell carcinoma, carcinosarcoma, cystadenoma, islet cell tumor, medullary carcinoma, medulloblastoma, meningioma, mesothelioma, benign mixed tumor and sarcoma. On the basis of these criteria, we generated 2X2 contingency tables and computed a likelihood ratio test statistic using the Fisher's exact test (http://www.matforsk.no/ola/fisher.htm). We also compared the difference in intensity by comparing the positive cases with diffuse, intense staining in neoplastic tissues to the nonneoplastic tissues. A student's t test using Microsoft excel spreadsheet software was used to look for statistical significance in the cell culture and breast resection immunohistochemistry data.

Results

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Assessment of ATF5 expression in TMA

To survey the presence and localization of ATF5 protein in a wide variety of tumors, replicate TMAs were stained with specific antiserum to ATF5. Of the 314 specimens on the TMA, 248 were available for analysis on both TMAs. Overall, diffuse positive (>25% of cells, scores 2 and 3) nuclear ATF5 staining was observed in 137/248 (55%) specimens. Cytoplasmic ATF5 expression was noted occasionally, but if present was consistently accompanied by strong nuclear ATF5 staining; hence only nuclear ATF5 labeling was taken into consideration. In addition, expression was more localized to tumor epithelium (intra-epithelial) than to the surrounding tissue. Tables I–III summarize the immunohistochemical data from the TMA.

Table II. ATF5 Expression By Immunohistochemistry in Adenocarcinoma of Different Tissue Types
Tumor typeTotal0123Percent positive (>25% cells) (%)
  1. Tissues represented on the TMA with 5 or more representative samples of adenocarcinoma are compared.

Adenocarcinoma-overall87236243467
Prostate14222871
Colon13304677
Endometrium13513454
Breast11311664
Ovary10402460
Gastric8022475
Pancreas6203167
Lung5103180
Table III. ATF5 Expression by Immunohistochemistry in Nonneoplastic Tissues
Tumor typeTotal0123Percent positive (>25% cells) (%)
  1. Nonneoplastic tissues represented on the TMA with 5 or more representative samples are compared.

Nonneoplastic-overall6236651532
Prostate10302570
Breast7211357
Liver761000
Kidney6210350
Colon5401020
Lung541000

In neoplastic tissues, positive (>25%, scores 2 and 3) nuclear ATF5 immunostaining was seen in 117/186 (63%) of the specimens (Table I and Fig. 1), which is significantly greater than the level of expression seen in 20/62 (32%) of nonneoplastic specimens (p-value < 0.001 Fisher's exact test). When analyzed by histological subtype, nuclear ATF5 labeling occurred in significantly more adenocarcinomas (67%; p-value < 0.001), transitional cell carcinomas (65%; p-value 0.02), squamous cell carcinomas (59%; p-value 0.02) and metastatic carcinomas (73%; p-value 0.02) than in the nonneoplastic tissues. This correlation remained significant for adenocarcinomas when we looked at the difference in intensity and compared the proportion of strongly positive ATF5 staining (score 3) to that of nonneoplastic tissues (p-value 0.002 Fisher's exact test). Table II shows the ATF5 expression in adenocarcinomas of various tissue types on the TMA. In contrast, there was no statistically significant difference between the ATF5 expression in seminomas, lymphomas and renal cell carcinomas when compared to the nonneoplastic tissues.

thumbnail image

Figure 1. Immunohistochemical staining of tumor specimens in a TMA for ATF5. Many carcinomas, including adenocarcinomas (a, b), basal cell carcinomas (c) and squamous cell carcinomas (d), showed nuclear immunoreactivity for ATF5. However, seminomas (e) and lymphomas (f) showed no or minimal nuclear ATF5 staining. The inset in (b) and (f) show positive immunoreactivity for Ki67. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

Figure 2 illustrates an example of the staining pattern seen in adenocarcinomas of the prostate, colon and breast, when compared to their respective nonneoplastic counterparts. While ATF5 staining in these adenocarcinomas shows a strong nuclear signal in the majority of cells, the staining in nonneoplastic epithelia is present in a smaller proportion of cells with variable intensities. In addition, the ATF5-positive cells in nonneoplastic glands are preferentially located at sites (e.g. base of colonic crypts, prostatic basal cells) from where these tissues are known to regenerate.

thumbnail image

Figure 2. Comparison of ATF5 staining in pathologic and nonpathologic tissues. The immunostaining for ATF5 is more uniform and intense in neoplastic adenocarcinomas (b, d, f) than in comparable nonneoplastic tissue (a, c, e) of the corresponding tissue type: prostate (a, b), colon (c, d), breast (e, f). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

As noted earlier and in Tables I and III, positive nuclear ATF5 staining was present in nonneoplastic tissues. Although the number of samples available for analysis of many particular tissue types was limited, for each case with an n > 5, the highest proportion of expression (Table I) included breast (3/7), kidney (7/12) and prostate (5/10). The tissues with the least expression included lung (0/7) and liver (0/7). Taken together, the TMA data indicate that ATF5 is expressed in subsets of a wide variety of normal and neoplastic tissues, and that the proportion and intensity of expression is elevated in many types of neoplasias when compared with nonpathologic tissues.

Expression of ATF5 in pathologic and nonpathologic breast tissue

On the basis of the data obtained from the TMA showing that ATF5 is overexpressed in many adenocarcinomas, we further investigated and quantified its expression in a series of human breast tissue specimens (Department of Pathology, Columbia University Medical Center). We randomly selected 10 cases of invasive ductal carcinomas and 7 cases of invasive lobular carcinomas from our files, and compared the staining to that of the surrounding nonneoplastic breast tissue in these 17 specimens and 5 additional mammoplasty specimens. In addition, we examined staining in a smaller number of in situ ductal and lobular carcinomas. All 17/17 invasive breast carcinomas (lobular and ductal) stained strongly for ATF5 (Figs. 3a3f). We then compared the proportion of ATF5-positive nuclei in nonneoplastic breast tissues to that of the in situ and invasive carcinomas of the breast (Fig. 4). Overall, the proportion of ATF5 positive nuclei was significantly greater in the invasive ductal (87% ± 2%, n = 10; p < 0.0001), invasive lobular (83% ± 3%, n = 7; p < 0.0001), in situ ductal (80% ± 4%, n = 5; p = 0.0001) and in situ lobular (73% ± 7%, n = 3; p = 0.03) carcinomas, when compared to the nonneoplastic breast tissue (45% ± 4%, n = 22). However, there was no direct correlation between ATF5 expression and markers ER/PR or HER2/neu (Table IV). The majority of the neoplastic breast specimens were hormone receptor positive (14/17 or 82% ER positive and 9/17 or 53% PR positive) and were negative for an overexpression of HER2/neu (12/17, 71%) in the invasive carcinoma cells.

thumbnail image

Figure 3. ATF5 expression in ductal and lobular carcinoma of the breast. Three paraffin embedded sections of ductal (ac) and lobular carcinomas (df) were stained with ATF5 antiserum. Nuclear ATF5 staining is evident in almost all tumor nuclei. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

thumbnail image

Figure 4. Comparison of ATF5 positivity in human breast tissue. The proportion of ATF5 positive nuclei was significantly greater in invasive ductal (87% ± 2%, n = 10; p-value 7.1 × 10−11), invasive lobular (83% ± 3%, n = 7; p-value 2.4 × 10−8), in situ ductal (80% ± 4%, n = 5; p-value 0.0001) and in situ lobular (73% ± 7%, n = 3; p-value 0.03) carcinomas, when compared to normal breast tissue (45% ± 4%, n = 22).

Download figure to PowerPoint

Table IV. Comparison of Hormone Receptor, HER2/NEU and ATF5 Expression in Ductal and Lobular Carcinoma
No.DiagnosisERPRHER2/neuPercent ATF5 positivity (in situ)Percent ATF5 positivity (invasive)
1Invasive ductal carcinoma2+9699
2Invasive ductal carcinoma++2+8087
3Invasive ductal carcinoma++1+7084
4Invasive ductal carcinoma++1+NA95
5Invasive ductal carcinoma07581
6Invasive ductal carcinoma0NA86
7Invasive ductal carcinoma++1+NA77
8Invasive ductal carcinoma++1+NA92
9Invasive ductal carcinoma+2+7989
10Invasive ductal carcinoma+1+NA84
11Invasive lobular carcinoma++07581
12Invasive lobular carcinoma++1+8590
13Invasive lobular carcinoma+1+NA85
14Invasive lobular carcinoma+2+NA91
15Invasive lobular carcinoma++1+6170
16Invasive lobular carcinoma+2+NA79
17Invasive lobular carcinoma++1+NA87

To investigate the properties of cells that express ATF5 in the nonneoplastic breast, 5 nonneoplastic breast specimens from reduction mammoplasties were triple stained by immunofluorescence for ATF5 (red), cytokeratin 8/18 (green) which is a marker for luminal cell differentiation14 and cytokeratin 5/6 (blue) which marks mammary progenitor cells.14 Figure 5 shows that the ATF5-positive nuclei are primarily located in the luminal ductal epithelium, colocalize with CK8/18 and/or CK5/6, but are not present in the surrounding stroma of the nonneoplastic breast. Since CK5-positive cells in the breast have been shown to represent a progenitor cell compartment,14 the abundance of CK5-positive progenitor cells in these mammary reduction specimens may be explained by the expansion of the progenitor pool in these hyperplastic mammary glands, which required surgical intervention. Overall, ATF5 is expressed both in the less differentiated CK5-positive cells, as well as in the more differentiated CK8/18-positive compartments, but not in the stroma.

thumbnail image

Figure 5. Cellular localization of ATF5 positive cells in human breast tissue. Nonneoplastic breast specimen was coimmunostained for ATF5 (red-B, D, F), CK8/18 (green-C, E, F) and CK5/6 (blue-A, D, E, F). The ATF5 positive nuclei are primarily located in the ductal epithelium (D) surrounding the lumen (L) and colocalize with both CK8/18 and CK5/6. However, the breast stroma (S) appears to be negative for all markers. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

Interference with ATF5 activity promotes cell death of breast carcinoma cell lines, but not of nonneoplastic breast cell lines in vitro

Interference with ATF5 function results in selective death of multiple rodent- and human-derived GBM cell lines, but not of activated astrocytes.4 To determine whether this might also be the case for a different type of tumor cell, we investigated the effect of loss-of-ATF5 function in neoplastic and nonneoplastic breast cell lines. This was achieved by transfection with a previously described dominant-negative form of ATF5 (LeGFPAzip). For all 5 neoplastic lines tested (Sum1315, BC5, BC39, BC13 and HCC1143), the apoptotic index was significantly greater with the dominant-negative ATF5 than in the same cell lines transfected with the control (LeGFP) construct (Fig. 6). In contrast, survival of nonneoplastic cell lines (MCF10A and HMEC) was not significantly altered (Fig. 7). All neoplastic and nonneoplastic cell lines were variably positive for ATF5 expression (Fig. 6c). Mouse ES cells, which also express ATF5 (Fig. 6b), also did not show enhanced death when transfected with dominant-negative ATF5 (Fig. 7).

thumbnail image

Figure 6. Transfection of neoplastic and nonneoplastic cell lines with a dominant-negative form of ATF5 (a, b) Effect of transfected dominant-negative ATF5 on survival. Nonneoplastic mammary cell line HMEC (a), and the neoplastic mammary SUM 1315 cell line (b) were transfected with dominant-negative ATF5-eGFP (D–F) or with vector expressing eGFP alone (A–C). Four days later, the cultures were fixed and immunostained for eGFP (green, left A and D), ATF5 (red, middle B and E) and Hoechst dye (blue, right C and F). The proportion of apoptotic nuclei in the neoplastic cells transfected with the dominant-negative construct (D–F) was significantly greater than in cells transfected with the control construct (A–C). The arrow in (b) D–F shows representative nuclei with condensed chromatin. (c) Expression of ATF5 in cell lines transfected with dominant-negative ATF5. Nonneoplastic and neoplastic cell lines were transfected with either dominant-negative ATF5-eGFP (LeGFP-Azip, white bars) or eGFP alone (LeGFP, black bars) and assessed 4 days later for proportions of transfected cells expressing ATF5. Lines mouse ES, HMEC and MCF10A are nonneoplastic, whereas lines HCC1143, BC13, BC39, BC5 and Sum1315 are neoplastic. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

thumbnail image

Figure 7. Comparison of the effects of transfection with dominant-negative ATF5 on apoptotic cell death in neoplastic and nonneoplastic human mammary cell lines. The indicated lines were transfected with either dominant-negative ATF5-eGFP (LeGFP-Azip, white bars) or eGFP alone (LeGFP, black bars) and assessed 4 days later for proportions of transfected cells with apoptotic nuclei. Lines HMEC and MCF10A are nonneoplastic, whereas lines HCC1143, BC13, BC39, BC5 and Sum1315 are neoplastic. For each cell line, p values are shown for the difference between cells transfected with eGFP and with dominant-negative ATF5-eGFP.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Our study provides an analysis of ATF5 expression in a variety of neoplastic and nonneoplastic tissues. The use of a TMA provided a large number of tissues and a variety of histologic subtypes of cancer, which allowed a widespread analysis of the distribution of this transcription factor. Overall, nuclear ATF5 appears to be widely present in both neoplastic and nonneoplastic cells. However, it appears that there is a statistically significant difference in the proportions of neoplasms and nonneoplastic tissues that express ATF5. When analyzed by histologic subtypes of cancer, adenocarcinomas, transitional cell carcinomas, squamous cell carcinomas and metastatic tumors showed more extensive ATF5 expression than other types of neoplasia and nonneoplastic tissues. This difference was statistically most significant for adenocarcinomas, when ATF5 distribution (scores 2 and 3), intensity (scores 1 and 3) or both (score 3) were considered. These findings extend previous studies that have shown the expression of ATF5 in GBM multiforme4 and follicular carcinoma of the thyroid.7

The predominance of ATF5 in epithelial carcinomas, particularly adenocarcinomas, and the scarcity of positivity in seminomatous germ cell tumors and lymphomas is particularly notable. Conceivably, this difference may be due to heterogeneity within particular tumor types and the samples on the TMA may not be truly representative. Alternatively, the relative paucity of ATF5 positivity in lymphomas and seminomas more likely reflects a true difference in these types of tumor cells. Studies with neuroprogenitors and activated astrocytes indicate that ATF5 expression may be important for cell cycle progression. Its absence in some tumors may simply be explained by the overexpression of different proteins that can compensate or activate different pathways involved in proliferation and/or survival. This is supported by the positive immunostaining for Ki67, which proved that the lymphomas are mitotically active like many of the adenocarcinomas (Figs. 1b and 1f, inset), yet seem to be negative for ATF5, possibly due to alternative molecular mechanisms. In addition, a recent study showed that both lymphoid malignancies and seminomas overexpress the protein, TCL1 (T-cell leukemia/lymphoma 1), which is an Akt kinase activator, and that its dysregulation may be important in these neoplasms.15 Thus, our results may highlight the fact that some neoplasms, such as seminomas and lymphomas, rely on different pathways and different proteins for their neoplastic phenotype than seen in most carcinomas.

Even when ATF5 was strongly expressed in tumors, it was generally not present in all cells. One possibility is that this reflects the presence of cells at various stages of the cell cycle. It was reported that ATF5 is expressed only during the G1/S phases of the cycle.12 However, examination of GBM cell lines indicated that in this case, ATF5 could be expressed also during the M phase and in all likelihood, other phases of the cycle as well.4 Alternatively, there may be true heterogeneity within tumors with respect to ATF5 expression. Recent studies have suggested that neoplasias may arise from and contain subsets of “stem” cells that are responsible for tumorigenic growth. In this regard, it may be of relevance that in cultures of neural stem cells, ATF5 was coexpressed with CD133, a putative stem cell marker.2 Moreover, in normal breast tissue, ATF5 was present in CK5+ ductal cells, which also appear to be progenitor/stem cells.14

It remains to be seen whether the degree of ATF5 expression within a range of tumor types correlates with patient outcome. However, in a series of GBMs, ATF5 was one of the 314 genes upregulated in the cells surrounding areas of necrosis (pseudopalisading necrosis) and was one of a few genes that were significantly associated with longevity of patient survival.8

When ATF5 was present in normal adult tissues, it was nearly always limited to epithelial structures and ductal epithelia. In the breast, in particular, ATF5 expression was present primarily within the luminal ductal epithelium and was expressed in thenuclei of ductal cells in the less differentiated (CK5+) and more differentiated (CK8/18+) compartments. In other nonneoplastic tissues, the ATF5-positive cells appear to be preferentially located at sites (e.g. base of colonic crypts, prostatic basal cells) from where cells are thought to play a role in repair and regeneration.

To extend past findings showing that ATF5 loss-of-function promotes apoptotic death of GBM cells, but not of normal proliferating activated astrocytes, we carried out a study of 5 breast tumor cell lines and 2 nontransformed breast-derived cell lines. In all cases, ATF5 loss-of-function caused a large and significant increase in apoptotic index in the tumor-derived lines, but not in the nontransformed lines. In addition, such treatment did not cause death of mouse ES cells, which also express ATF5. Thus, studies with 2 tumor types (brain and breast) of very different origins reveal that ATF5 is required for their survival, but is not required for the survival of a corresponding nonneoplastic cell type.

In conclusion, our data shows that ATF5 is expressed in a variety of neoplastic and nonneoplastic tissues, with the greatest extent of positivity and intensity in adenocarcinomas. We also showed that ATF5 is highly expressed in a series of breast carcinomas and that interfering with its function elicits an apoptotic effect in neoplastic breast-derived cell lines, but not in nontransformed lines from the same tissue. The widespread ATF5 expression in carcinogenesis may indicate that it is part of the “molecular signature” in many neoplasms and, if crucial for the maintenance of their survival, may be an attractive therapeutic target.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We thank Dr. Yi-Ji Shi for her excellent technical assistance. In addition, we also thank Dr. Jennifer Yu and Dr. Ramon Parsons for the breast cancer cell lines. We also thank Drs. Cecile Martinat and Asa Abeliovich for the cultures of mouse embryonic stem cells. We acknowledge Mr. Angelo Arias for help in the staining of the TMAs. Supported in part by grants from the NIH (L.A.G.) and the American Cancer Society Institutional Award (J.M.A.).

References

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • 1
    Angelastro JM, Klimaschewski L, Tang S, Vitolo OV, Weissman TA, Donlin LT, Shelanski ML, Greene LA. Identification of diverse nerve growth factor-regulated genes by serial analysis of gene expression (SAGE) profiling. PNAS 2000; 97: 104249.
  • 2
    Angelastro JM, Ignatova TN, Kukekov VG, Steindler DA, Stengren GB, Mendelsohn C, Greene LA. Regulated expression of ATF5 is required for the progression of neural progenitor cells to neurons. J Neurosci 2003; 23: 4590600.
  • 3
    Angelastro JM, Mason JL, Ignatova TN, Kukekov VG, Stengren GB, Goldman JE, Greene LA. Downregulation of activating transcription factor 5 is required for differentiation of neural progenitor cells into astrocytes. J Neurosci 2005; 25: 388999.
  • 4
    Angelastro JM, Canoll PD, Kuo J, Weiker M, Costa A, Bruce JN, Greene LA. Selective destruction of glioblastoma cells by interference with the activity or expression of ATF5. Oncogene 2006; 25: 90716.
  • 5
    Mason JL, Angelastro JM, Ignatova TN, Kukekov VG, Lin G, Greene LA, Goldman JE. ATF5 regulates the proliferation and differentiation of oligodendrocytes. Mol Cell Neurosci 2005; 29: 37280.
  • 6
    Ayers M, Symmans WF, Stec J, Damokosh AI, Clark E, Hess K, Lecocke M, Metivier J, Booser D, Ibrahim N, Valero V, Royce M et al. Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22: 228493.
  • 7
    Barden CB, Shister KW, Zhu B, Guiter G, Greenblatt DY, Zeiger MA, Fahey TJ,III. Classification of follicular thyroid tumors by molecular signature: results of gene profiling. Clin Cancer Res 2003; 9: 1792800.
  • 8
    Dong S, Nutt CL, Betensky RA, Stemmer-Rachamimov AO, Denko NC, Ligon KL, Rowitch DH, Louis DN. Histology-based expression profiling yields novel prognostic markers in human glioblastoma. J Neuropathol Exp Neurol 2005; 64: 94855.
  • 9
    Sarkar FH, Li Y. Indole-3-carbinol and prostate cancer. J Nutr 2004; 134(12 Suppl ): S3493S3498.
  • 10
    Li Y, Hong X, Hussain M, Sarkar SH, Li R, Sarkar FH. Gene expression profiling revealed novel molecular targets of docetaxel and estramustine combination treatment in prostate cancer cells. Mol Cancer Ther 2005; 4: 38998.
  • 11
    Chow LS, Lam CW, Chan SY, Tsao SW, To KF, Tong SF, Hung WK, Dammann R, Huang DP, Lo KW. Identification of RASSF1A modulated genes in nasopharyngeal carcinoma. Oncogene 2006; 25: 3106.
  • 12
    Persengiev SP, Devireddy LR, Green MR. Inhibition of apoptosis by ATFx: a novel role for a member of the ATF/CREB family of mammalian bZIP transcription factors. Genes Dev 2002; 16: 180614.
  • 13
    Lee SH, Lumelsky N, Studer L, Auerbach JM, McKay RD. Efficient generation of midbrain and hindbrain neurons from mouse embryonic stem cells. Nat Biotechnol 2000; 18: 6759.
  • 14
    Bocker W, Moll R, Poremba C, Holland R, Van Diest PJ, Dervan P, Burger H, Wai D, Ina Diallo R, Brandt B, Herbst H, Schmidt A et al. Common adult stem cells in the human breast give rise to glandular and myoepithelial cell lineages: a new cell biological concept. Lab Invest 2002; 82: 73746.
  • 15
    Narducci MG, Fiorenza MT, Kang SM, Bevilacqua A, DiGIacomo M, Remotti D, Picchio MC, Fidanza V, Cooper MD, Croce CM, Mangia F, Russo G. TCL1 participates in early embryonic development and is overexpressed in human seminomas. PNAS 2002; 99: 117127.