Organ-specific expression profiles of rat mammary gland, liver, and lung tissues treated with targretin, 9-cis retinoic acid, and 4-hydroxyphenylretinamide Free

A rexinoid, targretin, and two retinoids, 9-cis retinoic acid (9cRA) and 4-hydroxyphenylretinamide (4HPR), were examined for their effects on gene expression in rat mammary gland, liver, and lung tissues. The chemopreventive effects of these agents have largely been attributed to their ability to interact with retinoic acid receptors (RAR) and/or retinoid X receptors (RXR). Targretin interacts with the RXR receptors. 9cRA interacts with both the RAR and RXR receptors, whereas 4HPR has a moderate affinity primarily for RAR γ. Based on previous studies on mammary chemoprevention, targretin (150 mg/kg diet), 9cRA (100 mg/kg diet), and 4HPR (782 mg/kg diet), were administered to rats continually in their diet for 7 days. Tissue- and agent-specific expression differences were determined by comparing tissues from treated rats with those from rats given a control diet. There were significantly more changes associated with targretin than 9cRA or 4HPR. Only a limited number of expression changes were found with 4HPR treatment. For each organ, targretin- and 9cRA-treated tissues clustered closely together, whereas 4HPR-treated tissues clustered with the tissues from the control diet group. In contrast to 9cRA treatment, targretin treatment altered genes that involved fatty acid metabolism and modulation of various cytochromes P450 in the liver, clearly demonstrating the very disparate nature of these two retinoids. These expression signatures could provide useful pharmacodynamic biomarkers for retinoid treatment and chemoprevention. [Mol Cancer Ther 2006;5(4):1060–72]

Retinoids and rexinoids are perhaps the most studied agents for use in cancer chemoprevention (1–4). Retinoids are vitamin A analogues that function in regulating cell growth, differentiation, and apoptosis (1–4). Retinoids bind to specific nuclear receptors, i.e., retinoic acid receptors (RAR) and retinoid X receptors (RXR; refs. 5, 6). In turn, these receptors, which routinely function as heterodimers, bind to specific DNA sequences to regulate gene expression (7). One of the earliest retinoids examined for its ability to inhibit mammary tumorigenesis was 4-hydroxyphenylretinamide (4HPR). 4HPR minimally binds RXR receptors and seems to have only moderate affinity for the RARγ receptor. Almost two decades ago, 4HPR was shown to be an effective chemopreventive agent in the N-methylnitrosourea-induced rat mammary tumor model when given in the diet beginning 7 days prior to methylnitrosourea treatment (8). However, 4HPR shows more limited activity when given post-methylnitrosourea (Grubbs et al., data not shown). Based in part on these preclinical data, 4HPR was employed in a relatively large clinical trial (9).

The RXR receptors were identified 12 years ago and a naturally occurring retinoid, 9-cis retinoic acid (9cRA) was found to interact with these receptors with high affinity (10). 9cRA is a bifunctional retinoid activating both RAR and RXR receptors (11). Initial studies in animal models have shown that 9cRA is effective in preventing mammary tumors in the methylnitrosourea-induced rat mammary tumor model (12). However, the use of 9cRA in humans has been limited by toxicities including severe headaches and skin reactions (13).

Targretin, also known as LGD1069 or bexarotene, is a selective ligand for the RXR receptors. The targretin agent has proven to be highly effective in preventing methylnitrosourea-induced rat mammary tumorigenesis (14). Furthermore, targretin inhibited cell growth and caused complete regression in 72% of methylnitrosourea-induced rat mammary carcinomas at higher doses (14, 15). In addition, targretin effectively suppresses estrogen receptor–negative tumor development with minimal toxicity in mouse mammary tumor virus-erbB2 transgenic mice (16). These studies suggest that receptor-selective retinoids are promising agents for the prevention of breast cancer and that they may be particularly useful in preventing estrogen receptor–negative breast cancer. The RXR agonist targretin has shown some clinical efficacy both in lung cancer (17) and in the treatment of cutaneous T cell lymphoma (18).

Clinical trials with retinoids that primarily act on the RAR receptors have shown activity against certain cancers. For example, 13-cis-RA reduces aerodigestive tract tumors in patients with resected head and neck cancers (19). Other successful retinoid trials have been reported, including cervical dysplasia (20) and xeroderma pigmentosum (21). Treatment with all-trans-retinoic acid is a relatively standard therapy for the treatment of acute promyelocytic leukemias (22). All of these studies were done with agents that primarily work by activating the RAR receptors.

In the present study, we analyzed the gene expression profiles of normal tissues compared with targretin-, 9cRA-, or 4HPR-treated rat mammary gland, lung, and liver. The overall objective of this study was to employ Affymetrix oligonucleotide array analysis to determine whether targretin, 9cRA, or 4HPR: (a) gave distinctly different gene expression patterns and therefore are likely to cause their effects by different mechanisms; (b) exhibited significant overlap in altered gene expression for the same agent in different tissues, implying that the same mechanisms may be in effect in different tissues; (c) caused gene expression changes in multiple normal tissues (mammary gland, lung, and liver) that may reveal important clues that potentially underlie their varying chemopreventive efficacy; and (d) identified genes that may be useful as potential pharmacodynamic or surrogate biomarkers for use in clinical trials with these agents.

Agents (targretin, 4HPR, and 9cRA) were obtained from the National Cancer Institute Chemical Repository (Bethesda, MD). Teklad mash (4%) diet and Sprague-Dawley rats were obtained from Harlan Sprague-Dawley, Inc. (Indianapolis, IN). Diets were prepared by mixing appropriate amounts of targretin, 9cRA, or 4HPR with Teklad (4%) mash diet using a liquid-solid blender (Patterson-Kelly Co., East Stroudsburg, PA). Female Sprague-Dawley rats were housed in polycarbonate cages. The treatment was similar to that previously described (8, 23). Briefly, starting at 70 days of age, the rats received treatment of targretin (60 mg/kg body weight, gavage), 9cRA (100 mg/kg diet), or 4HPR (782 mg/kg diet) for 7 days. At the end of the 7-day treatment, normal mammary gland, liver, and lung tissues were collected. Total RNA was isolated using the TRIzol reagent (Life Technologies, Inc., Rockville, MD). The quantity of the RNA was measured using a spectrophotometer at wavelengths of 260 and 280 nm. The quality of the RNA was monitored by agarose gel electrophoresis.

The total RNA was used to generate cRNA probes. We have analyzed a total of 43 tissues samples (14 mammary gland, 14 lung, and 15 liver samples) with 3 or 4 samples from vehicle control, targretin, 9cRA, or 4HPR groups. All protocols used for mRNA reverse transcription, second strand synthesis, production of cDNA, cRNA amplification, hybridization, and washing were done as provided by the manufacturer (Affymetrix, Santa Clara, CA). Intensity data from CEL files were log-transformed, and normalized to make arrays comparable. Resulting scaled intensity data were used as a basis for computing model-based estimates of gene expression according to the method of Li and Wong (24). Gene expression values across samples were transformed to standard normal deviates and colors assigned on that basis, with green representing below average expression, red representing above average expression, and black representing near average expression. Transcripts with average expression ≤ 150 were excluded. Transcripts were considered differentially expressed if the mean gene expression for one group was at least 2-fold difference from the mean for the contrasted group and a statistical difference of P < 0.05 was obtained. The microarray data presented in this report has been deposited to the National Center for Biotechnology Information with an accession number of GSE3952.

The method has been previously described in detail (25). Briefly, total RNA was reverse-transcribed into cDNA with oligo15(dT) and Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI). In PCR, primer pairs for the selected genes and rat β-actin were used in each reaction. The PCR products were loaded on the agarose gel. The signals were collected and the relative intensities of the target products were then normalized to the level of β-actin control.

For protein immunoblotting (26), liver tissues were homogenized on ice using Dounce homogenizer, in cold lysis buffer [25 mmol/L Hepes buffer (pH 7.4) containing 150 mmol/L NaCl, 1% igepal, 10% glycerol, 2.5 mmol/L EDTA] and protease inhibitor cocktail (Roche Diagnostic, Mannheim, Germany). The homogenates were centrifuged at 16,000 × g for 20 minutes. Supernatants were collected and protein concentration measured with the bicinchoninic acid assay (Pierce Biotechnology, Rockford, IL). Equal amount of the supernatant proteins (25 μg per sample per lane) were resolved by SDS-PAGE (4–12% Bis-Tris gel, Invitrogen Life Technologies, Carlsbad, CA) and transferred to polyvinylidene difluoride membrane (Invitrogen). After blocking with 5% nonfat milk, the membranes were probed with monoclonal antibodies to insulin-like growth factor binding protein (IGFBP-2) and cyclin D1 (Santa Cruz Biotechnology, Santa Cruz, CA), monoclonal anti-rat cytochrome P450 (2B1) antibodies (Oxford Biomedical Research, Oxford, MI), or rabbit polyclonal antibodies to IGFBP-3 (Santa Cruz Biotechnology).

Figure 1 shows the expression profiles of transcripts (including known genes, unknown expressed sequences, and expressed sequence tags) of mammary gland (Fig. 1A), liver (Fig. 1B), and lung (Fig. 1C). Targretin modulated 316 transcripts in the mammary gland, 155 in the liver, and 120 in the lung. 9cRA modulated 48 transcripts in the mammary gland, 73 in the liver, and 24 in the lung. 4HPR modulated only a single transcript in mammary gland, 34 in the liver, and none in the lung. Untutored clustering of samples from liver, lung, and mammary gland exhibited that samples from each organ were clearly separated (Fig. 2) presumably reflecting tissue specific gene expression. The dendrogram showed that four untreated tissues and 4HPR-treated tissues clustered closely, whereas the targretin-treated tissues clustered closely together with the 9cRA-treated tissues. These findings indicate robust data sets that can distinguish untreated tissues from retinoid-treated tissues, particularly from targretin- and 9cRA-treated samples after 7 days of exposure to retinoid. These studies also illustrate that roughly half of the genes modulated by 9cRA were similarly modulated by targretin, with 21 of 48 genes in the mammary gland, 39 of 73 in the liver, and 10 of 24 in the lung, which implies a significant overlap.

Of the genes altered in the mammary gland, a total of 316 were modulated by targretin, presumably reflecting gene changes in various cell types. Among the 316 genes modulated by targretin, 182 (98 up-regulated and 84 down-regulated) are known genes (Table 1). A total of 48 altered transcripts were modulated by 9cRA, whereas only 1 was modulated by 4HPR. As shown in Fig. 1A, several transcripts were modulated by more than one retinoid in the mammary glands. There were 21 genes modulated by both targretin and 9cRA in the mammary gland which represented >40% of the genes modulated by 9cRA but <10% of the genes modulated by targretin (Table 2). The top 36 genes modulated by targretin were clustered into three major groups (Fig. 3). Clusters A and B were up up-regulated, whereas genes in cluster C were down down-regulated. Cluster A contained genes involved in fatty acid metabolism (acyl carrier protein domain of fatty acid synthetase and aldolase), genes coding for ribosomal proteins (ribosomal protein L18a, ribosomal protein L6), and genes involved in biochemical functions (noninducible carbony reductase, a regulatory subunit of protein phosphatase 2A, sulfotransferase). Cluster B consisted of genes that encode for either structural proteins (collagen and α-actin) or growth factors (Wistar transforming growth factor β-3 and glioma-derived vascular endothelial cell growth factor). Cluster C contained P450f, P450 mRNA, P450 2c39, hydroxysteroid sulfotransferase, and α-1 inhibitor III. Certainly, the most highly induced genes represent potential pharmacodynamic markers for potential clinical trials. A potential pharmacodynamic marker would be gene expression changes that were observed at effective doses of an agent. These specific markers do not necessarily reflect either a specific mechanism for efficacy nor serve as surrogates for efficacy.

Among the total 155 genes modulated by targretin in the liver, there were 70 (43 up-regulated and 27 down-regulated) known genes. As shown in Table 3, the top regulated genes were involved in fatty acid metabolism (mitochondrial carnitine palmitoyltransferase II, peroxinyl 3-ketoacyl-CoA thiolase, acyl carrier protein domain of fatty acid synthetase, cytochrome P450 IVA1, malic enzyme). All of these genes can be induced by PPARα agonists, which themselves form a heterodimer with RXR agonists. Another group of genes involves a wide range of P450-related enzymes [cytochrome P450e (CYP2B2), cytochrome P450b (CYP2B1), and NADPH cytochrome P450 reductase], all of which can be induced by phenobarbital-like compounds via the nuclear receptor CAR. The CAR receptor, similarly to the PPARα receptor, forms a heterodimer with the RXR receptors. The third distinct group of genes that was decreased in targretin-treated rats were P450-related enzymes (CYP 1B1 and flavin monooxygenase), which were under the control of nuclear receptor designated AhR. The AhR receptor, in contrast to the PPARα and CAR receptors, does not directly interact with the RXR receptors.

When we examined genes whose expression was altered by 9cRA, there are few genes involved in fatty acid metabolism or phenobarbital type induction that were induced in these animals (data not shown). There are 73 transcripts that were modulated by 9cRA. Thirty-nine transcripts overlapped between targretin and 9cRA and 15 overlapped between 4HPR and 9cRA (Fig. 1B). Fifty-two percent of the genes modulated by 9cRA (39 of 73) were modulated by targretin, whereas only 28% (39 of 144) of the genes modulated by targretin were modulated by 9cRA. In the limited examples of overlap, the genes modulated by both targretin and 9cRA were modulated more strongly in livers from rats treated with targretin. Among the genes uniquely modulated by 9cRA were tyrosine aminotransferase, at least one form of glutathione S-transferase, arginase, glycine methyltransferase, UDP-glucuronosyltransferase (phenobarbital-inducible form), nerve growth factor–induced factor A, and metallothionein-2 and metallothionein-1 genes, etc. 4HPR modulated the fewest genes (16 transcripts) with 3 overlapped with targretin (MHC class II-associated invariant chain, cytocentrin, and fatty acid synthase), 2 overlapped with 9cRA (sulfotransferase family 1A, member 2, and purine-selective sodium/nucleoside cotransporter). However, 13 4HPR-modulated transcripts were overlapped with both targretin and 9cRA (IGFBP, DNA polymerase α, cytochrome c oxidase, stearyl-CoA desaturase, P49 cytochrome P450, ribosomal protein L21, and L1 retroposon ORF2, plus 6 unknown genes).

A total of 120 transcripts were modulated by targretin in the lung. Among them, 110 genes were targretin-specific genes. Sixty-six (36 up-modulated and 30 down-modulated genes) were known genes (Table 4). There were 14 transcripts specifically modulated by 9cRA. These included vascular α-actin, ribosomal protein L21, liver stearyl-CoA desaturase, unr protein, ATPase, Na⁺K⁺ transporting, α2, eukaryotic translation elongation factor 1α1, ATP-binding cassette, subfamily D (ALD), and seven unknown genes. There were 10 transcripts modulated by both targretin and 9cRA, including the phosphatidylinositol 3-kinase p55 and squalene synthetase, and no genes were modulated by 4HPR (Fig. 1C).

In order to assess whether gene expression data from the Affymetrix oligonucleotide arrays were an accurate depiction of the transcription, we did semiquantitative RT-PCR on 10 randomly selected genes (cytochrome p450 2C22, cysteine sulfinic acid decarboxylase, β defensin-2, phenobarbital-induced cytochrome P450b, phenobarbital-induced cytochrome p450e, Schwannoma-derived growth factor, hydroxysteroid sulfotransferase, cytochrome p450 2C22, prostaglandin D synthase, and IGFBP). Nine of the 10 reactions worked and yielded a specific PCR product. The PCR primers for IGFBP did not work and failed to present a unique RT-PCR product. The results from eight of the nine genes confirmed the results of the array data including cytochrome p450 2C22, cysteine sulfinic acid decarboxylase, β defensin-2, phenobarbital-induced cytochrome P450b, phenobarbital-induced cytochrome p450e, Schwannoma-derived growth factor, hydroxysteroid sulfotransferase, and cytochrome p450 2C22. Figure 4 shows RT-PCR results on five of the confirmed genes (cysteine sulfinic acid decarboxylase, β defensin-2, phenobarbital-induced cytochrome P450b, phenobarbital-induced cytochrome p450e, and cytochrome p450 2C22). In addition, we confirmed the microarray results using immunoblotting. Four antibodies, IGFBP (IGFBP1), cyclin D1, cytochrome P450 (CYP2B1), and IGFBP-3, were randomly selected for immunoblotting on proteins isolated from five liver tissues treated with targretin, and four liver tissues treated with 9cRA, plus five normal diet livers as control (Fig. 5). In targretin-treated liver tissues, altered expressions of all four tested genes in protein immunoblotting are in agreement with the microarray data. On the other hand, in 9cRA-treated liver tissues, altered expressions of three out of four tested genes (e.g., IGFBP, CYP2B1, and IGFBP-3) in protein immunoblotting are in agreement with the microarray data. Although microarrays did not show any significant difference of cyclin D1, immunoblotting showed an obvious decreased expression of cyclin D1 in 9cRA-treated livers (Fig. 5). In summary, 17 genes/tissues were subjected for independent confirmation, and 15 of them were validated by either RT-PCR or immunoblotting. The confirmation rate was ∼88%.

In this study, we employed histologically normal liver, lung, and mammary glands to investigate gene expression changes associated with targretin, 9cRA, and 4HPR treatments. The duration of exposure was 7 days because we felt that this subchronic exposure was closer to the proposed clinical preventive trials than short-term exposures of 24 hours or less. The liver is relatively homogeneous in terms of cell types, the lung represents a wide variety of cell types, and the mammary gland has not only epithelial cells but also a high percentage of adipocytes. The gene expression data obtained by using normal mammary gland and normal lung as well as liver have an immediate implication because monitoring of altered gene expression in histologically normal “at-risk” tissues have been proposed for use in phase II chemoprevention trials. The use of in vivo bioassays is of particular importance because it does not allow one to employ unrealistically high levels of a given agent, which are often employed in in vitro studies. Also, the gene changes reflect not only the effects of the agent on the specific cells of interest, but also reflect secondary changes due to effects on stromal elements as well as more physiologic changes due to alterations in organs such as the liver. We found distinctly different patterns for gene expression modulation for each of the three agents. Targretin and 9cRA were expected to have substantial overlap based on the fact that both would interact with the RXR receptors. However, they do not, which may reflect the relative doses of each that can be achieved in vivo. Thus, a dose of >120 ppm of 9cRA in the diets of rats causes severe weight effects, whereas a dose of targretin >250 ppm is necessary to cause any significant weight effects (Grubbs et al., data not shown). These studies show that array analysis is relatively powerful in yielding very distinct biological patterns for agents within a given class. The very distinct pattern difference between targretin and 9cRA probably indirectly argues for a distinct mechanism of activity between the two agents.

Targretin, 9cRA, and 4HPR have differing abilities to prevent mammary tumorigenesis in the methylnitrosourea-induced mammary carcinogenesis in rats. Targretin (150 ppm), 9cRA (100 ppm), and 4HPR (782 ppm), when given in the diet beginning 5 days following methylnitrosourea treatment, reduced tumor multiplicity by 80%, 50%, and 20%, respectively.⁴

Targretin and 9cRA exhibited a higher efficacy than that seen for 4HPR. However, one must be aware that when initiated after giving methylnitrosourea in the standard rat mammary model, in 50-day-old rats, 4HPR is relatively ineffective (23) in contrast to strong activity when given prior to methylnitrosourea (8). Targretin was highly effective in the rat mammary tumor model (14, 15) and shows a significant efficacy in mouse lung tumor model (27). Targretin seems to have an effect both in NSCLC (in conjunction with standard chemotherapy) and in cutaneous lymphomas. 4HPR has proved to be strongly effective in the rat mammary tumor model when given beginning at the time the carcinogen is administered (8), and has proven to be effective in other animal models of cancer as well (28). Thus, there seems to be an organ specificity regarding the chemopreventive efficacy in animal models for each of these three agents.

The differences in chemopreventive efficacy for targretin, 9cRA, and 4HPR in the rat mammary tumor model can, at least in part, be attributed to the known differences in receptor activity. 4HPR and 9cRA are retinoids based on a related retionic acid backbone, whereas targretin is a synthetic rexinoid. 4HPR interacts with moderate affinity to RAR γ, 9cRA interacts with high affinity to both the RAR and RXR receptors, whereas targretin interacts with high affinity to the RXR receptors and has much lower affinity for the RAR receptors. Both 9cRA and targretin interact with the RXR receptors with high affinity. Therefore, we hypothesized that 9cRA and targretin were likely to have strongly overlapping gene expression patterns. Because the liganded RXR receptors themselves form a heterodimer with the widest range of nuclear receptors, including PPARs, CAR, LXR, VDR, TR, etc. (29), we felt that we would obtain a relatively varied group of genes with altered expression but with striking similarities between 9cRA and targretin. However, the results were somewhat surprising. When we first looked at the liver, we found that targretin altered the widest variety of genes. Many of the identified genes were related to the expected RXR interactions. Thus, many genes related to fatty acid metabolism that can be induced by PPARα agonists, e.g., mitochondrial palmitoyl transferase II, peroxinyl 3-keto-acyl-CoA thiolase, acyl carrier protein domain of fatty acid synthetase, cytochrome P450 IVA1, malic enzyme, fatty acid synthase, long chain acyl-CoA thioesterase, and carnitine ocranoyl-transferase, were similarly modulated by targretin. Targretin modulated the expression of various cytochrome P450–related enzymes, which are directly, related to the CAR receptor, e.g., cytochrome P450e (CYP 2B2), cytochrome P450b (CYP 2B1), and NADPH cytochrome P450 reductase (30). Thus, the addition of targretin alone seemed to partially substitute for the effects of administering PPARα and CAR agonists (e.g., phenobarbital, diphenylhydantoin, and dichlorodiphenyltrichloroethane, etc.). However, the level of induction of the CAR-related genes by targretin seems to be considerably less than the levels of induction achieved by optimal doses of phenobarbital. Thus, we have recently found that CYP 2B1 is induced almost 80× by an optimal dose of phenobarbital.⁵

We expected that we might see a similar pattern from treatment with 9cRA on the rationale that it interacted with both RXR receptors and RAR receptors. Surprisingly, we observed only a limited overlap when examining which genes were simultaneously modulated by targretin and 9CRA. Thus, when addressing genes related to fatty acid metabolism, 9cRA significantly induced fatty acid synthase, long-chain acyl-CoA thioesterase and carnitine ocranoyl-transferase, all at substantially lower levels than that achieved by targretin exposure. 9cRA failed to completely induce genes related to the CAR receptor, e.g., cytochrome P450e (CYP2B2), cytochrome P450b (CYP2B1), and NADPH cytochrome P450 reductase. In contrast, we found that roughly 50% of the genes modulated by 9cRA were similarly modulated by targretin. Our results imply that although 9cRA may look like a weak RXR agonist and show overlap, a stronger RXR agonist induced a variety of additional genes. This result may be in some contrast with certain in vitro studies, which implied a greater overlap between the two agents, but this may not reflect the toxicities that limited the dose of 9cRA which one can employ in vivo.

One of the questions was whether the same gene was modulated in different target tissues when treated with the same agent. As shown in Table 5, one observes limited overlap regarding the modulation of genes with regard to targretin. There were a few genes related to fatty acid metabolism that were modulated in both liver and mammary gland. We identified only one gene, the sIRR-2 (insulin receptor–related receptor alternatively spliced product), with altered expression level in mammary gland, liver, and lung after receiving targretin for 7 days (Table 5). There were 6 genes (acyl-CoA oxidase, pre-pro–epidermal growth factor, and squalene synthetase) with altered expression levels in both liver and mammary gland, 13 genes (e.g., inhibitory glycine receptor, ActRIIB, GHRH, mitochondrial fumarase, etc.) in both mammary gland and lung, and 6 genes (taurine transporter, cytochrome P450-LA-ω, TPM-γ, Kcnj16, and two unknown sequences) altered in both liver and lung after receiving targretin for 7 days (Table 5). This result implies that the RXR interactions with the wide raft of nuclear receptors are significantly tissue-specific. This is not completely surprising because the RXR receptors may work primarily by interacting with other nuclear receptors. Because the levels of those other receptors vary from tissue to tissue, this alone will substantially affect the response. The CAR receptor seems to function almost solely in the liver. Similarly, PPARα, which is a major nuclear receptor in liver, seems to be minimally expressed in the mammary gland. This decreases the likelihood that one could use the same pharmacodynamic marker in different tissues, but it does not negate the possibility of determining a series of pharmacodynamic markers in different tissues.

The results from this study offer reasonable clues about relevant biological mechanisms of the agent. For example, several genes were up-regulated including mitogen-activated protein kinase kinase kinase 1 (MEKK1), calreticulin, caveolin-1, and cellular retinol-binding protein. MEKK1, a member of the mitogen-activated protein kinases, was up-regulated by targretin in mammary gland. This effect was not seen in either 9cRA-treated or 4HPR-treated groups in any tissue. Previous studies indicated that retinoids cause activation of mitogen-activated protein kinases in the myeloid leukemia cell line HL60. Most importantly, this activation is necessary for RA-induced growth arrest and cellular differentiation. The expression of an active form of either MEKK1 or MEKK4 in P19 cells mimicked the action of RA by inducing these embryonal carcinoma cells to differentiate into primitive endoderm, via protein kinase C and the extracellular signal-regulated kinase of the mitogen-activated protein kinase family genes (31). Thus, modulation of MEKK1 expression by targretin may contribute to the preventive response observed. Calreticulin is a newly identified member of the antigen processing machinery (32), which was implicated in tumor immunology and in apoptosis (33). Calreticulin was expressed at a lower level in cell lines of the melanoma and lung carcinoma (34). We have shown here that calreticulin was up-regulated by targretin, specifically in the mammary gland, indicating that targretin may have an effect in increasing immuno-defense and apoptosis. Caveolin-1 is a tumor suppressor gene in breast cancer (35) and its expression was down-regulated in many tumors and in oncogene-transformed and tumor-derived cells. Caveolin-1 inhibits transformation-dependent processes (anchorage-independent growth, cell proliferation rate, and capacity to form colonies in soft agar; ref. 36). However, as we have previously shown, targretin seems to mediate its chemopreventive effects by directly affecting lesions because removal of targretin results in the rapid outgrowth of lesions (15). Thus, gene expression changes in normal mammary glands, which seem relevant to the efficacy of targretin, may not significantly contribute to the overall efficacy of this compound if the primary target is the lesions themselves.

In summary, we wished to reiterate the questions which we initially raised and to explain how the present data helped to elucidate these questions: (a) do different “retinoids” yield similar gene expression profiles following in vivo treatment? Different agents, although all of which are called retinoids and each has a different interaction with various RAR and RXR receptors, yield substantially different gene expression patterns and probably work differently. (b) Does a specific agent induce similar gene changes in different tissues? Targretin and 9-cis RA generally alter the expression of different genes in various tissues. (c) Can some of the gene expression changes observed in normal tissue be used to define a likely mechanism of action for an agent? Although some of the gene changes observed in normal tissue may make some sense mechanistically, e.g., increased calreticulin in mammary glands of targretin-treated rats, the effects of targretin are probably on the lesions themselves (15). (d) Could the gene changes defined by the arrays prove to be useful pharmacodynamic markers for clinical trials? Thus, their modulation in clinical trials would at least imply that you had achieved a significant level of the preventive agent, although that would not assure that it would work in humans.