C/EBPβ is expressed in numerous tissues, including the liver, adipose tissue, the intestine, the ovary, the lung, the mammary gland, skin, the brain, the kidney, the heart and hematopoietic tissues. Transcription of the intronless C/EBPβ gene results in a single 1.4 kb mRNA that can be translated via alternative start sites into three isoforms: full-length, 38 kDa liver-enriched activating protein (LAP)1; 35 kDa LAP2; and 20 kDa liver-enriched inhibitory protein (LIP).

The most abundant and easily detected isoforms in the mouse mammary gland are LAP2 (35 kDa) and LIP (20 kDa). In human breast tissue, the LAP isoforms are larger (approximately 42–46 kDa) and the LIP isoform remains unchanged at 20 kDa. This increase in size of the LAP isoforms is due to an additional 49 amino acid residues located in the N-terminal portion of the human C/EBPβ protein that are excluded from the LIP isoform. A larger LAP isoform (55 kDa) has also been reported in nontransformed, human mammary cells [5].

Both the LAP and LIP isoforms have the same DNA binding and dimerization domains but, due to translation at an alternative downstream start codon, LIP lacks much of the trans-activation domain (Fig. 1). LIP is consequently unable to activate gene transcription [6]. As with all C/EBP family members, LAP and LIP are capable of forming homodimers or heterodimers with each other. Also, LIP can antagonize the transcriptional activating potential of LAP by competing for DNA binding sites as a homodimer or by its interaction in a LAP/LIP heterodimer [6]. Because of an increased DNA affinity of the LIP isoform, this inhibition of transcriptional activity can occur even at substoichiometric ratios of LIP/LAP, thereby suggesting a dominant-negative function for LIP [6]. LAP and LIP thus have opposing actions on gene activation, and the LAP/LIP ratio, rather than their absolute amounts, may be an important indicator of C/EBPβ transcriptional activity.

C/EBPβ mRNA expression levels are detectable in the virgin gland. These expression levels increase during pregnancy, decline during lactation, and increase again during the onset of involution [7].

C/EBPβ-LAP2 is detected in whole cell extracts of the virgin rat gland and is elevated twofold to threefold with the onset of pregnancy. Although LAP levels decrease at parturition, they are readily detectable throughout lactation and involution [8]. The LIP isoform is expressed at very low levels in the virgin gland and is dramatically upregulated (100-fold) during pregnancy [8]. LIP then decreases to nearly undetectable levels at parturition and remains low throughout lactation [8]. Similar protein expression profiles have been observed in the mouse mammary gland [9].

In summary, the C/EBP-LAP isoform, although variable in expression levels, is detectable at all stages of mammary gland development. The LIP isoform, however, is most readily detected during pregnancy, a period of rapid epithelial cell proliferation. The expression profile of the C/EBPβ isoforms is in part regulated by lactogenic hormones (e.g. glucocorticoids), and it reflects the importance of the isoforms in the expression of milk protein genes such as β-casein and whey acidic protein [8,10,11].

C/EBPα is not as widely expressed as C/EBPβ, and is detected in the liver, adipose tissue, the mammary gland, the lung, the intestine, the adrenal gland, the placenta, and peripheral blood mononuclear cells. The intronless C/EBPα gene encodes a 2.7 kb mRNA that is translated into 42 and 30 kDa isoforms via a leaky ribosome scanning mechanism [6]. As was observed for C/EBPβ-LIP, the truncated 30 kDa isoform of C/EBPα is not as potent a transactivator of gene expression as the full-length 42 kDa isoform [12,13].

Reports on the expression profile for C/EBPα in the mammary gland are not in complete agreement. Some data suggest that C/EBPα levels are the most easily detected in the virgin gland, but are generally low and remain unchanged throughout mouse mammary gland development [9,14]. It has also been hypothesized that the low level of expression observed for C/EBPα mRNA during lactation is the result of a dilutional effect from overly abundant milk protein RNAs [9]. Other workers have demonstrated that C/EBPα mRNA is high during the early stages of lactation, expressed at lower levels during the late stages of lactation/early involution, and increases again during late involution [15]. It is consequently uncertain whether the cascade model of C/EBP regulation, which has been proposed in adipocyte differentiation and which results in activation of C/EBPα by C/EBPβ and C/EBPδ[16], occurs in the mammary gland.

Few studies have examined C/EBPα protein expression during mammary development. Raught et al. demonstrated that the 42 kDa protein isoform is expressed at high levels during lactation and at lower levels during involution in the rat mammary gland [8].

This intronless gene is transcribed as a single mRNA (1.2 kb) that encodes a 269 amino acid protein and is expressed in the intestine, adipose tissue, the lung and the mammary gland. The DNA binding affinity of C/EBPδ is less than that for C/EBPβ and C/EBPα, but the transactivation potentials are comparable [17]. The expression of C/EBPδ mRNA during mammary development is different from that reported for the protein. The reasons for this discrepancy are unclear.

C/EBPδ expression is low throughout pregnancy and lactation, but it increases dramatically (>100-fold) within the first 12 hours of involution. This precedes the induction of C/EBPβ at 24–48 hours [7].

Developmental expression of C/EBPδ protein in the rat mammary gland mimics that of C/EBPβ protein. Low expression levels are present in the virgin/involuted gland, with the highest levels observed during pregnancy, followed by a decrease in expression during early lactation and an increase in expression towards late lactation [8].

C/EBPβ-deficient mice were generated by two independent groups in 1995 [18,19], and were found to have defects in mammary gland development. Mammary glands from C/EBPβ^-/- virgin mice exhibited delayed ductal outgrowth, distended ducts and decreased branching [9,14]. The abnormal ductal morphogenesis in the virgin gland was in part due to a lack of normal ovarian function in the C/EBPβ^-/- mice, and could be partially rescued using wild-type ovarian transplants and treatment with estradiol and progesterone. The proliferative rate, the lobuloalveolar development, and the expression of milk proteins, β-casein and whey acidic protein, were dramatically inhibited in the C/EBPβ^-/- mouse despite hormone replacement [9,14]. An elegant series of transplantation experiments were conducted by both these groups to test whether the impairment in ductal morphogenesis and lobuloalveolar development was intrinsic to the epithelial cells or was mediated via stromal and systemic interactions. C/EBPβ^-/-, C/EBPβ^-/+, or C/EBPβ^+/+ epithelia were transplanted into wild-type stroma (C/EBPβ^+/+) of either host nude mice or syngeneic mice that had been 'cleared' of endogenous epithelial cells. In the reverse experiment, C/EBPβ^+/+ donor epithelia were transplanted into the cleared stroma of C/EBPβ^-/- mice. Both types of transplantation experiments demonstrated that the defects in ductal morphogenesis and in lobuloalveolar development were due to the loss of C/EBPβ in the epithelia and were not a result of defects in stromal signaling that may occur due to C/EBPβ deficiency [9,14]. In conclusion, the C/EBPβ knockout mouse has proven a useful model in determining that C/EBPβ is critical for mammary gland development.

Mice deficient for C/EBPα were also developed in 1995. Studies of mice homozygous for the targeted deletion are limited, however, because C/EBPα knockout mice die within 8 hours of birth due to the defective induction of liver enzymes essential for glucose metabolism [20]. The analysis of mammary gland development in C/EBPα^-/- mice was consequently possible only via transplantation experiments. Seagroves et al. transplanted neonatal mammary anlagen into the cleared fat pads of syngeneic hosts and conducted subsequent serial transplantation of these first generation ductal outgrowths into the cleared fat pads of additional syngeneic hosts [9]. On whole mount analysis of the second generation tissue, no differences in mammary development were detected between the C/EBPα^+/+ or C/EBPα^-/- transplanted glands in virgin, pregnant, lactating or involuted mice [9]. These experiments demonstrate that although C/EBPα is expressed in mammary epithelial cells, the deletion of this gene does not affect normal mammary gland development.

C/EBPδ^-/- mice are viable, fertile and, it is assumed, could lactate and nurse their pups because no mammary gland abnormalities have been reported [21]. In contrast, 85% of C/EBPβ^-/-C/EBPδ^-/- double knockout mice died at the perinatal or early postnatal stage. No gross abnormalities were observed in the surviving neonates, but the mice exhibited defective adipocyte differentiation [21]. No mammary gland abnormalities have been reported for these double knockout mice, and it is unknown whether the gland has been analyzed for defects in development.

Little is known about how the LAP and LIP isoforms independently regulate cell-cycle progression in the intact mammary gland. Unfortunately, only one transgenic mouse study has been published to date that investigates the functional significance of overexpression of the C/EBPβ isoforms in the mammary gland [32]. Likewise, no transgenic mice have been generated that overexpress either C/EBPα or C/EBPδ in the mammary gland.

C/EBPβ-LIP was overexpressed in the mouse mammary gland under the control of whey acidic protein sequences that preferentially target expression to the mouse mammary gland at day 10 of pregnancy, extending throughout lactation [32,33]. Mammary gland abnormalities were detected in the involuted glands of multiparous, aged females. These abnormalities included focal and diffuse hyperplasia in 30–40% of the glands examined and, less frequently, mammary intraepithelial neoplasias (similar to ductal carcinoma in situ) and invasive and non-invasive carcinomas in 9% of the glands [32]. These results demonstrated that overexpression of LIP itself was not sufficient to induce tumorigenesis. Elevated LIP expression can, however, induce proliferation and hyperplasias that may be more susceptible to additional oncogenic hits.

Both in vivo proteolysis and in vitro proteolysis via a calpain protease have been proposed as mechanisms to generate the LIP isoform. Specific in vivo proteolytic cleavage of full-length C/EBPβ has been shown to occur in pre-natal and neonatal mouse liver, and is dependent on C/EBPα expression [44]. Other groups have reported that artifactual in vitro proteolytic cleavage of C/EBPβ can occur and is dependent on the cell lysis procedure [26,45].

It has been suggested that mixing of the cytoplasmic and nuclear compartments could expose the nuclear C/EBPβ to cytoplasmic proteases or could activate nuclear, calcium-dependent proteases via cytoplasmic calcium [45]. Interestingly, Baer and Johnson also demonstrated that proteolytic cleavage of LIP was dependent on the presence of the leucine zipper region, and suggested that dimerization could alter the protein's folded structure and render it more susceptible to proteolysis [45].

Translational control of C/EBP isoform expression has been most thoroughly studied in C/EBPα and C/EBPβ. Because C/EBPα does not play a critical role in mammary development or breast cancer, the present review will emphasize the translational regulation of C/EBPβ.

Translation of mRNA can be regulated by AUG codons within transcript leaders, referred to as upstream AUGs, and associated upstream open reading frames (uORFs) [46]. Upstream AUGs are commonly associated with oncogenes and genes that regulate cellular growth and differentiation [46]. C/EBPβ mRNA contains an out-of-frame uORF that codes for a short, nine amino acid peptide located between the LAP1 and the LAP2 translation start sites [25]. It has been hypothesized that both the LAP1 and the LAP2 translation start sites, as well as the uORF, are embedded within a stem loop structure that may play an important role in the regulation of AUG recognition [25] (Fig. 1).

Several different mechanisms have been described to account for the differential expression of the C/EBPβ isoforms. A leaky ribosome scanning mechanism was initially proposed to give rise to translational initiation at alternate AUG codons [6]. It was suggested that if a leaky ribosome scanning complex were to translate the uORF, it may favor reinitiation further downstream at the LIP AUG codon, which is positioned within a perfect Kozak consensus sequence (see Fig. 1) [25,46]. However, Lincoln et al. have found that the uORF does not facilitate translation of the LIP isoform [47].

The laboratory of Leutz recently conducted an elegant study to address the controversies surrounding the role of the uORF in translational regulation of the LIP isoform versus the proposed mechanism of proteolytic cleavage to generate LIP [34]. Strong evidence was presented to support the hypothesis that the C/EBPβ isoforms arise via differential usage of translation start sites rather than by proteolysis. Leutz and coworkers showed that translation of the LIP isoform in cultured cells is strictly regulated by the uORF, but translation of LAP can occur either by leaky ribosome scanning or by reinitiation of ribosomes immediately after translation of the uORF [34]. Interestingly, increased levels of eIF-4E and eIF-2 favor LIP translation and function through associations with the uORF (Fig. 1) [34]. Both eIF-2 [25] and eIF-4E [48] are overexpressed in breast cancer and support our observation of increased LIP translation in human tumors.

The final translational mechanism is centered around a number of recent observations suggesting that LIP translation may be controlled by specific RNA binding proteins that interact with the 5' region of C/EBPβ mRNA [49]. One of these proteins, the CUG triplet repeat binding protein (CUGBP1), has been characterized in detail. CUGBP1 binds to the CCG repeat located within the uORF of C/EBPβ mRNA [49]. The binding of CUGBP1 to the 5' region of C/EBPβ mRNA induces the translation of the LIP isoform in rat and mouse liver during the acute phase response, and in cell-free translation systems [49,50]. These observations suggest that the production of the dominant-negative LIP isoform can be controlled at different levels through several mechanisms.