Prolactin signalling to milk protein secretion but not to gene expression depends on the integrity of the Golgi region

Cell sites of signal transduction, after the binding of a hormone to its receptor, have been believed to occur mainly at the plasma membrane. However, the internalisation of ligand-receptor complexes into the endosomal apparatus may be consistent with intracellular signalling (Baas et al., 1995; Bevan et al., 1996). As prolactin (PRL) is internalised in lactating mammary epithelial cells (MEC) and transcytosed to the lumen (Ollivier-Bousquet, 1998), these cells provide a model to study whether, during its transport through the cell, PRL initiates distinct signalling cascades that lead to different biological responses. PRL exerts many biological actions through the prolactin receptors (PRLR), such as the control of casein gene expression (Hennighausen et al., 1997) and the in vitro stimulation of protein secretion, which is referred to as the secretagogue effect (Ollivier-Bousquet, 1978). Transduction of the PRL message within the cell from the cell membrane to the nucleus, which causes activation of the transcription of milk protein genes, includes oligomerisation of the receptor and activation of the self-phosphorylating Janus kinase 2 (Jak2), which in turn phosphorylates tyrosine residues on the receptor and the signal transducers and activators of transcription (STATs), Stat5a and Stat5b. These factors dimerise, translocate to the nucleus and then bind to specific promoter elements on PRL-responsive genes (Hennighausen et al., 1997; Goffin and Kelly, 1997). However, molecular events that lead to the transduction of the secretagogue effect of PRL remain partly unknown. They include a very rapid and transient release of arachidonic acid within a few minutes of the addition of PRL, and the metabolism of this polyunsaturated fatty acid (PUFA) (Blachier et al., 1988). Moreover, arachidonic acid, when added to the incubation medium of lactating MEC, stimulates casein secretion, probably after being metabolised in products of the lipoxygenase pathway (Ollivier-Bousquet, 1982; Ollivier-Bousquet, 1984). Together these results strongly suggest that prolactin stimulates casein secretion by the release of arachidonic acid.

In contrast to hormones that are endocytosed and then rapidly degraded in lysosomes (Burgess et al., 1992), PRL is carried through the MEC by transcytosis and then released in the milk in either intact or cleaved molecular forms (Sinha, 1995). PRL is transported via a vesicular pathway that includes not only endosomes, late endosomes and multivesicular bodies, but also vesicles located in the Golgi region and secretory vesicles containing casein micelles (Ollivier-Bousquet, 1998). The transcytosis of PRL can be slowed down or inhibited in tissues incubated at low temperatures (25°C) for 1 hour (Seddiki et al., 1991), or in mammary tissues from rats that receive a PUFA-deficient diet (Ollivier-Bousquet et al., 1993; Ollivier-Bousquet et al., 1997). Under these two conditions, in which PRL accumulates in late endosomes and multivesicular bodies, basal secretion is not affected but the secretagogue effect of the hormone is inhibited (Seddiki et al., 1991; Ollivier-Bousquet et al., 1993; Ollivier-Bousquet et al., 1997). One question that arises is whether the journey of PRL or the PRL-PRLR complex across the cell may be related to the PRL signalling. As PRL passes through the Golgi region and secretory vesicles during transcytosis, we wondered whether the integrity of these organelles is required for the transmission of PRL effects. To address this point, the effect of disorganising the Golgi region by brefeldin A (BFA) (Pauloin et al., 1997) was assessed in terms of PRL endocytosis, transcytosis, signalling to milk gene expression and the secretagogue effect.

Hanks’ medium was purchased from Gibco (BRL-Life technologies, Cergy Pontoise, France). L-[4,5−³H]leucine, 1−¹⁴C arachidonic acid, horseradish peroxidase (HRP)-conjugated streptavidin and an enhanced chemiluminescence detection (ECL) kit were purchased from Amersham (UK). Ovine PRL was a gift from Dr A. F. Parlow (National Hormone and Pituitary Program, Baltimore, MD, USA). Bromocriptine (CB154) was a gift from Sandoz Pharmaceuticals. Biotinylated ovine PRL (bioPRL) was prepared using a biotinylation kit (Amersham). Nitrocellulose membranes were purchased from Schleicher and Schuell (Veen, NH). The 5G2 monoclonal antibody against oPRL was a gift from Dr J. G. Scammell. Goat polyclonal anti-rabbit prolactin receptor antibody 46 was kindly provided by Dr J. Djiane. Antiphosphotyrosine monoclonal antibody 4G10 was purchased from Upstake Biotechnology (Lake Placid, NY). All other reagents were obtained from Sigma.

Mammary acini and mammary fragments were prepared from New Zealand rabbits at day 14 of their first lactation, and treated with bromocriptine or vehicle (Waters et al., 1995). The ethical aspects of animal care complied with the relevant guidelines and licensing requirements laid down by the Ministère de l’Agriculture, France. Mammary tissues dissected free of connective and adipose tissues were cut into small fragments. For the preparation of enzymatically dissociated acini, mammary fragments were incubated for 90 minutes at 37°C in Hanks’ medium containing 200 UI/ml collagenase IV and 200 UI/ml hyaluronidase under an atmosphere of 95% O ₂/5% CO ₂, washed and then filtered through a strainer. Isolated cells were separated from the acini by three successive decantations for 15 minutes at 20°C in Hanks’ medium. Acini or mammary fragments were incubated in Hanks’ medium in the presence or absence of 5 μM BFA for 25 minutes at 37°C, and then incubated for 30 minutes at 20°C in the presence of 5 μg/ml bioPRL. They were then washed extensively and incubated in the presence or absence of 5 μM BFA, for 5, 15 and 60 minutes at 37°C. In order to label basolateral membranes of MEC, FITC-conjugated rabbit IgG (20 μg/ml) was added to the incubation medium 1 minute before the end of each incubation period.

Having previously established that 5 μM BFA dismantles the Golgi stacks in a few minutes, whereas preformed secretory vesicles are still detectable after 20 minutes, and that 50 μM BFA exerts the same effect by acting more rapidly (Pauloin et al., 1997), we used these two concentrations to evaluate the effect of BFA over various periods of time with regards to the intracellular transport of caseins.

Mammary fragments were also incubated similarly in Hanks’ medium in the presence or absence of 5 μM BFA, then incubated in the continuous presence of bioPRL at 37°C for 5 and 60 minutes.

Mammary fragments are used to measure casein secretion. Until now, it has not been possible to obtain in vitro, differentiated and polarised MEC in culture system where the apical medium is accessible. Consequently, incubated fragment system remains the more physiological system to study milk secretion. Fragments were first incubated for 30 minutes in Hanks’ medium, labelled for 3 minutes with 0.74 MBq/ml of L-[4,5−³H]leucine (1.7-3.1 TBq/mmol), and extensively rinsed in the same medium. Then, they were incubated at 37°C for the time periods indicated, in the same medium in the presence or absence of 5 μg/ml PRL and 5 μM BFA, added individually or together. Some fragments were incubated after the pulse in the presence or absence of 25 μM or 100 μM arachidonic acid and 5 μM BFA, added individually or together for the time periods indicated.

Labelled tissue proteins were assayed after TCA precipitation. The labelled newly synthesised caseins secreted into the medium were precipitated at pH 4.6 or directly analysed using SDS-polyacrylamide gel electrophoresis (SDS-PAGE) followed by fluorography (not shown). With these two procedures, the quantities of newly synthesised caseins expressed as a percentage of total protein incorporation in tissues are similar. The rate of secretion was expressed using methods previously described (Seddiki et al., 1991; Pauloin et al., 1997; Clegg et al., 1998).

Mammary fragments and acini were homogenised in 10 mM Hepes, 10 mM EDTA and 150 mM NaCl buffer supplemented with a mixture of protease inhibitors (5 μl/ml). Homogenisation was carried out in glass potters using a Heidolph blender at high speed. The supernatant was spun at 800 g for 10 minutes at 4°C after the addition of 1% IGEPAL CA-630. SDS-PAGE and electrotransfer were carried out as previously described (Lkhider et al., 1996). Nitrocellulose membranes were incubated with either HRP-conjugated streptavidin (1:300) in phosphate-buffered saline (PBS)/0.3% Tween/5% no fat milk for 45 minutes or with the anti oPRL monoclonal antibody 5G2 (1:800) in the same buffer for 1 hour 45 minutes, washed and then incubated with HRP-conjugated anti-mouse IgG (1:1000) for 30 minutes. Proteins were detected using ECL.

Mammary fragments from bromocriptine-treated rabbits subjected to different treatments were homogenised and solubilised as previously described (Waters et al., 1995). The solubilised fraction was immunoprecipitated by polyclonal anti-receptor antibody 46 (2 μl/mg protein in the supernatant) overnight at 4°C, followed by the addition of 50 μl protein G-sepharose 1:1 suspension in PBS for 1 hour at 4°C. The beads were washed and analysed using SDS-PAGE on 8% gels and electrotransfer onto a nitrocellulose membrane. The membrane was probed with the 4 G10 anti-phosphotyrosine monoclonal antibody and immune complexes were revealed by ECL after incubation with anti-mouse peroxidase antibody. Blots were stripped in Tris-HCl 50 mM (pH 6.8) SDS 2%, β-mercaptoethanol 0.1 M for 30 minutes at 56°C, rinsed in washing buffer and reblotted with the polyclonal anti-receptor antibody 46 (1:2000), followed by anti-goat peroxidase antibody. Immune complexes were revealed using ECL.

Mammary fragments from bromocriptine-treated rabbits subjected to different treatments, were frozen and then potterised in ice-cold buffer A (10 ml/g tissue): 10 mM Hepes (pH 7.7), 25 mM KCl, 2 mM EDTA, 0.5 mM EGTA, 0.3 M sucrose, supplemented with a mixture of protease and phosphatase inhibitors. IGEPA CA-630 was added to 0.2% final concentration and the homogenate was solubilised for 5 minutes at 4°C and then layered on a sucrose cushion made using the same buffer as A, except that it contained 1 M instead of 0.3 M sucrose. Following centrifugation at 700 g for 15 minutes, the supernatant was removed and then centrifuged for 60 minutes at 20,000 g to obtain cytosolic extracts. Nuclear extracts were prepared as previously described (Jolivet et al., 1996), and protein concentrations were determined using the BCA protein assay kit (Pierce). For shift reactions, 1 μg of nuclear or 6 μg of cytosolic extract were incubated with approximately 50 pg of end-labelled, double-stranded DNA αs1 casein probe (50,000 cpm) for 25 minutes at room temperature, in 10 μl reaction buffer containing 10 mM Hepes (pH 7.7), 50 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, 8% Ficoll and 1 μg polydI-dC as the nonspecific competitor. The probe used was a double-stranded oligonucleotide, representing the sequence −104 to −85 in αs1 casein promoter (5′-GAG AAT TCT TAG AAT TTA AA-3′) and it was end-labelled using T4 polynucleotide kinase. Complexes and free DNA were separated by electrophoresis on 6% non-denaturing polyacrylamide gels with 0.25×TBE electrophoresis buffer at 200 V. Gels were transferred to DE 81 paper (Whatman), dried and autoradiographed.

After 30 minutes of preincubation in Hanks’ medium, the fragments (50-70 mg/ml) were labelled for 45 minutes in the same medium, using 37 kBq/ml of [1−¹⁴C]arachidonic acid, specific activity 2.00 GBq/mmol washed extensively and incubated for 15 minutes in the presence or absence of 5 μM BFA. PRL (5 μg/ml) was then added to the incubation medium, and further incubated for periods of 15 seconds, 1 minute, 5 minutes and 15 minutes. These time points were chosen according to results previously described (Blachier et al., 1988; Ollivier-Bousquet et al., 1991). Lipid extraction and thin-layer chromatography were carried out as previously described (Blachier et al., 1988). The radioactivity of the spots corresponding to the different lipid classes (compared with standards) was measured with an Instant Imager (Software Packard). The radioactivity of the fatty acid was evaluated as a percentage of total radioactivity.

After incubation, acini and mammary explants were fixed and treated as previously described (Clegg et al., 1998). Acini were cytocentrifuged, fixed and permeabilised with 0.1% Triton X-100. The sections and the permeabilised acini were sequentially incubated with 50 mM NH4Cl, 1% bovine serum albumin (BSA), tetramethylrhodamine isothiocyanate-conjugated (TRITC) ExtrAvidin 1:300 in PBS/1% BSA and mounted in Vectashield medium.

Mammary fragments were fixed and treated as previously described (Clegg et al., 1998).

For each protein- and fatty acid-labelling assay carried out with explants from a single rabbit, controls and treatments were paired and replicated as indicated. The resulting pairs of values were compared using Student’s t-test for paired samples. Differences were considered significant at P<0.05.

Enzymatically dissociated acini were used to monitor transcytosis of bioPRL in the presence or absence of BFA.

These acini present two advantages. First, the spherical organisation, which is a prerequisite for MEC functioning, is preserved. Second, the whole acinus, cytocentrifuged, permeabilised and treated for immunocytochemical localisation of bioPRL, make it possible to visualise the entire cell volume contrary to sections which show only a thin slice.

After 30 minutes incubation at 20°C in the presence of bioPRL, labelled hormone was detected in the cells. Interestingly, hormone was detectable in long, tubulo-vesicular structures of the cells (Fig. 1A). In the presence of a 100-fold excess of ovine PRL most of the acini were not or were very faintly labelled (Fig. 1B), thus confirming the specificity of bioPRL binding to receptors. Acini were preincubated for 25 minutes at 37°C in the absence or presence of 5 μM BFA, then bioPRL was added to the same medium for a further 30 minutes incubation at 20°C. To determine the precise localisation of bioPRL in cells, FITC-IgG was used as a basolateral marker. Because binding of IgG appears to take place through a Fc receptor localised in the basolateral membrane (Larson, 1992), 20 μg/ml FITC-IgG was added to the incubation medium 1 minute before the end of each chase period. In these conditions FITC-IgG stained the basolateral membrane in green (Fig. 1C-F). After a 5 minute chase at 37°C, bioPRL labelling (red) was detected in control cells (not shown) and in cells incubated in the presence of BFA (Fig. 1D). BioPRL labelling was detectable in spots close to the basolateral membranes and in a nonpolarised tubulo-vesicular network. However, whereas in untreated acini, bioPRL accumulated in the apical region of the cells after 15 minutes chase (Fig. 1C), in BFA-treated acini the localisation of bioPRL labelling was not modified (not shown). After a 60 minute chase at 37°C, very little or no bioPRL labelling remained in untreated cells (Fig. 1E). In contrast in BFA-treated acini, bioPRL labelling remained accumulated in tubulo-vesicular structures, suggesting that bio-PRL was retained in these structures during the entire 60 minute chase period (Fig. 1F). In support of this observation, gold-labelled PRL, endocytosed and chased during 60 minutes in BFA-treated cells, was localised, by electron microscopy, in vesicular and tubular structures (not shown).

Proteins from acini were analysed using SDS-PAGE and blotting with either the anti-oPRL monoclonal antibody 5G2, or HRP-conjugated streptavidin, which binds to biotin (Fig. 2). In control acini, a major 23 kDa protein band was revealed by both probes after a 5 minute chase, suggesting that PRL was internalised in an intact molecular form. An intracellular intact molecular form of PRL was also detected after a 5 minute chase in BFA-treated acini (Fig. 2), attesting that a substantial amount of the hormone was internalised independent of the BFA presence. After a 60 minute chase, substantial PRL labelling was detected in BFA-treated compared with untreated acini extracts (Fig. 2). Thus, both, the localisation by fluorescence and the biochemical detection of PRL in acini showed that the early events (endocytosis) of PRL transport were not abolished by BFA treatment, whereas the late events (transcytosis) were impaired.

As in acini, bioPRL was detected in mammary fragments, after a 5 minute chase, as a major 23 kDa molecular form, irrespective of BFA treatment (not shown). Fluorescence detection of bioPRL on sections of mammary fragments incubated in the absence or presence of BFA for 25 minutes, followed by incubation with bioPRL for 5 and 60 minutes at 37°C, revealed that after 5 minutes of bioPRL incubation, labelling accumulated in the basal region of the cells, irrespective of BFA treatment (Fig. 3A,B). After 60 minutes of bioPRL incubation, labelling was detected in the apical region of untreated cells, whereas, in BFA-treated tissues, labelling accumulated in the cells (Fig. 3C,D). Thus, BFA strongly impaired the transcytosis of PRL in mammary fragments, as was the case in acini.

Newly synthesised caseins accumulate in vitro in the rough endoplasmic reticulum (RER) in less than 5 minutes, in the Golgi region in about 15 minutes and in secretory vesicles in between 25 and 45 minutes. They then gradually accumulate in the lumen and begin to appear in the incubation medium. Thus, after 60 minutes, substantial amounts of secreted protein accumulate in the medium, making it possible to evaluate the effect of BFA and PRL on protein secretion.

As previously shown (Ollivier-Bousquet, 1978; Seddiki et al., 1991; Ollivier-Bousquet et al., 1993; Ollivier-Bousquet et al., 1997; Pauloin et al., 1997), PRL added at the end of a 3 minute pulse, exerted a secretagogue effect (Fig. 4). Preincubation of mammary tissues with low concentrations of BFA (<50 nM) did not affect basal secretion but inhibited the secretagogue effect of PRL. Similar results were observed with the addition of micromolar concentrations of BFA at the end of a 3 minute pulse (Pauloin et al., 1997). Given the inhibitory effect of BFA on the secretagogue effect of PRL, we questioned whether BFA extended this by impairing the transmission of PRL signalling to stimulation of secretion. To examine this point, 5 μM BFA and PRL were added, respectively, 25 minutes and 45 minutes after the beginning of the pulse, when radioactive cargoes have already reached secretory vesicles. Addition of PRL alone, 45 minutes after the beginning of the pulse, increased casein secretion, as did PRL added immediately after the pulse. Addition of 5 μM BFA 20 minutes prior to PRL inhibited the PRL-stimulated secretion of radioactive caseins accumulated in the secretory vesicles (Fig. 4). However, in these experimental conditions, BFA alone slightly decreased the basal level of casein secretion, suggesting that general disorganisation of the exocytotic pathway might be responsible of the lack of PRL stimulatory effect. Three data reveal that this is not the case. First, addition of a higher concentration of BFA (50 μM) 35 minutes after the pulse, when radioactive cargoes are accumulated in secretory vesicles located close to the apical membrane, did not affect basal level of casein secretion but inhibited the secretagogue effect of PRL added 50 minutes after the pulse (Fig. 5). Second, the aspect of the cells after 10 minutes of preincubation in the presence of 5 μM BFA revealed the morphological features induced by the drug as previously described: the disappearance of the Golgi saccules and the presence of some tubulo-vesicular networks (Pauloin et al., 1997). However, small secretory vesicles containing casein micelles, the fusion of these vesicles with the apical membrane and the release of micelles via the exocytotic route were always detectable (Fig. 6A). Third, the effect of arachidonic acid on casein secretion in mammary cells incubated in the presence of BFA was investigated. This PUFA has previously been shown to mimic the effect of PRL on exocytosis (Ollivier-Bousquet, 1984). In agreement with these data, arachidonic acid exogenously added at the end of the pulse increased casein secretion (Fig. 6B, column 2). The same effect was obtained when arachidonic acid was added 45 minutes after the beginning of the pulse, consistent with an effect on late events in the secretory pathway (Fig. 6B, column 3). After 20 minutes preincubation with 5 μM BFA added 25 minutes after the beginning of the pulse, arachidonic acid increased casein secretion (compare Fig. 6B, column 4 with column 5). This increase is a function of arachidonic acid concentration (25 μM and 100 μM) and time of action (10 minutes and 30 minutes) (Fig. 6C). All together, these results show that arachidonic acid increases casein secretion by acting on late events of exocytosis and suggest that the final steps of the exocytotic pathway were not completely disorganised by BFA. As the effect of arachidonic acid does not require the integrity of the Golgi region, whereas the effect of PRL does (compare Fig. 4 with Fig. 6B), these results suggest that the inhibition by BFA of PRL-stimulated secretion is rather due to impairment of the transmission of PRL-mediated signals.

Overall, these results demonstrate that PRL, after binding to its receptor and endocytosis in the mammary cell, requires integrity of the Golgi apparatus to exert its secretagogue effect on caseins contained in compartments downstream of the Golgi apparatus.

In order to determine whether or not the effects of BFA were confined to the stimulation of protein secretion by PRL, we examined whether BFA affected the early (receptor phosphorylation) and late (Stat5 DNA-binding activity) events of PRL signalling to expression of the genes for milk proteins using fragments from bromocriptine-treated lactating rabbits stimulated by PRL in vitro.

In untreated fragments, peak PRL-induced receptor phosphorylation was observed at 5 minutes, and peak Stat5 activation at 15 minutes. The pre-treatment of mammary fragments with 5 μM BFA for 15 minutes did not significantly change PRL-induced phosphorylation of the receptors (Fig. 7A). Immunoblotting of the same blot with the anti-receptor antibody 46 showed that the amount of total receptor protein immunoprecipitated did not change upon BFA or PRL treatment during this period of time (Fig. 7A). Furthermore, such pre-treatment did not affect PRL-induced Stat5 DNA-binding activity in either cytosolic or nuclear extracts (Fig. 7B). Thus, BFA pre-treatment did not affect the early and late events of PRL signalling on the expression of milk protein genes.

An early in vitro event in response to PRL and in response to a phospholipase A₂ that mimics the secretagogue effect of PRL, is the rapid release of arachidonic acid, from membrane phospholipids, between 15 seconds and 5 minutes (Blachier et al., 1988; Ollivier-Bousquet et al., 1991). Metabolites of this fatty acid are involved in the secretagogue response of PRL (Blachier et al., 1988). As PRL was incapable of stimulating casein secretion in BFA-treated fragments, whereas arachidonic acid exerted this effect, we questioned whether BFA treatment affected the PRL-induced release of this fatty acid. In the absence of BFA, PRL induced the release of free arachidonic acid very early on. In contrast, in fragments preincubated with 5 μM BFA, the level of free arachidonic acid did not increase, even after 60 minutes incubation with PRL (Fig. 8). Thus, when the Golgi apparatus is disorganised by BFA treatment, PRL can no longer induce the release of arachidonic acid.

Disorganisation of the Golgi apparatus by BFA did not interfere with PRL-induced receptor phosphorylation and Stat5 activation, nor did it impair the first stage of internalisation of PRL. However, this disorganisation prevented transcytosis of PRL, PRL-induced arachidonic acid release and the subsequent stimulatory effect of PRL on the exocytosis of caseins contained in secretory vesicles. These results show, for the first time to our knowledge, that two signalling pathways of PRL in the lactating MEC can be dissociated. They reveal that PRL binding to its receptor and the subsequent phosphorylation of PRLR are insufficient to induce signalling to the secretagogue effect of PRL, and suggest that the emission of various PRL-mediated signals occurs in different cellular compartments.

Plasma-borne PRL is transcytosed across the lactating MEC and released in milk in an intact molecular form of 23 kDa and in cleaved molecular forms (Sinha, 1995). Thus, PRL is only partially degraded during its passage across the cell. Observations to date have revealed that PRL was endocytosed in a complex, tubulo-vesicular network that appears to be very similar to the continuous endosomal network described for the transport of transferrin and epidermal growth factor (Hopkins et al., 1990). However, the subsequent intracellular stages of transport seem to be very different. In the former system, ligands and their receptors were rapidly recycled from the endosomal compartment, either to the plasma membrane or to the lysosomes, through a single vacuolar intermediate, the multivesicular bodies (Futter et al., 1996). In lactating MEC, PRL (internalised in endosomes) escapes from the lysosomal pathway and enter the secretory pathway (Lkhider et al., 1996; Ollivier-Bousquet, 1998). Our results make it possible to highlight different possibilities for the transport of PRL from the endosomes to the secretory pathway. PRL transcytosis may use routes into and across the Golgi apparatus. In the presence of BFA, Golgi stacks disappeared and this route was blocked. Fusion between endosomes and Golgi stacks has already been suggested in immunoglobulin-secreting myeloma cells, where the transferrin receptor and transferrin from the endosomal pathway appear in stacked Golgi cisternae (Woods et al., 1989). In rat endocrine pancreatic B cells, membrane markers (ferritin-conjugated concanavalin A, lectin and cationised ferritin) were internalised in trans median and cis cisternae of the Golgi apparatus, and also in vesicles of the trans middle and cis side of Golgi stacks (Pavelka et al., 1998). Alternatively, endosomes carrying PRL may fuse with the trans Golgi network (TGN). Considerable traffic between the plasma membrane and the TGN is well illustrated by the delivery of TGN 38 (which is TGN specific) to the plasma membrane and its recycling through the endocytic compartment to the TGN (Reaves and Banting, 1994; Ghosh et al., 1998). BFA has been shown to cause fusion of the Golgi stacks and RER, while the TGN fuses and mixes with the endosomal tubule system (Lippincott-Schwartz et al., 1991; Wood et al., 1991). It could be hypothesised that tubules containing labelled PRL, which are detectable in the presence of BFA, might correspond to this mixed endosome-TGN tubular compartment. Finally, endosomes containing PRL may fuse directly with secretory vesicles. A direct fusion of endocytic vesicles labelled with membrane markers and mature secretory vesicles has been described in other cell types (Pavelka et al., 1998). At present it is impossible to choose between these three options.

Our results show that BFA does not alter PRL-induced tyrosine phosphorylation of the receptor and the activation in the cytosol and transport in the nucleus of cellular targets, such as Stat5, which lead to the expression of the genes for milk proteins. Thus, PRL signalling to the gene for casein is not dependent on Golgi integrity. STAT nuclear translocation mechanisms are not well understood, but it is possible that Stat5 forms transcriptionally active complexes with the glucorticoid receptor (GR). In support of this, Stat5 and GR are reported to have a mutual effect on nuclear translocation (Wyszomierski et al., 1999). Our results emphasise the fact that Stat5 nuclear translocation does not require transport by BFA-sensitive vesicles.

In contrast to its effect on PRL signalling to genes, BFA impairs the secretagogue effect of PRL. The addition of PRL at the end of a 3 minute pulse increases the basal level of casein secretion measured after a 60 minute chase period. Similarly, ovine PRL and recombinant bovine PRL exerts a stimulatory effect on release of newly synthesised triacylglycerol and proteins, in rat mammary gland slices (Da Costa et al, 1995). The one group not able to confirm the secretagogue effect of PRL on protein release in MEC failed to verify whether PRL receptors were always present on the plasma membrane in their assays (Burgogne and Wilde, 1994). These authors show results obtained with acini labelled for 1 hour, a delay too long to monitor an hormonal effect on a defined time window. Finally, it is obvious that the secretagogue response to PRL by lactating MEC requires an atmosphere containing 95% O2/5% CO2, conditions that are not clearly mentioned by these authors. In our study, PRL exerted its secretagogue effect when added 45 minutes or 50 minutes after the beginning of the pulse, demonstrating that the hormone exerts its effect on the late events in the exocytotic pathway. This effect is improved when PRL is added 50 minutes after the beginning of the pulse, suggesting that, at this time, radioactive proteins are accumulated in secretory vesicles close to the apical membrane, ready to be released into the lumen. It was also evidenced that exogenous added arachidonic acid similarly stimulated these late events, in agreement with the hypothesis that arachidonic acid, released after PRL stimulation, or its metabolites, mimics the effect of PRL on exocytosis. BFA impaired the release of arachidonic acid. The question arises whether the release of arachidonic acid occurred at the plasma membrane or if the fatty acid was released from intracellular membranes. The very rapid release of free arachidonic acid after binding of the hormone to its receptor might suggest that it occurred at the plasma membrane. However, if arachidonic acid was being released at the same cell site as receptor phosphorylation, BFA might have no effect on this event. This was not the case, and raises several questions.

1. A cytoplasmic signal emitted at the plasma membrane after the binding of PRL to its receptor may trigger the release of arachidonic acid from intracellular membranes. BFA may be ineffective on the emission of this signal, but may disorganise these target intracellular membranes which would then not be able to release arachidonic acid. In this case, PRL transcytosis and the emission of signals could be independent processes.

2. Another possibility is that a signal may be emitted intracellularly during transport of the PRL or PRL-PRLR complex. The fate of the PRL-PRLR complex during its transport across the MEC, and the intracellular site of dissociation between the hormone and the receptor remain speculative. Because cleaved molecular forms of the hormone have been described in mammary tissues (Lkhider et al., 1996), it can be postulated that endoproteolytic processing occurs in the intracellular sites where cleaving activities reside. The journey of PRL may be associated with cleavage, leading to the formation of molecular forms with specific biological properties. Whether these forms of PRL are able to interact with intracellular receptors and participate in the transduction of an intracellular hormonal message is an unanswered question. However, an internalised receptor, that is or is not associated with the hormone, may interact with a topographically distinct substrate and give rise to different biological actions. These hypotheses do not contradict the time needed for PRL or PRLR to reach a possible target in the Golgi region, as it has been shown that a membrane marker (cationised ferritin) is internalised within 60 seconds in Golgi stacks in lactating MEC (Ollivier-Bousquet, 1998). Transport into and across the Golgi apparatus appears to be essential for a number of drugs and toxins to exert their biological effects (Pelham et al., 1992). The intoxication of target cells by a number of toxic proteins such as cholera toxin, Shiga toxin, pertussis toxin or ricin require an intact Golgi region (Sandvig et al., 1991; Nambiar et al., 1993; Garred et al., 1995; el Baya et al., 1997). It has also been suggested that endocytic routes for polypeptide hormones in intracellular signal transduction participate in signalling (Bevan et al., 1996). In conclusion, these results demonstrate that in lactating MEC, two signalling pathways of PRL can be dissociated. They also suggest that a compartmentalisation of signalling may occur in the lactating MEC.

We are grateful to S. Delpal for electron microscopy, to Drs R. Boisgard, E. Chanat, F. Lavialle, A. Pauloin and D. Rainteau for helpful discussions, and to I. Blondeau for preparing the manuscript.