/05/$15.00/0 Molecular Endocrinology 19(5): Copyright 2005 by The Endocrine Society doi: /me

Transcription

1 /05/$15.00/0 Molecular Endocrinology 19(5): Printed in U.S.A. Copyright 2005 by The Endocrine Society doi: /me Unmodified Prolactin (PRL) and S179D PRL-Initiated Bioluminescence Resonance Energy Transfer between Homo- and Hetero-Pairs of Long and Short Human PRL Receptors in Living Human Cells Dunyong Tan, David A. Johnson, Wei Wu, Lingfang Zeng, Yen Hao Chen, Wen Y. Chen, Barbara K. Vonderhaar, and Ameae M. Walker Division of Biomedical Sciences (D.T., D.A.J., W.W., L.Z., Y.H.C., A.M.W.), University of California, Riverside, Riverside, California ; Mammary Biology and Tumorigenesis Laboratory (B.K.V.), National Cancer Institute, Bethesda, Maryland 20892; and Oncology Research Institute (W.Y.C.), Clemson University, Greenville, South Carolina We have used bioluminescence resonance energy transfer (BRET) to examine the interaction between human prolactins (PRLs) and the long (LF) and two short isoforms (SF1a and SF1b) of the human PRL receptor in living cells. cdna sequences encoding the LF, SF1a, and SF1b were subcloned into codon-humanized vectors containing cdnas for either Renilla reniformis luciferase (Rluc) or a green fluorescent protein (GFP 2 ) with a 12- or 13- amino acid linker connecting the parts of the fusion proteins. Transfection into human embryonic kidney 293 cells demonstrated maintained function of Rluc and GFP 2 when linked to the receptors, and confocal microscopy demonstrated the localization of tagged receptors in the plasma membrane by 48 h after transfection. All three tagged receptors transduced a signal, with the LF and SF1a stimulating, and SF1b inhibiting, promoter activity PRLRs [PROLACTIN (PRL) RECEPTORS] are members of the class I cytokine superfamily and, as such, share a number of structural features with more classical cytokine receptors (1 3). The structural features of this superfamily include: 1) all are single transmembrane domain receptors that associate with effector kinases rather than having intrinsic kinase activity and 2) homo- or hetero-receptor isoform multimerization is thought to be initiated by ligand binding, which in turn leads to transphosphorylation of the First Published Online February 3, 2005 Abbreviations: B/F, Bioluminescence/fluorescence; BRET, bioluminescence resonance energy transfer; DPBS, Dulbecco s PBS; ECD, extracellular domain; GFP, green fluorescent protein; HEK293, human embryonic kidney 293 cells; LF, long isoform of the human PRLR in living cells; PRL, prolactin; PRLR, PRL receptor; Rluc, Renilla reniformis luciferase; SF1a and SF1b, short isoforms of the human PRLR in living cells; Stat, of signal transducer and activator of transcription; U- PRL, unmodified PRL. Molecular Endocrinology is published monthly by The Endocrine Society ( the foremost professional society serving the endocrine community. of an approximately 2.4-kb -casein-luc construct. Both unmodified PRL (U-PRL) and the molecular mimic of phosphorylated PRL, S179D PRL, induced BRET with all combinations of long and short receptor isoforms except SF1a plus SF1b. No BRET was observed with the site twoinactive mutant, G129R PRL. This is the first demonstration, 1) that species homologous PRL promotes both homo- and hetero-interaction of most long and short PRLR pairs in living cells, 2) that both U-PRL and S179D PRL are active in this regard, and 3) that there is some aspect of SF1a- SF1b structure that prevents this particular hetero-receptor pairing. In addition, we conclude that preferential pairing of different receptor isoforms is not the explanation for the different signaling initiated by U-PRL and S179D PRL. (Molecular Endocrinology 19: , 2005) associated kinase and initiation of several intracellular signaling cascades (4, 5). For the PRLR, dimerization is considered sufficient to initiate signaling (3). This model of dimerization activation of the PRLR was originally developed by analogy to closely related hormones and is supported by studies showing the formation of a ternary complex between a related hormone, ovine placental lactogen, and two copies of the extracellular domain (ECD) of the rat PRLR (6), functional analysis of mutations in PRL (7), solution analysis of species heterologous PRL-PRLR-ECD interactions (8), ligand-receptor cross-linking studies (9) and, most recently, the demonstration of placental lactogen-induced heterodimerization of PRLRs with GH receptors using fluorescence resonance energy transfer (10). To date, however, there is no evidence for species-homologous PRL-induced PRLR dimerization because all attempts to demonstrate this have failed. The ability of species heterologous, but not homologous PRL, to cause measurable dimerization has been attributed to a more stable interaction with the second receptor (11). Thus, species heterologous interactions may not be representative of species ho- 1291

2 1292 Mol Endocrinol, May 2005, 19(5): Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms mologous ones. In this regard, a number of mutations, which have no influence on biological effect generated in heterologous ligand-receptor interactions, have been demonstrated to have marked effects in homologous ligand-receptor interactions (12). Likewise, biological activity in a heterologous system can be different from that in a homologous system (12, 13). From the work published to date, therefore, it remains possible that dimerization is an artifact of either the use of species heterologous PRL or the use of cross-linking and immunoprecipitation protocols. Alternatively spliced PRLRs, possessing different cytoplasmic tails, have been recognized in rodent species for some time (11, 14) but have only recently been cloned from normal human tissue. In the case of human PRLR, two short isoforms have been identified (15, 16), which are in addition to a long form and an intermediate form cloned from human breast cancer cells (17, 18). The designations as long, intermediate, and short isoforms are a little confusing when referring to the human receptors, because one of the short forms is actually longer than the intermediate form. Thus, the short 1a (SF1a) isoform is 352 amino acids, the short 1b (SF1b) isoform is 264 amino acids, the intermediate form is 325 amino acids, and the long isoform (LF) is 598 amino acids (15, 16). In addition to the question of whether hormone-induced dimerization occurs with species homologous PRL and PRLRs, it is unclear whether human PRL induces pairing of all combinations of PRLR isoforms, i.e. homo- and hetero-pairs of receptors. Also unclear is whether both unmodified PRL (U-PRL) and a molecular mimic of phosphorylated PRL (S179D PRL) cause the same homo- and hetero-pair interactions because these two forms of human PRL initiate different intracellular signals (19). To address these issues, we assessed the ability of human PRL forms to induce a bioluminescence resonance energy transfer (BRET) signal using various combinations of the LF and SF1a and SF1b isoforms of human PRLRs. Chimers of Renilla reniformis luciferase (Rluc) or a variant of green fluorescent protein (GFP 2 ) linked to the LF, SF1a, and SF1b isoforms of the PRLR were engineered. We found species homologous PRL-dependent BRET as a result of homo- and hetero-pairing of all receptor isoform pairs except the SF1a-SF1b hetero-pair in living cells. This occurred with both U-PRL and S179D PRL, indicating that pairing of specific receptors is not what produces the different signaling activities of these two forms of the hormone. Additionally, no pairing occurred with the mutant PRL, G129R, engineered to eliminate binding to the second PRLR (20), demonstrating that each PRL molecule must interact with two receptors to induce BRET. RESULTS Biological Characterization of Rluc- and GFP 2 - Tagged PRLRs in Response to U-PRL To assess the ability of the cells transfected with the Rluc- and GFP 2 -tagged receptors (Fig. 1) to produce a PRL-induced intracellular signal, U-PRL-dependent activation of the -casein gene promoter was measured with a -casein promoter-luciferase assay. Figure2(upper panel), shows that the -casein luciferase activity was significantly (P 0.05) increased in cells Fig. 1. Vector Map for Rluc and GFP 2 Receptor Constructs cdna sequences encoding LF, SF1a, and SF1b were PCR-amplified from original plasmid, PEF6-LF, PEF4-SF1a, and pcr2.1 1b, respectively, by using forward and reverse primers harboring unique MluI and KpnI sites and subcloned into a codon-humanized expression vector containing cdnas for either the bioluminescent Rluc protein or GFP 2 such that a 12- or 13-amino acid linker would occur between the two parts of the resultant fusion proteins. The amino acid sequence of the linker in the GFP 2 - and Rluc-tagged receptor are Trp-Tyr-Arg-Gly-Pro-Gly-Ile-Pro-Pro-Val-Ala-Thr and Trp-Tyr-Arg-Gly-Pro-Gly-Ile- Pro-Pro-Ala-Arg-Ala-Thr, respectively. The sizes were 6171 bp for LF-GFP 2, 5443 bp for SF1a-GFP 2, 5178 bp for SF1b-GFP 2, 6810 bp for LF-Rluc, 6082 bp for SF1a-Rluc, and 5817 bp for SF1b-Rluc. SV40, Simian virus 40; Zeo, zeocin; Kan, kanamycin; amp, ampicillin; cmv, cytomegalovirus; TK, thymidine kinase; Mlu I, restriction enzyme from Micrococcus luteus; Kpn I, restriction enzyme from Klebsiella pneumoniae.

3 Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Mol Endocrinol, May 2005, 19(5): Fig. 2. Ability of the GFP 2 - and Rluc-Tagged Receptors to Activate -Casein Promoter in Response to U-PRL HEK 293 cells were cotransfected with fusion constructs (GFP 2 -PRLRs or prluc-prlrs) and -casein-luc, which contains a -casein gene promoter and a luciferase coding cdna sequence. Twenty-four hours after transfection, the medium was changed to DMEM without serum containing 45.7 nm (final concentration) of U-PRL. After a further 24 h, the cells were washed with DPBS three times, lysed with the reporter lysis buffer and the luciferase activity was measured using a Monolight 2010 Luminometer (Analytical Luminescence). For PRLR-GFP 2 constructs (upper panel), the LF cdna was the same with and without any SF cdna. Total DNA was kept constant by using empty vector. Data are the mean of three to six determinations SEM. *, Different from control with P 0.05; **, different from control with P 0.01; ##, different from LF-GFP 2 or LF-Rluc with P RLU, Relative luminescence unit. Data presented are relative -casein promoter activity in the absence vs. presence of the receptor isoform indicated. All cultures were treated with U-PRL as indicated above. cotransfected with the construct combinations of LF- GFP 2 / -casein-luciferase and SF1a-GFP 2 / -caseinluciferase. The relative luciferase activity units were increased from in the control to and , respectively, demonstrating that receptors tagged with GFP 2 are capable of transducing a signal and are therefore functional. In the cells cotransfected with SF1b-GFP 2 and -caseinluciferase, the relative luciferase activity units were decreased from the control value to , indicating, in agreement with our previously published observation (15), that SF1b acts as a negative regulator of baseline -casein activity. Similar results were observed in the cells cotransfected with LF-Rluc and -casein-luciferase, SF1a-Rluc and -casein-luciferase and SF1b-Rluc and -casein-luciferase (Fig. 2, lower panel). The ability of the SF1a form to increase promoter activity was not due to the attachment of Rluc or GFP because it also occurred with receptors without attached fusion proteins (data not shown). There is no way to accurately determine the activity of hetero-pairs because transfection with two forms of the receptor does not guarantee that the observed signal derives from heterodimerization. Instead, it could arise from the mean of multiple homodimerization and heterodimerization events. Bearing in mind this caveat, however, one can observe a reduced signal to -casein luciferase in cells cotransfected with the LF-GFP 2 and either SF1a-GFP 2 or SF1b-GFP 2 vs. those transfected with only the LF-GFP 2 (Fig. 2, upper panel). This occurs for the LF-SF1a pair despite the fact that the SF1a-GFP 2 form alone signals weakly to -casein. Cotransfection of SF1a-GFP 2 and SF1b- GFP 2 produced no increase in -casein promoter activity in response to U-PRL (data not shown). Inclusion of a control vector normalized for total DNA transfected. This vector coded for a nonspecific protein and therefore also controlled for levels of competing protein expression. Cellular Expression of the GFP 2 and Rluc- Tagged PRLR The cells were cotransfected with equimolar amounts of cdna in various combinations of GFP 2 - and Rluctagged receptors. A GFP 2 -Rluc fusion construct, in which the GFP 2 protein was directly linked to the Rluc protein, once expressed, was used as a control from which a reference bioluminescence/fluorescence (B/F) ratio, indicative of equimolar expression of GFP 2 and Rluc, was derived. For all experiments shown in Fig. 3, the longer-lived and more economical substrate of Rluc, coelenterazine h, was used for analyses. Luminescence peaks at 475 with this substrate. Coelenterazine h is suitable for this analysis, but because of its spectral properties is less suitable for the BRET analyses presented (see Figs. 6 8 and 10) for which the DeepBlueC (PerkinElmer, Wellesley, MA) substrate was used. It should also be noted that the spectral properties of GFP 2, with excitation at 405 nm and peak emission at nm, are different from those of the more commonly used EGFP. As shown in Fig. 3, both the fluorescence (Fig. 3A) and bioluminescence (Fig. 3B) intensities of GFP 2 - and Rluc-tagged receptors were lower than that of GFP 2 -Rluc, indicating a higher expression level of the cytoplasmically expressed control construct relative to any of the PRLR receptors. The peak fluorescence intensities for GFP 2 - Rluc, LF-GFP 2 /LF-Rluc, SF1a-GFP 2 /SF1a-Rluc, and SF1b-GFP 2 /SF1b-Rluc transfected cells were , , and , respectively, and the bioluminescence peak intensities were , , , and , respectively, resulting in similar ratios of B/F for all the construct

4 1294 Mol Endocrinol, May 2005, 19(5): Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Fig. 3. Fluorescence and Bioluminescence Measurements with the Constructs HEK293 cells were transfected with equal moles (0.48 pmol) of GFP 2 -Rluc, LF-GFP 2 /LF-Rluc, SF1a-GFP 2 /SF1a-Rluc, and SF1b-GFP 2 / SF1b-Rluc, respectively. The spectra were examined within 48 h after the transfection. Samples were excited at 405 nm. For bioluminescence measurements, coelenterazine h was present at a concentration of 5 M. A, Fluorescence; B, bioluminescence; C, luminescence/fluorescence (L/F) ratio. pairs. These were ,8 10 4, and , respectively (Fig. 3C), strongly suggesting that equimolar expression of donor and acceptor was achieved by equimolar concentrations of plasmid DNA used for the transfections. Different receptors, however, were expressed at different levels. The SF1a isoform, whether tagged with Rluc or GFP, was expressed at lower levels than either the SF1b or LF isoforms. This is in agreement with data from the Dufau lab (16), which also showed a decreased expression of the SF1a and is likely therefore a function of the receptor per se and not just the fusion protein. It is also important to note that the spectral properties of luciferase or GFP were not altered by their linkage to the different receptors. Subcellular localization of LF-GFP 2 (Fig. 4A), SF1a-GFP 2 (Fig. 4B) and SF1b-GFP 2 (Fig. 4C) was investigated. Panels A C of Fig. 4 show single confocal images from 48 h after transfection. Panel D shows an autofluorescence control. At this time point, the confocal imaging showed that a substantial number of PRLRs were present in the region of the plasma membrane (white arrows), although there was, of course, intracellular staining consistent with synthesis in the rough endoplasmic reticulum and movement through the Golgi. Plasma membrane localization of the receptors was confirmed by the production of a BRET signal in response to added PRL (see below). Large quantities of receptor were sequestered in a juxtanuclear compartment. Sequestration of receptors within a juxtanuclear compartment, such as the Golgi apparatus, is typical of a wide variety of cells because receptors mature through the Golgi and are recycled through the Golgi (21 23). Substantial Golgi localization is also typical in cells overexpressing receptors (24). To be sure that the levels of expression of the constructs used in this study did not produce a nonspecific BRET signal due to overexpression and a resultant decrease in the average distance between expressed proteins, levels of transfecting DNA of half and double those used to measure BRET (see below) were examined. We reasoned that if nonspecific BRET occurs, increasing concentrations of transfecting DNA should increase BRET. Cells were therefore cotransfected with GFP 2 and Rluc plasmids at either half (2 g/well) or double (8 g/well) the amount used for the subsequent BRET experiments. Figure 5 shows that, although the bioluminescence from the Rluc increases with the concentration of transfecting DNA, no detectable GFP emission and therefore no BRET was observed, suggesting that the concentrations of plasmid we were using for the GFP 2 - and Rluc-tagged PRLRs should not produce nonspecific BRET. This does not mean that further increases in transfecting DNA would not have produced ligand-independent BRET. The GFP 2 -Rluc fusion construct was used as a positive control to indicate where a BRET signal would have been observed, had one occurred.

5 Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Mol Endocrinol, May 2005, 19(5): Fig. 4. Cellular Localization of the GFP 2 -Tagged Receptors HEK 293 cells, cultured on a coverslip ( 12 mm), were transfected as described in Materials and Methods. Forty-eight hours after transfection, the cellular localization of the expressed fusion proteins of LF-GFP 2 (A), SF1a-GFP 2 (B), and SF1b-GFP 2 (C) were examined with a confocal fluorescence microscope (Zeiss 510) using filter sets designed for regular GFP. This was not optimal but was sufficient to demonstrate subcellular localization. D, Mock transfected control with no measurable fluorescence. White arrows, Fluorescence in the region of the plasma membrane. Bars in all panels, 10 m. Spectral Properties of the Rluc- and GFP 2 - Tagged PRLR Constructs The addition of DeepBlueC (5 M) to all the Rluctransfected cells was associated with a bioluminescence emission peak at 395 nm, a value identical with what has been reported for the untagged Rluc (PerkinElmer Application notes). The fluorescence excitation and emission maxima of all the GFP 2 -tagged PRLRs were also the same at 405 nm and 515 nm, respectively (data not shown). Unmodified PRL Induces BRET between Homo- Pairs of PRLRs Shown in Fig. 6 are a set of representative bioluminescence scans of suspensions of LF-Rluc and LF- GFP 2 cotransfected and nontransfected human embryonic kidney 293 (HEK293) cells. In the absence of U-PRL, primarily Rluc emission (peak 395 nm) was observed with barely detectable amounts of GFP 2 emission ( 515 nm; BRET) (Fig. 6A). Addition of U- PRL, however, was associated with a large increase in GFP 2 emission (Fig. 6B), indicative of a large increase in BRET and therefore a reduction in the distance between donor and acceptor, each attached to the cytoplasmic tails of paired receptors. Figure 6 also illustrates the averaged effects of U-PRL on the BRET ratio of cells transfected with RLuc and GFP 2 tagged to the same receptor isoforms (homo-pairs): LF (Fig. 6C), SF1a (Fig. 6D), and SF1b (Fig. 6E). PRL increased the BRET ratio of LF, SF1a, and SF1b homo-pairs by 9-, 4-, and 4-fold respectively. Unmodified PRL Induced Hetero-Pairing of PRLRs To assess the ability of U-PRL to induce pairing of hetero-receptor isoforms, HEK293 cells were cotransfected with equal amounts of LF-GFP 2 plus SF1a- Rluc, LF-GFP 2 plus SF1b-Rluc, or SF1a-GFP 2 plus SF1b-Rluc plasmids. As illustrated in Fig. 7, U-PRL increased the BRET ratios of cells cotransfected with SF1a-Rluc plus LF-GFP 2 (11-fold) and SF1b-Rluc plus LF-GFP 2 (5-fold) plasmids, but no significant BRET was observed with SF1b-Rluc plus SF1a-GFP 2. Similar results were also observed in the reverse combinations of the constructs including LF-Rluc plus SF1a- GFP 2, LF-Rluc plus SF1b-GFP 2, and SF1a-Rluc plus SF1b-GFP 2, showing that it made no difference which receptors were used as energy donors or acceptors (data not shown).

6 1296 Mol Endocrinol, May 2005, 19(5): Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Fig. 5. Overexpression of the Constructs Does Not Result in a BRET Signal from Random Collisions Cells were cotransfected with GFP 2 -N1 and Rluc-N1 plasmids at either half (2 g/well) or double (8 g/well) the usual amount for each construct. The bioluminescence and BRET 2 signal were examined within 48 h after the transfection in the presence of 5 M of DeepBlueC. S179D PRL and G129R PRL Effects on Homoand Hetero-Pairing of PRLRs and Activation of the -Casein Promoter When similar experiments were conducted with the molecular mimic of phosphorylated PRL, S179D PRL, the same results were obtained. Figure 8A illustrates homo-pairing and Fig. 8B, hetero-pairing. S179D PRL increased the BRET ratio of LF, SF1a and SF1b homopairs by 8-, 3-, and 3-fold, respectively. S179D PRL, like U-PRL, induced hetero-pairing between LF-SF1a and LF-SF1b, but not between SF1a and SF1b. The S179D PRL-induced increases in BRET were 3-fold for LF-SF1a and 5-fold for LF-SF1b. The only significant difference between the effects of U-PRL and S179D PRL was that S179D PRL was less efficient at producing BRET with the LF-SF1a hetero-pair (3-fold vs. 11- fold for U-PRL). As was true for U-PRL, interaction of S179D PRL with both LF and SF1a homo-pairs, and LF-SF1a hetero-pairs stimulated -casein promoter activity (Fig. 9). S179D PRL was an equivalent stimulator with LF homo-pairs, and a better stimulator of promoter activity with SF1a homo-pairs, than U-PRL (compare the fold control in Figs. 2 and 9). In contrast to U-PRL, there was no measurable dominant-negative activity when LF-SF1a hetero-pairs were engaged by S179D PRL. Like U-PRL, however, S179D PRL still caused a dominant-negative effect with LF-SF1b. No effect of S179D PRL on promoter activity was seen with SF1b homo-pairs or SF1a-SF1b hetero-pairs (data for the latter not presented). G129R PRL is a mutant human PRL, engineered to eliminate binding to the second PRLR (20). This form of human PRL was unable to induce BRET with any homo- or hetero-receptor pairs (data mostly not shown). Data for the LF homo-pair are shown in Fig. 10. Figure 10 also shows that G129R PRL competed in a dose-related fashion with U-PRL for the generation of a BRET signal by LF homo-pairs of the receptor. This result is not due just to increasing concentrations of PRL and the undoing of dimers because in other experiments 90 nm U-PRL produced an equivalent BRET signal to that seen with 45 nm U-PRL (data not shown). G129R PRL also failed to increase -casein promoter activity with any form of the receptor (Fig. 9). DISCUSSION PRL is usually described as inducing PRLR dimerization through a sequential process that includes initial high-affinity binding to one receptor followed by lower affinity binding to a second, recruited receptor (25). The lower affinity interaction with the second receptor has the potential to be undone by excess ligand because higher affinity 1:1 ligand-receptor interactions would prevail in a competitive process. Because receptor dimerization has been linked to one of the main signaling pathways, excess ligand would therefore be predicted to result in at least some decreased biological responses, a prediction in keeping with the longrecognized biphasic dose response curves observed for PRL (26, 27). The lower affinity binding of the second receptor to the PRL-PRLR binary complex has prevented any demonstration of PRL-induced dimerization with homologous PRL and PRLRs because the lower affinity interaction easily dissociates during sample processing. Heterologous PRL-PRLR interactions, on the other hand, have been shown to have a higher affinity for the second receptor (11), and this has allowed cross-linking and immunoprecipitation studies to produce some evidence of dimerization. This phenomenon has also made it possible to demonstrate a 1:2 ratio between PRL and the ECD of PRLRs using solution analysis or surface plasmon resonance (8). What is missing, however, is evidence of effective pairing of PRLRs with the biologically relevant ligand in living cells in real time.

7 Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Mol Endocrinol, May 2005, 19(5): Fig. 6. U-PRL-Induced Dimerization of the Homo-Receptor Pairs A, Bioluminescence from the DeepBlueC reaction without U-PRL; B, bioluminescence after the addition of U-PRL (45.7 nm). Note the absence of a BRET signal in the absence of U-PRL. C (LF), D (SF1a), and E (SF1b) show the averaged results from five to nine separate determinations on different preparations of cells. The BRET ratio was calculated as (Emission nm Background nm )/(Emission nm Background nm ). *, P 0.05 for difference with and without PRL. In the current study, we used BRET to examine the rapid interaction of homologous PRL with cell surface receptors in living cells. We observed very little BRET in the absence of ligand and increased BRET in response to ligand, thereby demonstrating that the BRET reflected events at the plasma membrane. The technique, however, cannot distinguish between the ability of a ligand to recruit monomers into dimers and a conformational change in the cytoplasmic regions of preexisting dimers. Either way, however, the technique measures close approximation (within 100 Å) of the signal transducing portions of the receptor and hence an important interaction. Our goal in applying BRET was to determine whether species homologous ligands could produce such interactions. Also, we have examined the ability of three specific forms of human PRL to cause interactions between the long and recently cloned short receptors, both as homoand heterodimers. To be sure that the different tagged PRLRs had biological activity, we examined -casein expression in response to U-PRL. Activation of the casein promoter has only been previously reported for the LF of the human PRLR; previous studies have failed to demonstrate activation of the casein promoter by either SF1a or SF1b after transient transfection of Chinese hamster ovary (15), COS-1 and HEK293 cells (16). However, we observed activation of -casein by both the LF and, to a lesser extent, the SF1a isoform of the receptor. By contrast, the SF1b isoform reduced casein expression below that seen in the controls, consistent with a previously reported dominant-negative activity (15, 16). The reason for the difference in activity of the SF1a reported herein does not reflect its attachment to the tags because untagged versions were similarly active. Rather, it reflects the use of the approximately 2.4 kb construct because there is no activity with a shorter 344 to 1 region of the pro-

8 1298 Mol Endocrinol, May 2005, 19(5): Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Fig. 7. U-PRL-Induced Heterodimerization of PRLRs Transfections were carried out as described in Materials and Methods and BRET 2 assays were performed within 48 h after transfection in the presence of DeepBlueC (DBC; 5 M) with or without U-PRL (45.7 nm). **, P 0.01 vs. for difference with and without PRL. moter (data not shown). In this regard, Hu et al. (16) did not define the size of the -casein promoter used, and their experiments were conducted with a different receptor construct isolated in their laboratory. In addition, Hu et al. (16) report that the amino acid sequence of the exon 11 portion of their SF1a results in more rapid turnover of the receptor. In similar studies, using Chinese hamster ovary cells and the construct isolated in our laboratory, conducted in the presence of ligand, we have been unable to demonstrate any significant difference in half-life of the LF and SF1a isoforms (Long, G., and B. K. Vonderhaar, unpublished). A significant basal activity of the -casein promoter in the control suggests a significant degree of signal transducer and activator of transcription (Stat) 5 activation in the absence of PRL and the PRLR in these cells, a phenomenon that was documented by immunoprecipitation and Western blot analyses (data not shown). Thus, SF1a may increase activity at this longer promoter by activating additional signaling cascades that can contribute to -casein expression (19, 28). Although SF1a at 352 amino acids is the longer of the SFs, current consensus, based on PRLRs in other species, would suggest that it is unable to efficiently activate Stat 5 (18, 29, 31). Whatever the details of signal transduction from the SF1a form turn out to be, however, the important issue for the current study is that all tagged forms of the receptor are able to transduce a signal, as reflected either positively or negatively in -casein promoter activity. Analysis of the hetero-pairs for biological activity is problematic because one cannot be sure that the activity derives from hetero-pairing. Instead, it could be the average result of signals generated from two sets of homo-pairs and some hetero-pairs. That said, however, it appears that cotransfection of either SF with the LF reduces the signal to -casein as compared with the LF alone when U-PRL is used as the ligand. This dominant-negative activity of the SFs for -casein expression has been reported previously (15, 16). Because the same amount of LF DNA was transfected in Fig. 8. S179D PRL-Induced Homo- and Heterodimerization of PRLRs Transfections were carried out as described in Materials and Methods and BRET 2 assays were performed 48 h after transfection in the presence of DeepBlueC (DBC; 5 M) with or without S179D PRL (45.7 nm). *, P 0.05; **, P 0.01 the absence and presence of SF1a and SF1a transduces a signal to -casein, one might have expected the LF-SF1a combination to have stimulated more promoter activity than LF alone, but this was not the case. Whether the dominant-negative activity of the SFs is caused by hetero-pairing, the generation of alternate signals or interference with cell surface display or synthesis of the more active LF remains to be established. This dominant-negative activity toward Fig. 9. Activation of the -Casein Promoter by S179D PRL and G129R PRL with All Homo-Pairs and BRET-Producing Hetero-Pairs of the Receptor This experiment was conducted as described in the legend for Fig. 2 using S179D PRL and G129R PRL at 45.7 nm. **, P 0.01 compared with control; ##, P 0.01 compared with LF.

9 Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Mol Endocrinol, May 2005, 19(5): Fig. 10. Competition between G129R PRL and U-PRL for the Generation of BRET at the LF Receptor Cells transfected with the LF were exposed to either 45.7 nm U-PRL (concentration used for all previous analyses), 45.7 nm G129R PRL, or 45.7 nm U-PRL plus increasing concentrations of G129R PRL between 4.57 and 45.7 nm and BRET produced was measured. For the sake of clarity on the figure, 45.7 nm was written as 45 nm and all proportions of 45.7 nm were similarly represented. *, P 0.05 compared with G129R PRL alone; #, P 0.05 compared with U-PRL alone. -casein promoter activity with LF-SF1a was not present when S179D PRL was used as the ligand, a result that may be related to the reduced BRET seen with S179D PRL and this hetero-pairing vs. that seen with U-PRL. In our experiments, U-PRL significantly increased the BRET ratio with all the homo-pairs of the tagged receptors and most of the hetero-pairs. Only receptors at the cell surface contributed to the BRET signal because there was very little signal in the absence of ligand and measurements were taken immediately after addition of ligand. The combined SF1a and SF1b hetero-pair, however, did not produce a BRET signal. This implies that U-PRL has a much greater capacity to cause interactions between homopairs of either SF1a or SF1b isoforms than it does between SF1a-SF1b hetero-pairs. Because the methodology cannot distinguish between actual dimerization and conformational change in the cytoplasmic regions of preexisting dimers, we cannot say that SF1a-SF1b dimers don t exist, but we can say that the signal transducing parts of the receptor remain further apart than 100 Å upon addition of PRL. One mechanism by which added PRL could distinguish between SF homo- and heterodimers would be by different conformations of their respective ECDs. The ECD of each of the receptors has the same amino acid sequence, but a change in conformation could be the result of attachment to the very different cytoplasmic domains. The ability of the intracellular domain of the PRLR to influence the conformation of the extracellular domain has been previously described (32 34). One study, which examined the effect of deletion of 55 of 57 of the amino acids in the intracellular domain of the long receptor, demonstrated that this increased the affinity of the receptor when this was examined by the binding of ovine PRL to microsomes derived from transiently transfected cells (32). Another study demonstrated that the PRL-binding protein present in rabbit milk, which is probably equivalent to the ECD of the receptor (15, 33), interacted differently with a monoclonal antibody when compared with the membrane receptor (34). This suggested that the membrane-embedded and free ECD had different conformations. Finally, a mutant form of the rat receptor with a reduced cytoplasmic tail has been demonstrated to have a greater affinity for ligand than the full-length version (35). For each of these examples, the amino acid sequence of the extracellular domain of the full-length and foreshortened version was the same. Thus, it appears extremely likely that the length and conformation of the intracellular domain affects the conformation of the ECD, and hence that SF homodimers and SF heterodimers can have different conformations such that the ligand cannot interact properly with the SF hetero-pair. When S179D PRL was used as the ligand, we found it to have the same ability to homo-pair all receptor forms and to heteropair all but the SF1a-SF1b pair. Thus, with two different ligands, this incompatibility is evident. S179D PRL has been shown to act as an antagonist to U-PRL-mediated cell proliferation and associated Stat 5 activation (36) but has also been shown to act as an agonist for -casein expression [Ref. 19 and data herein (Fig. 9)]. Before the current study, the easiest way to explain this would have been if antagonistic signals were brought about by blockade of specific receptor pairing, and agonistic properties, by preferential pairing of other receptors. Our BRET results, however, indicate that S179D PRL interacts with all of the same receptor pairs as U-PRL and so this is not what accounts for the different effects of the two ligands. Our current working model involves the generation of different receptor conformations with each ligand, resulting in inhibition and activation of different sets of signaling molecules. Indeed, the generation of different signals by S179D PRL and U-PRL has been demonstrated (19, 36). The idea that different receptor conformations would result in different signaling is supported by an elegant study analyzing the effect of stable one amino acid turns in the transmembrane helix of the erythropoietin receptor (37). In this study, one conformation produced full Jak-Stat and ERK 1/2 signaling, whereas another favored ERK 1/2 signaling. In other words, one conformation produced signals like U-PRL and another, signals like S179D PRL (19). A mechanistic cartoon of differential signaling in response to U-PRL and S179D PRL can be found in a recent review (38). If there are different conformations produced by U-PRL and S179D PRL, however, they are not large enough to see absolute differences in BRET generated by each ligand. This suggests that approximation of the cy-

10 1300 Mol Endocrinol, May 2005, 19(5): Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms toplasmic domains of two receptors is important for all signaling. Some aspects of the different conformations, however, may be reflected in the degree of BRET. The only significant BRET difference between the two ligands uncovered in the present study was that S179D PRL was about one third as efficient as U-PRL at creating a BRET signal from LF-SF1a hetero-pairs and, unlike U-PRL, did not show a dominant-negative effect with this hetero-pairing. It is possible therefore that the reduced BRET generated by the interaction of S179D PRL with LF-SF1a hetero-pairs reflects an altered conformation vs. that generated by U-PRL. G129R PRL was designed to substantially lower the affinity of binding to the second PRL receptor to block ternary complex formation (20) and, as such, would be expected to be inactive in BRET assays. This was confirmed in the present study and is a useful control indicating that the BRET signal required interaction of the ligand with both receptors in a pair and not just a conformational change in one of the two receptors. When titrated against a constant amount of U-PRL, G129R PRL competed with U-PRL such that a reduced BRET signal was produced. This demonstrated that G129R PRL, by occupying some receptors, interfered with the ability of U-PRL to form as many receptor pairs. The only surprising aspect of this experiment was the inhibitory potency of G129R PRL when analyzed in this manner because other work has suggested that more than a 1:1 ratio of G129R PRL to U-PRL is necessary to see significant antagonism (39). Because activation of the -casein promoter only reflects some of the signals generated by PRL at the LF and very little is yet known about signaling from either SF1a or SF1b, the current study cannot address the question of whether dimers can ever be formed without generation of a signal. What can be said, however, is that any signal generated by dimerization of the SF1b form does not lead to activation of the casein promoter. For the LF, where we know that signaling results in activation of the -casein promoter, we can also say that those forms of PRL that produce BRET, also activate the promoter, whereas the one that does not produce BRET, does not. The conformation of the cytoplasmic domains of the PRLRs has not been examined but is unlikely to be linear. It is interesting in this regard that the BRET ratio generated with either U-PRL or S179D PRL is very similar whether a LF-LF dimer or a LF-SF1b dimer is formed, despite the 334-amino acid difference in length between the LF and S1b receptor. Thus, it seems that with folding, the intracellular tags in all likelihood are not as far apart as the length of the amino acid chain would imply In summary, in our present study, U-PRL- and S179D PRL-induced BRET was observed with all combinations of long- and short-receptor isoforms except the combination of SF1a plus SF1b isoforms. This is the first demonstration that human PRL promotes homo- and heterodimerization of most long and short human PRLR pairs in living cells and that this occurs using species homologous hormone and receptors. The results also highlight the incompatibility of SF1a and SF1b and exclude the formation of different receptor pairs as an explanation for the different signaling abilities of U-PRL and S179D PRL. MATERIALS AND METHODS Materials The codon-humanized pgfp 2 -N1 (h) and prluc-n1 (h) vectors, containing multiple cloning sites upstream of the GFP 2 gene or Rluc gene, and the luciferase substrate, DeepBlueC, were purchased from PerkinElmer. Coelenterazine h, another substrate for Rluc was purchased from Assay Designs, Inc. (Ann Arbor, MI). The PRLR cdnas, encoding LF, SF1a and SF1b cloned into pef6c, pef4c, and pcrscript were established as described previously (13). A -casein-luc plasmid, which contains 2.4 kb of the -casein promoter upstream of a luciferase reporter gene (40), was a generous gift from Dr. Linda Schuler (Madison, WI) and Jeffrey Rosen (Baylor, TX). U-PRL and S179D PRL were purified from Escherichia coli (DH5 ) as described previously (30). G129R PRL was purified from E. coli as described by Zhang et al. (39). BRET Assay In a BRET assay, excited-state electron energy generated by an Rluc substrate oxidation reaction is transferred to an energy acceptor molecule when it is within about 100 Å. The excited acceptor molecule then emits light at its specific emission wavelength. The BRET 2 technique we used improves upon the original BRET by the use of a new substrate, DeepBlueC, and a new acceptor fluorophore, a modified GFP, GFP 2. In the original BRET, bioluminescence from Rluc using the older coelenterazine h substrate peaks between 475 nm and 480 nm and the fluorescence from the acceptor, yellow fluorescent protein between 520 nm and 530 nm, resulting in poor peak resolution (only nm) between donor and acceptor emission. In addition, the emission from Rluc using coelenterazine h is very broad, and overlaps with yellow fluorescent protein emission. The DeepBlueC substrate, by contrast, has a peak emission at nm and the GFP 2 acceptor has a peak emission at nm. Thus, peak resolution is increased to about 115 nm and overlapping donor and acceptor emission is markedly reduced, leading to greater BRET sensitivity. Also in the current study, we used a scanning spectrofluorometer so as to record whole spectral scans with each set of cells. This controlled for two things: coexpression of Rluc- and GFP 2 -tagged receptors so that we could properly interpret a lack of GFP 2 emission, and spectral shifts that may have resulted from association or dissociation of other proteins from the receptors upon initiation of signaling. As it turned out, no significant spectral shifts were observed and so future experiments can be conducted by monitoring only peak wavelengths. Construction of PRLR Eukaryotic Expression Vectors Tagged with GFP 2 and Rluc To prepare PRLR tagged with GFP 2 (LF-GFP 2, SF1a-GFP 2, SF1b-GFP 2 ) and Rluc (LF-Rluc, SF1a-Rluc, SF1b-Rluc) at the carboxy terminus, the entire coding sequences of the LF (1866 bp), SF1a (1128 bp) and SF1b (864 bp) receptors without their stop codons were amplified by PCR from the original plasmids, pef6-lf, pef4-sf1a and pcr2.1-sf1b,

11 Tan et al. Homo- and Hetero-Pairing of PRLR Isoforms Mol Endocrinol, May 2005, 19(5): respectively, by using Taq DNA polymerase (Promega, Madison, WI) and sense/antisense primers harboring unique MluI (at sense) and KpnI (at antisense) restriction sites. The extra DNA base pairs corresponding to MluI (ACGCGT) and KpnI (GGTACC) were designed to be upstream of the initiator codon on the sense and to replace the stop codon on the antisense. An extra triplet ACC was also designed to be immediately before the initiator codon, ATG, to facilitate and therefore increase the transcription rate in transfected mammalian cells. The primers (sense/antisense) were as follows (cleavage sites for MluI and KpnI are shown in boldface): LF: 5 -GAC ACGCGT ACC ATG AAG GAA AAT GTG- 3 (sense)/5 -AAC GGTACC A GTG AAA GGA GTG TGT-3 (antisense). SF1a: 5 -GAC ACGCGT ACC ATG AAG GAA AAT GTG-3 (sense)/5 -AAC GGTACCA CTG GAC TGT GGT CAA-3 (antisense). SF1b: 5 -GAC ACGCGT ACC ATG AAG GAA AAT GTG- 3 (sense)/5 -AAC GGTACCA AGG GGT CAC CTC CAA-3 (antisense). GFP 2 -N1 and Rluc-N1, which have a cytomegalovirus early promoter and a polyadenylation site, as well as the linearized PCR product of PRLR cdna, were then digested with MluI and KpnI and purified from an agarose preparative gel. The fragments then were subcloned in-frame into the MluI/KpnI site of the GFP 2 -N1 and Rluc-N1 vectors to give fusion plasmids of LF-pGFP 2 N1, SF1a-pGFP 2 N1, SF1bpGFP 2 N1, LF-pRlucN1, SF1a-RlucN1 and SF1b-pRlucN1. Sequence analyses were performed to verify the correct orientation and open reading frame of the newly made constructs. The fusion proteins contain a linker peptide of 12 and 13 amino acids in GFP 2 - and Rluc-tagged receptor, respectively (Fig. 1). Cell Culture and Transfection HEK293 cells were maintained in DMEM (Invitrogen, Carlsbad, CA) containing high glucose, 1 mm sodium pyruvate, and pyridoxine hydrochloride, 2 mm L-glutamine, supplemented with 10% fetal bovine serum, 100 U/ml penicillin, and 0.7 M streptomycin. The cells were seeded at a density of cells per well of six-well (or as indicated) tissue culture dishes. Transient transfections were performed the following day when the cells were 90 95% confluent using Lipofectamine 2000 (Invitrogen) in accordance with the protocol provided by the vendor. Briefly, g DNA was used per well. The DNA was initially incubated in 250 l of Opti-MEM I Reduced Serum Medium (Invitrogen) (without serum and antibiotics) and mixed gently. A 1:25 dilution of Lipofectamine 2000 in 250 l of the same medium was then prepared. After a 5-min incubation at room temperature, the diluted DNA and Lipofectamine 2000 solutions were mixed and incubated at room temperature for 20 min to allow the DNA-Lipofectamine 2000 complexes to form. The complexes were then added to the cells (in medium with serum without antibiotics) and mixed by rocking the plate back and forth. After a 48-h incubation of the cells at 37 C in a CO 2 incubator, the cells were subjected to bioluminescence, fluorescence, and BRET 2 analyses. PRL-Induced Expression of -Casein-Luciferase Assay The biological activities of tagged PRLRs were analyzed using a functional bioassay based on cotransfection with an approximately 2.4-kb portion of the -casein gene promoter fused to a luciferase reporter. For the analysis of heterodimers, equal amounts of LF cdna were transfected in the LF alone and LF plus SF. Total transfected DNA was kept constant by adjustment with a control vector. Twenty-four hours after transfection, the medium was changed to DMEM without serum containing 45.7 nm recombinant PRLs. After a further 24 h, the cells were washed three times with Dulbecco s PBS (DPBS), and then reporter lysis buffer (15 l/cm 2 ) was added to the plate. To ensure complete lysis, a freeze thaw was performed followed by a 15-min shake at room temperature. The lysed cells were scraped off the dish and centrifuged (12,000 g) for 5 min. Twenty microliters of the supernatant were added into luminometer tubes containing 50 l of luciferase assay reagent containing luciferin, and the relative luminescence signal was measured using a Monolight 2010 luminometer (Analytical Luminescence, San Diego, CA). For the activity assay of Rluc-PRLRs, three separate transfections including LF-Rluc only, SF1a-Rluc only and SF1b-Rluc only were performed and the normalized counts were subtracted from normalized LF-Rluc/ -casein-luc, SF1a- -casein-luc, and SF1b- -casein-luc, respectively. Expression Level of GFP 2 - and Rluc-Tagged Receptors The relative expression levels of GFP 2 - and Rluc-tagged PRLR constructs in the cells were estimated by comparison of the fluorescence and bioluminescence from cells cotransfected with tagged PRLRs to cells transfected with a GFP 2 - Rluc fusion plasmid. The fluorescence emitted by GFP 2 was measured with a FluorMax II spectrofluorometer (JY/Horiba, Edison, NJ) with excitation at 405 nm. Bioluminescence of the transfected cells was measured after the addition of coelenterazine h (5 M) using the same spectrofluorometer but with a black card placed in the excitation-beam path. Coelenterazine h and not DeepBlueC was used in these measurements because coelenterazine h oxidation by Rluc is slower than DeepBlueC and thus provides the necessary extra time for more complete spectroscopic measurements. Also, the much longer emission wavelength associated with coelenterazine h oxidation (475 nm) reduces the possibility of donor quenching and increases the accuracy of the measurements. Confocal Imaging Confocal microscopy (Zeiss 510; Zeiss, Jena, Germany) was applied to check the cellular expression and localization of GFP 2 -tagged PRLRs. HEK293 cells were plated at a density of cells/well on polylysine-coated coverslips ( 12 mm) and cultured in DMEM as described above. One day after plating, when the cells reached about 90% confluence, the cells were transfected with 0.8 g DNA/35-mm well using Lipofectamine 2000 as described above. At various times after transfection, microscopic observation was performed. Fluorescence and BRET 2 Measurements Transfected cells were harvested within 48 h of transfection by washing with DPBS (three times), detachment with DPBS containing 2 mm EDTA centrifugation at 800 rpm and resuspension in BRET 2 buffer (DPBS containing 0.9 mm CaCl 2, 0.5 mm MgCl 2 6H 2 O and 5.5 mm D-glucose) at a density of /ml. Before the BRET 2 measurements, the cells were incubated at 37 C for at least 1 h. For each measurement, 0.5 ml cell suspension was loaded into a 0.5-cm 2 quartz cuvette. Fluorescence and bioluminescence spectral scanning were performed by using a FluoroMax-2 spectrofluorometer (Jobin Yvon Inc., Edison, NJ). Bioluminescence scanning and BRET 2 signal detection were carried out immediately after the addition of the ligand and 5 M of the cell permeant luciferase substrate, DeepBlueC, with a black card in the path of the excitation light beam. Data were collected with the slits set at 5 nm, datum points collected every 5 nm, and signal integration for 0.5 sec per datum point). Energy transfer was defined as the BRET ratio (Emission 500 nm-520 nm -background 500 nm-520 nm) / (Emission 385 nm-420 nm -background 385 nm-420 nm ). The signals obtained from the nontransfected cells were considered background.