Determinants of 5-Lipoxygenase Nuclear Localization Using Green Fluorescent Protein/5-Lipoxygenase Fusion Proteins*

5-Lipoxygenase catalyzes the first two steps in the biosynthesis of leukotrienes, potent extracellular mediators of inflammation and allergic disorders. The unanticipated observation of 5-lipoxygenase in the nucleus of some cell types including bone marrow-derived mast cells (Chen, X. S., Naumann, T. A., Kurre, U., Jenkins, N. A., Copeland, N. G., and Funk, C. D. (1995) J. Biol. Chem. 270, 17993–17999) has raised speculation about intranuclear actions of leukotrienes or the enzyme itself. To explore the entry of 5-lipoxygenase into the nucleus we have transfected various cell types with expression vectors encoding native 5-lipoxygenase and green fluorescent protein/5-lipoxygenase (GFP-5LO) fusion proteins. 5-Lipoxygenase and green fluorescent protein/5-lipoxygenase co-localized with the nuclear DNA stain Hoechst 33258 in each cell type. The three main basic regions of 5-lipoxygenase were incapable of acting as “classical” nuclear localization signal sequences. Mutations that abolished enzyme activity/non-heme iron resulted in proteins that would no longer enter the nucleus. An NH2-terminal 5-lipoxygenase fragment of 80 residues was sufficient for directing nuclear localization of green fluorescent protein but not cytosolic pyruvate kinase. The combined data suggest that 5-lipoxygenase enters the nucleus not by a classical nuclear localization signal but by a non-conventional signal located in the predicted β-barrel domain that may be masked by structural alterations.


5-
Mutations that abolished enzyme activity/nonheme iron resulted in proteins that would no longer enter the nucleus. An NH 2 -terminal 5-lipoxygenase fragment of 80 residues was sufficient for directing nuclear localization of green fluorescent protein but not cytosolic pyruvate kinase. The combined data suggest that 5-lipoxygenase enters the nucleus not by a classical nuclear localization signal but by a non-conventional signal located in the predicted ␤-barrel domain that may be masked by structural alterations. 5-Lipoxygenase (arachidonate:oxygen 5-oxidoreductase, EC 1.13.11.34) is a non-heme iron enzyme found primarily in white blood cells, macrophages, and mast cells that converts arachidonic acid first to 5-hydroperoxyeicosatetraenoic acid (5-HPETE) 1 and then to leukotriene (LT)A 4 (5,6-oxido-7,9,11,14eicosatetraenoic acid) (1). Subsequent conversion of leukotriene A 4 by leukotriene A 4 hydrolase yields the potent neutrophil chemoattractant leukotriene B 4 . Alternatively, conjugation of LTA 4 with glutathione by leukotriene C 4 synthase plus downstream metabolism leads to the cysteinyl leukotrienes that influence airway reactivity and mucus secretion especially in asthmatics (1)(2)(3).
5-Lipoxygenase was isolated originally from the cytosol fraction of human and porcine neutrophils (4,5). The enzyme was shown subsequently to undergo a calcium-dependent translocation to the nuclear envelope upon ionophore A23187 stimulation and was dependent on the 5-lipoxygenase-activating protein situated in this location for leukotriene biosynthesis (6,7). More recent work based on immunofluorescence techniques and cellular fractionation demonstrated that certain cell types capable of leukotriene formation (alveolar macrophages, rat basophilic leukemia cells, and mouse bone marrow-derived mast cells) express 5-lipoxygenase completely or partially in the nucleus (8 -11). Additionally, it was shown that cytosolic 5-lipoxygenase in rat neutrophils could enter the nucleus if they were first elicited in vivo with various inflammatory agents or subjected to adherence to glass in vitro (12).
The discovery of 5-lipoxygenase in the nucleus was a surprising observation since it is well known that leukotrienes must exit the cell once synthesized to act on cell surface G proteincoupled receptors to exert their actions on neutrophils or bronchiole smooth muscle (13,14). The possibility of 5-lipoxygenase itself or leukotrienes acting in the nucleus was raised. The recent observation that LTB 4 could bind to the nuclear peroxisomal proliferator-activated receptor-␣ indicated that intranuclear actions of leukotrienes are feasible (15).
Nothing is known about control of 5-lipoxygenase entry into the nucleus in some cell types but not others. Proteins enter the nucleus by nuclear localization signal (NLS) sequences that are recognized by specific importins, prior to nuclear pore docking, translocation through the pore and release from the pore's inner side (16,17). The NLS is typically a short basic region or bipartite basic sequence (18,19). Increasingly, however, novel NLS sequences are being recognized for import of particular classes of proteins; for example, the 38-amino acid M9 domain of heterogeneous nuclear ribonucleoprotein A1 (20,21). Here, we demonstrate primarily with the use of green fluorescent protein (GFP)/5-lipoxygenase fusion proteins the complexity of events for 5-lipoxygenase nuclear entry.

EXPERIMENTAL PROCEDURES
Construction of Expression Vectors-The full-length cDNA encoding human 5-lipoxygenase from pT3-5LO (22) subsequently cloned into pcDNA3 (Invitrogen) or a version with deletion of sequence encoding six amino acids at the carboxyl terminus (22) was cloned into the EcoRI/ ApaI sites of a GFP vector pEGFP-C2 (CLONTECH) to obtain pEGFP-C2/5LO (GFP-5LO) and pEGFP-C2/5LO(C6-deletion) (GFP-5LO(C6deletion)). To construct GFP-SV40/NLS including a sequence representing the NLS (PKKKRKV) from SV40 large T antigen (18) and GFP-5LO(128 -135) in which basic amino acid region 2 (see Fig. 3 and "Results") is included, two complementary pairs of oligonucleotides with restriction enzyme overhangs were synthesized as follows: AAT-TCCCCAAGAAGAAGCGAAAGGTGGG for SV40/NLS and AAT-TCAAGCAACACCGACGTAAAGAACTGGG for the second basic region of 5-lipoxygenase (bases for EcoRI and BamHI sites are underlined). The oligonucleotides with 5Ј-ends phosphorylated by T4 polynucleotide kinase were annealed and ligated into the EcoRI and BamHI sites of pEGFP-C2. Correct insertion of the oligonucleotides was verified by automated DNA sequencing (Applied BioSystems BigDye Terminator Ready reaction kit reagents; ABI 373 sequencer) using the facilities of the Department of Genetics.
The plasmid pcDNA3/Myc pyruvate kinase (PK) was used for construction of PK-5LO and PK-5LO mutant fusion proteins (20). Generally, cDNAs for 5-lipoxygenase and its mutants were cloned into the KpnI site that corresponds to codon 443 of PK and different restriction sites available in the downstream linker region.
In vitro mutagenesis was carried out using the QuikChange Sitedirected Mutagenesis kit (Stratagene) or by polymerase chain reaction on pEGFP-C2/5LO template. Truncations of human 5-lipoxygenase or platelet-type 12-lipoxygenase cDNA were performed by subcloning and/or polymerase chain reaction on pEGFP-C2/5LO, pcDNA3/5LO, and pcDNA/6His-12LX (23). Details of the procedures and the sequences of mutagenic primers are available upon request. DNA sequencing as mentioned above was used to verify the introduced mutations and polymerase chain reaction products.
Human embryonic kidney (HEK) 293 cells and NIH-3T3 fibroblasts were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum. Chinese hamster ovary (CHO) cells were cultured in Ham's F-12 medium with 10% fetal bovine serum. Cells growing in Petri dishes or 0.1% gelatin-coated glass chamber slides were transfected with plasmid DNA by calcium phosphate co-precipitation techniques (10). The transfected cells were used for GFP fluorescence microscopy (20 h post-transfection), protein preparation, 5-lipoxygenase activity assay, and immunocytochemistry (40 h post-transfection).
Fluorescence Microscopy-Indirect immunofluorescence analysis was carried out as described (10) for mouse BMMC with the following modifications: 5LO ϩ/ϩ BMMC, 5LO Ϫ/Ϫ transfected BMMC and HEK 293 cells transfected with pcDNA3/human or mouse 5LO plasmid were fixed with 2% paraformaldehyde in PBS for 20 min at room temperature followed by a 10-min incubation in 0.3% Triton X-100 in PBS. Cells were incubated with 3% bovine serum albumin for 30 min to block nonspecific binding and then incubated with rabbit anti-human 5LO antiserum (1:1000; see below) or monoclonal anti-Myc tag antibody (9E10; 1:1,500; (20)) in PBS with 3% bovine serum albumin for 1 h at room temperature, followed by a 30-min incubation with Cy3-conjugated donkey anti-rabbit IgG (1:4000; Jackson ImmunoResearch Laboratories) in PBS containing 3% bovine serum albumin. After three washes with PBS, cells were incubated with the DNA stain Hoechst 33258 (Molecular Probes, 0.5 g/ml in PBS) for 5 min and mounted with Gel/Mount (Biomedia Corp.) after brief rinses with PBS.
For initial assessment of time course of GFP-5LO expression, fluorescence was monitored in living cells with a Nikon inverted microscope equipped with fluorescence capability. For data documentation, HEK 293, NIH-3T3 fibroblasts, and CHO cells transfected with pEGFP-C2/ human 5LO or its mutant plasmids were fixed with 2% paraformaldehyde in PBS for 20 min followed by a 5-min incubation with Hoechst 33258 stain. Slides were mounted with Gel/Mount and kept at 4°C.
Slides were examined with an Olympus AX-70 analytical microscope equipped with a Sensys CCD-digital camera (Photometrics Ltd., Tucson, AR). Fluorescence was analyzed with a combination filter cube for Cy3 and Hoechst 33258/bound DNA (excitation/emission for these fluorophores, 554/568 and 346/460 nm) and a fluorescein isothiocyanate filter cube for GFP fluorescence (490/525 nm). Raw data photoimages were acquired through IPLab software (Signal Analytics Corp., Vienna, VA) and processed further using Adobe Photoshop and Canvas programs. The fluorescence data are based on transfections that were performed in multiple experiments (n ϭ 3-5).

Transfection of cDNA Encoding 5-Lipoxygenase into Various
Cell Types Results in Nuclear Localization-Since 5-lipoxygenase is situated in the cytosol of resting neutrophils, the nucleus of BMMC and in both compartments in rat basophilic leukemia-1 cells and macrophages, we sought to determine the cellular localization of 5-lipoxygenase in various cell types transfected with 5-lipoxygenase expression vectors prior to detailed analysis of NLS sequences. As previously determined (10), 5-lipoxygenase was detected in the nucleus of BMMC using indirect immunofluorescence studies (Fig. 1A). A similar expression pattern was observed in transfected BMMC obtained from 5LO Ϫ/Ϫ mice that had been generated by gene targeting techniques (Fig. 1A). Nuclear localization was also observed for either human or mouse 5-lipoxygenase in transfected HEK 293 cells (Fig. 1A).
A GFP Tag Attached to the NH 2 Terminus of 5-Lipoxygenase Does Not Alter Enzyme Activity or Cellular Localization in Transfected Cells-To simplify the localization studies we developed a construct encoding a GFP-5LO fusion protein. In transfected HEK 293 cells, CHO cells, and NIH-3T3 fibroblasts (Fig. 1B) the fusion protein co-localized with the nuclear stain Hoechst 33258 as in native 5-lipoxygenase-transfected cells. Comparison of the enzyme activity of native 5-lipoxygenase and GFP-5LO in HEK 293 sonicated cell supernatants indicated that arachidonic acid was converted to 5-HPETE and 5-HETE to approximately the same extent (Fig. 2). In agreement with previous studies, GFP alone exhibited a nonspecific cellular expression pattern (26).
Analysis of Basic Regions of 5-Lipoxygenase as NLS Sequences-The classical NLS is a short basic region of 3 or more residues or a bipartite basic region separated by a variable length spacer sequence (18,19). In the original cloning of the 5-lipoxygenase cDNA one of the key features we mentioned about the structure of this protein was the presence of pairs or small clusters of basic residues (27). We chose the three most likely candidate regions to represent a classical NLS that differed to some extent from other non-nuclear localized lipoxygenases for mutational analysis (Fig. 3). Three double mutant GFP-5LO constructs were prepared and transfected into HEK 293 cells (Fig. 4). Both the basic region 1 (R72Q/K73Q) and basic region 3 (K653Q/K654Q) constructs yielded nuclear localization patterns identical to GFP-5LO (Fig. 4). However, the double mutant of basic region 2 (R131Q/R132Q) yielded a predominantly cytosolic localization pattern. All proteins were expressed at similar levels in the transfected cells (Fig. 4C). When we tested the enzymatic activity of the double mutants the basic region 1 and 3 mutations did not influence 5-lipoxygenase activity but the R131Q/R132Q mutations abolished enzyme activity (Fig. 2, left panel). This result raised the possibility that basic region 2 was not really a NLS but that the 5-Lipoxygenase Nuclear Localization enzyme activity/folding of the protein might also influence nuclear localization.
Analysis of 5-Lipoxygenase Structural Alterations on Nuclear Localization-5-Lipoxygenase contains a non-heme iron atom essential for catalytic activity (28). Based on the crystal structures of soybean and rabbit reticulocyte lipoxygenases (29,30) and mutational analyses of 5-lipoxygenase (22,31,32), there are three important histidine residues (His-367, -372, and -550) and the COOH-terminal isoleucine (Ile-673) which participate in binding the iron atom and presumably which are essential for maintaining proper structural integrity for enzyme activity. Five constructs with mutations to abolish enzyme activity and influence iron binding were prepared and introduced into HEK 293 cells (Fig. 5). Mutations known to completely abolish iron presence in 5-lipoxygenase (H372Q, H550Q, and C6-deletion (deletion of the COOH-terminal 6 amino acid residues including Ile-673); Refs. 31 and 32) led to nearly exclusive cytosoliclocalized enzyme (Fig. 5B). Mutations known to abolish enzyme activity yet leave partial presence of iron in the protein (H367Q, H367N) showed graded distribution of 5-lipoxygenase in the nucleus and cytosol depending on the iron content ( Fig.  5B; Ref. 31). All the mutated 5-lipoxygenase constructs were expressed in the transfected cells to a similar extent (Fig. 5C) but had lost enzyme activity by HPLC analysis.
Are There Regions of 5-Lipoxygenase Sufficient for Directing Nuclear Localization?-Various portions of 5-lipoxygenase were fused to GFP to see if a specific region of the protein could direct nuclear localization. Initially, four constructs were designed based on restriction enzyme sites suitable for cloning FIG. 1. Immunofluorescence analysis depicts the subcellular location of 5-lipoxygenase or GFP-5LO in mouse BMMC, HEK 293, CHO, or NIH-3T3 fibroblast cells. A, BMMC from wild-type mice (non-transfected) or 5LO Ϫ/Ϫ mice were electroporated with pcDNA3/mouse 5LO and cytospun onto glass microscope slides (48 h post-electroporation). HEK 293 cells were cultured on chamber slides and transfected with pcDNA3/human 5LO or mouse 5LO. Cells were immunocytostained indirectly with anti-human 5LO antiserum and Cy3-conjugated secondary antibody, followed by incubation with the DNA stain Hoechst 33258 as detailed under "Experimental Procedures." Very few 5LO Ϫ/Ϫ BMMC were transfected successfully, but those that were exhibited intense fluorescence within the nucleus. To document the fluorescence in more than one cell on one micrograph the magnification was reduced (40 ϫ objective instead of ϫ 100) relative to other panels. B, NIH-3T3 fibroblasts and CHO cells cultured on chamber slides were transfected with pEGFP-C2/5LO and were stained with Hoechst 33258 at 40 h post-transfection. Co-localization of 5-lipoxygenase with DNA stain was observed in all cases.   Fig. 4 legend for details). A GFP monoclonal antibody was used for the immunoblot analysis. Asterisk, iron content of mutant proteins is based on Refs. 31 and 32 assuming no changes for the extra GFP addition. Mutants that retain partial structural characteristics (as estimated by iron content) can direct some nuclear localization whereas those proteins that have lost iron no longer enter the nucleus.

FIG. 6. Determination if 5-lipoxygenase fragments can direct GFP to the nucleus.
A-C, GFP-5LO fusion drawings, fluorescence micrographs, and immunoblot analysis of expressed 5LO proteins (see Fig. 4 legend and "Results" for details). A GFP monoclonal antibody was used for the immunoblot analysis. An NLS peptide from SV40 was used as a control. A sequence in platelet 12-lipoxygenase (P-12LO(1-75)) corresponding to residues 1-80 of 5-lipoxygenase that directed nuclear localization was used as another control.

5-Lipoxygenase Nuclear Localization
(5LO(1-80), 5LO(80 -673), 5LO(80 -564), and 5LO(564 -673)). None of these constructs, with the exception of 5LO(1-80) could direct a nuclear localization pattern (Fig. 6). The pattern of nuclear localization gave a mottled appearance as previously determined for 5-lipoxygenase in alveolar macrophages (11). As a control, we fused the 7-residue SV40/NLS to GFP and found that it directed a predominant nuclear expression pattern (data not shown). In contrast, an 8-residue sequence (basic region 2, Fig. 3; 5LO(128 -135)) did not direct nuclear localization although mutations at positions 131/132 in the GFP-5LO fusion altered cellular localization as mentioned above. Since the NH 2 -terminal region of 5-lipoxygenase appeared sufficient to direct nuclear localization we constructed a GFP fusion construct from the exact region of another lipoxygenase known not to enter the nucleus ("platelet-type" 12-lipoxygenase; P12LO(1-75)). This construct did not enter the nucleus (Fig.  6B). Mutation of basic region 1 located in the 5LO(1-80) construct also did not alter nuclear targeting. However, extending the region 5LO(1-127) and 5LO(1-166) resulted in a diminished ability to target to the nucleus. Once again all the various constructs were expressed well in the transfected HEK 293 cells as judged by immunoblot analysis (Fig. 6C).

5-Lipoxygenase Domains Are Incapable of Directing the Cytosolic Protein Pyruvate
Kinase to the Nucleus-We tested the ability of various 5-lipoxygenase fusions to direct a commonly used cytosolic protein, pyruvate kinase (20,21) to the nucleus (Fig. 7). Neither the full-length 5-lipoxygenase, which retained enzyme activity with the PK fusion, nor shortened fragments of 5-lipoxygenase could direct the enzyme to the nucleus, although the M9 NLS sequence of heterogeneous nuclear ribonucleoprotein A1 could (data not shown). DISCUSSION We have carried out a study to investigate the nuclear targeting of 5-lipoxygenase. More than 25 fusion protein constructs were prepared and analyzed by immunofluorescence microscopy in various transfected cell lines. The combined data indicate that 5-lipoxygenase nuclear targeting is dependent on a number of complex factors. First, 5-lipoxygenase does not possess a "classical" basic region that functions as a NLS although the protein does contain at least three potential basic cluster regions that could act in this function. Second, the proper folding of the enzyme is critical for nuclear localization. This result is based on the fact that fusion proteins, which completely lack or partially retain the non-heme iron atom, show a predominantly cytosolic localization. Third, a short NH 2 -terminal region (first 80 amino acids) appears sufficient for directing the enzyme to the nucleus but when this segment is extended the recognition sequence(s) are lost. Fourth, any NLS sequence(s) in the NH 2 -terminal region of 5LO are relatively weak since only one of two proteins tested could be transported to the nucleus (e.g. GFP versus pyruvate kinase).
The nuclear import of proteins bearing the classical NLS such as those for the SV40 large T antigen and nucleoplasmin begins by binding to karyopherin-␣ which acts as the NLS receptor (17). Karyopherin-␤1 interacts further with the NLSbound karyopherin-␣ to form the ternary complex, which is targeted to the nucleoporins in the nuclear pore complex. Subsequent translocation into the nucleus through the nuclear pore complex depends on GTPase Ran and its modulators (16,17). Novel import pathways have been identified recently. For instance, heterogeneous nuclear ribonucleoprotein A1 interacts directly with karyopherin-␤2/transportin, which is one of the members in the ␤-karyopherin superfamily, through the substrate's distinct NLS known as M9 or NLS2 to target to the nuclear pore complex (20,21). Common features of the novel nuclear import pathways appear to be NLS sequences distinct from the classical pathway and direct interaction with an individual karyopherin-␤ form independent of interaction with the adapter karyopherin-␣. Perhaps 5-lipoxygenase is using a novel means of nuclear entry in this respect. 5-Lipoxygenase does not contain an M9 domain and it is not yet known if it can interact with karyopherin-␣.
5-Lipoxygenase is the only known mammalian lipoxygenase that resides in the nucleus. The NH 2 termini of lipoxygenases differ to the greatest extent in this region (33,34). Thus, this region appears to be the most important for 5LO nuclear targeting since there was one fragment (1-80) that could direct GFP to the nucleus. This was not a nonspecific result since both a COOH-terminal fragment from 5-lipoxygenase (5LO(574 -673), see Fig. 6) of similar size and a fragment corresponding to the same residues of platelet 12-lipoxygenase (1-75; in the NH 2 -terminal sequence five residues are not present in this lipoxygenase and several other mammalian lipoxygenases) did not direct nuclear localization. Platelet 12-lipoxygenase is known to reside in the cytosol, either soluble or membranebound, of human erythroleukemia cells, A431 cells, and epidermal homogenates (23,35,36). The basic cluster within the 5LO(1-80) sequence was not essential for nuclear localization.
Based on the three-dimensional x-ray crystal structures of soybean lipoxygenases and rabbit reticulocyte 15-lipoxygenase (29,30,37), the lipoxygenase family members possess two domains; a short ␤-barrel NH 2 -terminal domain of unknown function and a major catalytic domain that includes the nonheme iron atom. Gillmor et al. (30) have hypothesized that the ␤-barrel NH 2 -terminal region, with homology to lipoprotein lipase, may participate in binding lipid membranes to gain access to the source of substrate. In the case of 5-lipoxygenase, they suggested a possible site of interaction with 5-lipoxygenase-activating protein, a co-accessory protein in leukotriene biosynthesis that may help to "transfer" arachidonic acid substrate to the enzyme (7,38). The data here could extend the possible list of functions for this domain, in particular for 5-lipoxygenase, as aiding in nuclear localization. The putative ␤-barrel domain of 5-lipoxygenase is about 125 amino acid residues in length. A construct with this domain still directed nuclear localization of GFP. However, if it was extended to 166 amino acids the fusion protein remained cytosolic. This result suggests that the folding of the extra portion beyond the ␤-barrel may have masked any potential NLS. The context within which the NLS is situated is important for nuclear localization (39,40).
Precedents for weak NLS sequences in proteins are prevalent in the literature and 5-lipoxygenase seems to fit in this class of proteins. For example, a 29-amino acid stretch of GAL4 could direct cytosolic invertase to the nucleus but not ␤-galactosidase (40). Likewise, a NLS motif near the NH 2 terminus of fibroblast growth factor 3 conferred nuclear localization to cytoplasmic ␤-galactosidase but not pyruvate kinase (41). Signals in 5-lipoxygenase were capable of directing the GFP reporter protein to the nucleus but not pyruvate kinase. Complex patterns of nuclear targeting that may involve weak additive signals from opposite ends of the protein are known (e.g. fibroblast growth factor 3) (41). Examples of NLS "masking" by structural alterations and/or other cellular proteins, a prime example being the transcription factor NF-B/Rel bound by its inhibitor IB are known (42). The data herein could be consistent with some sort of unmasking of an NLS to gain nuclear entry and also do not rule out the possibility that 5-lipoxygenase is "piggybacked" to the nucleus by some other chaperone protein. A somewhat surprising result was that 5-lipoxygenase could be localized to the nucleus in four different transfected cell types. Perhaps, in human and rat neutrophils, 5-lipoxygenase is specifically bound by an inhibitor protein that prevents nuclear transport since the enzyme in these resting cells is exclusively cytosolic. Upon in vivo activation or adherence to glass in vitro (12), therefore, these cells would lose the capacity to behind the inhibitor and enter the nucleus. Alternatively, some sort of post-translational modification such as phosphorylation of 5-lipoxygenase, particularly in neutrophils, may influence NLS recognition. Phosphorylation of a nuclear 5-lipoxygenase fraction has been detected in HL-60 cells but the site of phosphorylation has not been identified (43). We are currently studying potential interactions of 5-lipoxygenase with other proteins using the yeast two-hybrid system with relevance to the nuclear targeting paradigm.
The use of 5-lipoxygenase inhibitors and leukotriene receptor antagonists has made its widespread debut in the clinical arena for asthma treatment in the last few years (44). The recent discoveries of 5-lipoxygenase in the nucleus, that LTB 4 can bind and activate a nuclear transcription factor (peroxisomal proliferator-activated receptor-␣), and the potential for other nuclear functions make it essential to understand the mechanisms for nuclear localization of 5-lipoxygenase. This study has raised new questions in beginning to answer these issues.