The effect of protein context on nuclear location signal function

Cell, Vol. 50, 465-475,

July 31. 1987, Copyright

0 1987 by Cell Press

The Effect of Protein Context on Nuclear Location Signal Function Bruce L. Roberts,’ William and Alan E. Smith” ’ integrated Genetics 31 New York Avenue Framingham, Massachusetts t Department of Zoology University College Gower Street London, WClE-6BT

D. Richardson,t

01701

Summary The effect of position and number and of another intracellular location signal on the activity of the nuclear location signal was investigated. A minimal signal was inserted into several sites within the polypeptide chain of pyruvate kinase. The results observed suggest that a nuclear location signal can function at a variety of positions within a protein but that in some locations its activity is masked. Multiple copies of a partially defective signal were integrated into pyruvate kinase. The data suggest that multiple signals can cooperate to enhance nuclear accumulation. A nuclear location signal failed to function when inserted into polyomavirus middle T but was active in an identical variant lacking the carboxy-terminal hydrophobic tail. We conclude that while a minimal nuclear location signal is sufficient for nuclear localization, its activity is crucially dependent on the protein context within which it is present. Introduction A short amino acid sequence involved in nuclear location has been identified in SV40 large T antigen (Kalderon et al., 1984a) and shown to be able to target large T variants to the nucleus, whether located at its original site or appended to the amino terminus (Kalderon et al., 1984b). The suggestion that the sequence Pro Lys Lys Lys Arg Lys Val Glu constitutes an independent structural element specifying nuclear location was confirmed by the finding that the sequence retains its ability to act as a nuclear targeting signal when transposed to the amino termini of 6-galactosidase and pyruvate kinase (Kalderon et al., 1984b). Synthetic peptides homologous to the SV40 large T nuclear signal and covalently coupled to proteins of various sizes can also target molecules to the nucleus (Goldfarb et al., 1986; Lanford et al., 1986). Several known nuclear proteins contain basic sequences analogous to the SV40 large T prototypic signal sequence (Smith et al., 1985) and in some cases such sequences have been implicated in nuclear location (e.g., polyomavirus large T, Richardson et al., 1986; SV40 VP2, Roberts et al., 1986; adenovirus 72K DNA-binding protein, Roberts et al., 1986; SV40 VPl, Wychowski et al., 1986). On the other hand, similar basic sequences are also pres-

ent in a number of non-nuclear proteins. For example, the capsid protein VP3 of polio virus type I contains the sequence Pro Pro Lys Lys Arg Lys Glu (Smith et al., 1985). In this and similar cases, the potential nuclear location activity of the basic sequence appears to be masked. Proteins destined to enter mitochondria, chloroplasts, or the lumen of the endoplasmic reticulum1 contain signal sequences that are often located at the extreme amino termini (Blobel, 1980; Watson, 1984). It would not appear that, in an analogous fashion, nuclear location signals need to be integrated into a specific region of a protein in order to function. For example, the nuclear location signal sequence of SV40 large T antigen is located at an internal position, while nucleoplasmin contains a carboxy-terminal domain specifying a nuclear location (Dingwall et al., 1982). In our study, a minimal nuclear location signal sequence of S/40 large T was inserted into five sites within chicken muscle pyruvate kinase to examine the effect of position on activity and on potential masking. Some nuclear proteins, such as SV40 VPl, contain homologous amino acid sequences that are not as basic as the SV40 large T prototypic nuclear signal (Smith et al., 1985). These sequences are characteristically located at the extreme terminal ends of the polypeptide chain, suggesting that if they are involved in nuclear location they may need to be very exposed in order to function. Alternatively, duplication of such sequences within a protein might be required. Indeed, the large T antigen of polyomavirus contains at least two basic sequences that contribute toward its nuclear accumulation (Richardson et al., 1986). To test whether a weak nuclear location signal is influenced by position and whether weak signals can cooperate to enhance nuclear accumulation, we inserted multiple copies of a partially defective nuclear location signal into various sites within pyruvate kinase. Polyomavirus middle T, unlike pyruvate kinase, is localized to the cellular membrane fraction of eukaryotic cells (Zhu et al., 1984; Dilworth et al., 1986). Acarboxy-terminal hydrophobic region of the protein appears to be responsible for this association and is indispensable for the transforming activity of the protein (Novak and Griffin, 1981; Carmichael et al., 1982; Templeton and IEckhart, 1982; Markland et al., 1986). To explore the effect of protein context on nuclear location signal function, the SV40 large T nuclear signal was inserted into the coding sequence of polyomavirus middle T at a position considered unlikely to interfere with its transforming activity. The subcellular distribution and biological activities of these variants, as well as truncated versions lacking the hydrophobic tail, were examined. Results Transposition of the SV40 Large T Nuclear Signal within Pyruvate Kinase Our previous experiments indicated that following microinjection into Vero cells of chicken muscle pyruvate kinase

Cell 466

B

/

Ph

I SacI

A2

C

Domain

Figure 1. Structural Features cle Pyruvate Kinase

1 Slul

cDNA, the encoded protein is stable and excluded from the cell nucleus (Kalderon et al., 1984b; Richardson et al., 1986). Xhol and EcoRl linkers of three different lengths each were inserted into various restriction enzyme sites within the pyruvate kinase cDNA (Figure 1). The Pvull and BstXl sites were selected because they correspond to regions of the coding sequence close to the amino- and carboxy-terminal ends of the protein (Figure 1). In addition, three internal restriction enzyme sites that are roughly equidistant from one another were selected. With appropriate pairs of Xhol and EcoRl linker insertion mutants, a small DNA fragment encoding an SV40 large T nuclear location signal of only ten amino acids could be inserted in the correct reading frame into these sites. Figure 2 indicates the subcellular distribution of the pyruvate kinase variants 18 hr after microinjection of the plasmid DNAs into Vero cells, together with the predicted amino acid sequences around the signal inserted at the different restriction enzyme sites. Figure 3 displays micrographs of cells expressing the various mutants. As shown previously, a minimal wild-type SV40 large T signal sequence (XR30) containing Lysine at the critical 128 equivalent position (Kalderon et al., 1984b) can direct pyruvate kinase to the nucleus of microinjected Vero cells when appended to the amino terminus of the protein (XR30-PK; Figure 2). Oligonucleotide-directed mutagenesis was used to generate an analogous small DNA fragment (XR35) encoding a Threonine at the Lysine-128 equivalent position of the SV40 large T nuclear signal. The XR35 signal did not discernibly alter the normal cytoplasmic distribution of pyruvate kinase (XR35-PK; Figure 2). A wild-type large T signal could also direct pyruvate kinase to the nucleus when appended to the extreme carboxy1 terminus of the protein (BsxXR30B; Figures 2 and 39). No such effect was observed when the defective XR35 signal was substituted for the wild-type counterpart (BsxXR35B; Figures 2 and 3h). The wild-type large T signal could also direct pyruvate kinase to the nucleus when it was introduced into the cDNA at either the BstEll or Stul sites (BseXR30 and StuXRSOB; Figures 2, 3c, se). However, the signal failed to direct pyruvate kinase to the nucleus when it was inserted into the Sac1 site of the cDNA (SacXRSOB; Figures 2 and 3a). A defective SV40 large T signal did not alter the distribution of pyruvate kinase when inserted into the Sacl, BstEll, or Stul sites (SacXR35B, BseXR35, and StuXR35B; Figures 2, 3b, 3d, 3f). Similar results were ob-

Mus-

The positions of predicted a helices and 6 stands are shown (Muirhead et al., 1966) as are the boundaries of the subunit domains. A partial restriction enzyme map of the chicken muscle pyruvate kinase cDNA is shown. Nucleotides are numbered according to the convention of Lonberg and Gilbert (1983).

cDNA I BstEll

of Chicken

I BstXI

MUTANT

AMINO ACID SEQUENCE

XR30 PK

; :: MDKAEFLEAPKKKRKVEDP~LH

XR35 PK

123 I I MDKAEFLEAPKTKRKVEDP~LH

SXXR3OB

GS&PKKKRKVEDPRNSREFRaA

L.OCATION 16 16

136

‘7

StxXR358

BseXR30

Y4

BeXR35

SfUXRJOB

Y

GDPEFLEAPKTKRKVEDPRNSGD

4/6 VEGLEAPKKKRKVEDPRNSAS

4;7

StuXR358

BsxXR30B

!3SXXR35B

Figure 2. Structure and Subcellular nal Insertion Mutations

Location

of Pyruvate

Kinase

Sig-

The predicted amino acid sequences at the site of signal insertion are shown. Linker-encoded amino acids are underlined, while sequences derived from SV40 large T are boxed. Flanking sequences of the mature pyruvate kinase polypeptide, lacking the initiator methionine, are numbered according to the convention of Lonberg and Gilbert (1963). Signal insertion mutants were either nuclear(N) or cytoplasmic (C), as judged by indirect immunofluorescence 16 hr after microinjection of plasmid DNAs. Note that the first three residues of XRIO-PK and XRWPK are derived from SV40 large T as described previously (Kalderon et al., 1964b).

tained in an earlier preliminary study in which a longer SV40 large T nuclear location signal consisting of 21 amino acids was inserted into the pyruvate kinase cDNA (Roberts et al., 1986). Expression of Signal Insertion Mutants We were previously able to verify that the variant SV40 large T molecules utilized in our studies remained intact by using monoclonal antibodies directed towards the extreme amino- and carboxy-terminal ends (Kalderon et al., 1984b). Since monoclonal antibodies directed against known epitopes in pyruvate kinase were not available for our study, we assessed the structural integrity of pyruvate

The Protein 467

Figure

Context

3. Subcellular

of Nuclear

Location

Signals

of Pyruvate

Photomicrographs showing the subcellular (d) BseXR35, (e) StuXR30B, (f) StuXR356,

Kinase

Signal

Insertion

Mutants

location of the pyruvate kinase variants (g) BsxXRSOB, and (h) BsxXR356.

kinase signal insertion variant proteins by expression in mouse cells, followed by immunoprecipitation and polyacrylamide gel electrophoresis. Signal insertion mutants were expressed in Cl27 cells by the use of bovine papilloma viral shuttle vectors. The pyruvate kinase signal insertion mutants exhibited similar subcellular distributions in transformed Cl27 cells and microinjected Vero cells. Foci of transformed cells were picked and shown to express pyruvate kinase as judged by indirect immunofluorescence. Major protein products of 60,000 daltons were detected by immunoprecipitation of cell extracts with antichicken muscle pyruvate kinase antibody (Figure 4). A minor 60,000 dalton species was also immunoprecipitated from extracts of Cl27 cells transfected with pRL66BPV, and most likely corresponds to the endogenous mouse isozyme. Truncated fragments of SV40 large T, less than 40 kilodaltons in size, are distributed in both the nuclear and cytoplasmic compartments of microinjected Vero cells whether they contain a wild-type or defective nuclear sig-

illustrated

m Figure

2. (a) SacXR306,

(b) SacXR356,

(c) BseXRSO,

nal (Kalderon et al., 1984b). If proteolytic degradation gave rise to fragments of pyruvate kinase, such molecules might be small enough to distribute by diffusion in both the nuclear and cytoplasmic compartments of microinjetted cells, However, as shown above, we were unable to detect such species by immunoprecipitation from radioactively labeled cells. This argues that selective degradation in the nuclear and/or the cytoplasmic compartment cannot account for the different subcellular distributions of the pyruvate kinase variants. Insertion of Multiple into Pyruvate Kinase

Nuclear

Signals

We showed previously that a mutated form of the minimal SV40 large T nuclear signal (XR12) is partially defective in its ability to direct pyruvate kinase to the nucleus (Kalderon et al., 1984b). Thus the fusion protein encoded by plasmid XR12-PK appears in both the nuclear and cytoplasmic compartments of Vero cells 18 hr following microinjection. We chose to insert the XR12 signal into

Cell 468

,A,B,C, ,NI

,D, I

NI

I M

NI

I Y

-69

BSXXR12B

i-u

XR12[1+2]

I

XR12[1*3]

I

WC

I

I

XR12[2+3]

1

NX

I

WC

I

N>C

XR12[1+2+3]

N

128 SRKRKVEOP I

*u

I I.--

Figure 4. lmmunoprecipitation Mutants

of Pyruvate

Figure 5. Structure and Subcellular ants Containing Partially Defective

Kinase

Signal

Insertion

Equal amounts of [35S]methionine-labeled cell extracts were immunoprecipitated with either normal rabbit serum (NI) or rabbit anti-chicken muscle pyruvate kinase antibody (I), and analyzed on 10% SDS containing polyacrylamide gels. BPV-transformed Cl27 cell lines expressing either XRSO-PK (A), SacXR30B (B), or BseXR30 (D) were analyzed, as was a Cl27 cell line transfected with the BPV expression vector alone (pRL66BPV) (C). M denotes molecular weight markers. Similar molecular weight species were detected in cell lines expressing the proteins encoded by pStuXR30B and pBsxXR30B (data not shown).

three sites within the structure of pyruvate kinase in order to determine whether a partially defective nuclear signal might be more susceptible to positional effects than a wild-type SV40 nuclear signal. When the XR12 signal was inserted into either the BstEll or BstXl sites of the PK cDNA, the encoded mutant proteins were localized to both the nuclear and cytoplasmic compartments of Vero cells 18 hr following plasmid microinjection (BseXR12 and BsxXRl2B; Figures 5, 6a, 6b). Hence the partially defective XR12 nuclear signal, like the wild-type XR30 signal, appears to function similarly when transposed to any one of three sites within the structure of pyruvate kinase, at least when assayed under these conditions. We tested the ability of multiple copies of a partially defective nuclear signal to cooperate in directing pyruvate kinase to the nucleus. A mutant (XR12[1+2]) encoded by a plasmid containing single copies of the XR12 signal inserted into both the Pvull and BstEll sites was located predominantly in the nucleus of microinjected cells (Figures 5 and 6~). Similarly, mutants encoded by plasmids containing partially defective signals at both the Pvull and BstXl (XR12[1+3]) or BstEll and BstXl (XR12[2+3]) sites were located predominantly in the nucleus (Figures 5, 6d,

Location of Pyruvate Nuclear Signals

Kinase

Vari-

Plasmids encoding pyruvate kinase variants containing either single, double, or triple copies of the XR12 signal (Kalderon et al., 1984b) were microinjected into Vero cells, and the subcellular distribution of the proteins was determined 18 hr later. Cells exhibited either nuclear and cytoplasmic (N+C), predominantly nuclear (N>C), or nuclear(N) fluorescence.

se). Hence two copies relative to a single copy of the XR12 signal could enhance nuclear accumulation. However, even two copies of the XR12 signal could not bring about the complete nuclear localization of pyruvate kinase under the conditions studied here. Thus a variant containing three partially defective signals was constructed (XR12[1+2+3]; Figure 5). In this case, the encoded pyruvate kinase protein was found only in the nucleus of microinjected Vero cells, as judged by indirect immunofluorescence (Figure 6f). An analogous mutant containing three copies of the defective XR35 signal was confined to the cytoplasm of microinjected cells (data not shown). Insertion of Nuclear Signals into Polyomavirus Middle T Antigen The SV40 large T nuclear signal can efficiently target the cytosolic protein, pyruvate kinase, to the cell nucleus. We wished to know whether the nuclear signal could function when introduced into the coding sequence of a protein that already contains information specifying an alternative cellular location. In our study, nuclear signals were inserted into the middle T antigen of polyomavirus. Middle Tcontains a putative membrane-binding domain proximal to its carboxyl terminus that appears to be responsible for its localization to the cellular membrane fraction of eukaryotic cells (Novak and Griffin, 1981; Carmichael et al., 1982; Templeton and Eckhart, 1982; Markland et al., 1986). Restriction enzyme sites were introduced into the coding sequence of middle T or of a truncation mutant thereof

The Protein 469

Figure

Context

6. Subcellular

of Nuclear

Locations

Signals

of Pyruvate

Kinase

Variants

Containing

Photomicrographs showing the subcellular distribution of the pyruvate XR12[1+2], (d) XR12[1+3], (e) XR12[2+3], and (f) XR12[1+2+3]

(RXT) lacking the hydrophobic tail (Markland et al., 1986), so as to permit the substitution of the DNA sequence encoding residues 294 to 304 of middle T with a small DNA fragment encoding a wild-type SV40 large T nuclear signal (XFi40; Kalderon et al., 19846). This region of middle T has been extensively mutated and the results obtained Indicate that it has little or no influence on the ability of middle T to transform (Ding et al., 1982; Nilsson et al., 1983). We therefore anticipated that the insertion of nuclear signals into this region would have little effect on the transformation properties of middle T, provided the location of the protein was not affected. Oligonucleotide-directed mutagenesis was used to generate a DNA fragment, analogous to the fragment encoding the XR40 signal, encoding a Threonine residue at

Partially kinase

Defective variants

Nuclear illustrated

Signals in Figure

5. (a) BseXR12.

(b) BsxXRlPB,

(c)

the Lysine-128 equivalent position of the nuclear signal (XR45). Analogous constructions to those described above were also created using the XR45 instead of the XR40 signal. Middle T mutants were cloned into an expression vector such that they were under the control of the SV40 early promoter, and were tested for their ability to transform as judged by focus formation on a monolayer of Rat-l cells. Mutant middle T species that contained a hydrophobic tail were indistinguishable from wild-type middle T in their ability to transform Rat-l cells whether they contained a wild-type (SVXD57-XR40-101) or mutated (SVXD57-XR45101) SV40 large T nuclear signal (Figure 7). By contrast, truncated middle T species were unable to transform Rat1 cells whether they contained a wild-type (SVXD57-XR40-

Cell 470

FOCVS LOCATION

FORMATION

100

100

96

SV-RXT

SYXD57-XR40-RXT

0

I-,

S”XD57-XR45-RXT I

N

0

C-N

0

(“I)

Figure 7. Structure and Subcellular of Polyomavirus Middle T Signal Mutants

Location Insertion

Transformation efficiencies on Rat-l cells are expressed as a percentage of that exhibited by wild-type middle T placed under the control of the SV40 promoter (pSV-101). Identical results were obtained whether 5 or 10 ng of plasmid DNA was used to transfect Rat-l cells. For those middle T species that were unable to transform, their subcellular distribution was determined by microinjection. Cells exhibited either perinuclear and cytoplasmic (PN+C), predominantly cytoplasmic (C>N), or nuclear (N) fluorescence.

PPKKKRKVEDP

PPKTKRKVEDP

RXT) or mutated (SVXD57-XR45-RXT) nuclear signal (Figure 7). These results are in agreement with previous studies showing that the hydrophobic tail is required for middle T-mediated transformation (Novak and Griffin, 1981; Carmichael et al., 1982; Templeton and Eckhart, 1982; Markland et al., 1988) and that the region encompassing residues 294 to 304 of middle T is insensitive to mutation (Ding et al., 1982; Nilsson et al., 1983) at least as judged by transforming activity. For those middle T species that were able to transform, foci were picked, expanded, and analyzed for the presence of middle T by indirect immunofluorescence. Middle T species containing either the XR40 or XR45 signal exhibited a perinuclear and cytoplasmic distribution in transformed cells (Figures 7, 8a, 8b). Previous studies have shown that wild-type middle T exhibits a similar perinuclear and cytoplasmic distribution in transformed cells (Dilworth et al., 1986). The subcellular distribution of middle T species that were unable to transform Rat-l cells was determined by the microinjection assay. A truncated middle T species containing the wild-type nuclear signal (SVXD57-XR40RXT) was located in the nucleus of Vero cells 6 hr following microinjection (Figure 8d). By contrast, an analogous truncated middle T species containing the mutated nuclear signal (SVXD57-XR45-RXT) was located predominantly in the cytoplasm (Figure 8~). However, some nuclear accumulation of the protein encoded by plasmid SVXD57-XR45-RXT was evident. This is consistent with previous microinjection data indicating that 40 kilodalton proteins lacking nuclear location signals can slowly accumulate in the cell nucleus, probably by diffusion (Kalderon et al., 1984b). The studies with the truncated species indicate that the nuclear location signal was not inactivated by a positional, masking effect when located in the middle T coding sequence corresponding to residues 294 to 304. We interpret this to mean that the wild-type signal was unable to

target full-length middle T to the nucleus because its activity was dominated by the membrane-binding signal. Discussion Effect of Position on Nuclear Signal Function We have shown here that an SV40 large T nuclear location signal can suffice to direct the cytoplasmic protein pyruvate kinase to the nucleus of microinjected Vero cells when appended to the amino or carboxyl terminus of the protein, and when inserted into two of three internal sites. However, the nuclear signal failed to direct pyruvate kinase to the nucleus when inserted into the polypeptide chain at a position corresponding to the Sacl site of the cDNA. This argues that amino acid sequences with the ability to act as nuclear location signals may not always function as such because their activity is sensitive to the structural environment within which they are present in any given protein. Pyruvate kinase was chosen in part for these studies because the three-dimensional structure of the cat muscle enzyme has been determined (Lonberg and Gilbert, 1985; Muirhead et al., 1986). We therefore attempted to equate the results obtained in these experiments with the known structure of the protein. It is reasonable to suppose that for a signal sequence to be recognized as such, it must be exposed on the surface of a protein. Since the amino- and carboxy-terminal ends of the pyruvate kinase polypeptide are exposed at least in the context of the monomeric subunit (Lonberg and Gilbert, 1985) it is perhaps not surprising that the signals function when appended to these positions. Extrapolating from the crystallographic structural data obtained for cat muscle pyruvate kinase (Lonberg and Gilbert, 1985) the BstEll site of the chicken cDNA corresponds to a hydrophilic exposed region of the polypeptide chain between 8 strand 3 and a helix 4 of domain A2 (Figure 1). Similarly, the Stul site of the cDNA corresponds to an exposed region between a

The Protein 471

Figure

Context

8. Subcellular

of Nuclear

Location

Signals

of Polyomavirus

Middle

T Signal

Insertion

Mutants

The subcellular distribution of the polyomavirus middle T mutants illustrated cells was determined by indirect immunofluorescence. (a) SVXD57-XR40-101,

in Figure 7 in either transformed (a and b) or microinjected (c and d) (b) SVXD57-XR45-101, (c) SVXD57-XR45RXT, (d) SVXD57-XR40-RXT.

helix 2 and 8 strand 1 of domain C (Figure 1). This is consistent with the nuclear location signal functioning when introduced at these sites. By contrast, the Sac1 site at which the signal is inactive corresponds to a position within a region of hydrophobic anti-parallel 8 strands that is buried within domain B (Figure 1; Lonberg and Gilbert, 1985). Although this observation is also consistent with the notion that functional signals must be exposed, we should point out that these secondary structure assignments are not definitive because unlike domains A and C, a high-resolution electron density map is not yet available for domain B (Lonberg and Gilbert, 1985). Furthermore, we do not know the effect of nuclear location signal insertion on inter-subunit interaction, or whether our interpretation of structural environments within the context of the monomeric protein can be extended to the tetrameric form. In addition, we should caution that we have no evidence to demonstrate that the conformations of the pyruvate kinase variant proteins are similar to those of the native enzyme, or that the chicken and cat enzymes have identical structures. The only suggestive evidence is the observation that all the mutant proteins are stable and can be recognized by a polyclonal antibody (Figure 4). In spite of this, our interpretation of the data presented here is that in certain structural contexts, nuclear location signals can be masked. Presumably, a sequence as basic as the SV40 large T nuclear location signal would normally tend to be exposed at the hydrophilic surface of a protein. Even in this circumstance the signal could be masked by intra- or intermolecular interactions. For example, the putative nuclear

location signal present in poliovirus VP3 (Smith et al., 1985) may be masked by interaction with other parts of the poliovirus virion, either protein or nucleic acid. For this to be the case, it would follow that the forces favoring the intra- or intermolecular interaction would dominate those favoring movement to the nucleus. This conclusion highlights the difficulty in obtaining unequivocal data to demonstrate that a given sequence in any particular protein acts in thatprotein as a nuclear location signal. It is not sufficient to show homology with the SV40 prototype, nor is it sufficient to show nuclear location activity when appended to a cytoplasmic protein. While both these methods show potential nuclear location signal activity, they ignore the influence of protein context and possible masking. The isolation of subtle, preferably point mutations mapping to the putative nuclear location sequence is necessary to provide unequivocal evidence for the function of the domain. Few studies in the literature have been this thorough. Cooperation between Nuclear Signals We have shown here that two partially defective nuclear location signals in any one of three different spatial combinations can cooperate to enhance the nuclear accumulation of pyruvate kinase. This suggests that these signals function independently of one another, and is in agreement with our previous observation that polyomavirus large T contains two basic sequences that contribute toward its nuclear accumulation (Richardson et al., 1986). Next to nothing is known about how nuclear location signals function. The SV40 large T signal permits the nu-

Cell 472

clear entry of proteins that would otherwise be too large to accumulate in the nucleus by passive diffusion. This process exhibits a high degree of selectivity since a single point mutation at codon 128 of the signal sequence can completely abolish activity (Kalderon et al., 1984a; Lanford and Butel, 1984). Thus a highly efficient and selective cellular apparatus must be responsible for the recognition and nuclear translocation of proteins containing nuclear location signals. The likelihood that this translocation apparatus will recognize a given protein will increase if it contains not one but two or more targeting sequences. This could account for the fact that multiple relative to single copies of a partially defective nuclear location signal can enhance the nuclear accumulation of pyruvate kinase. The finding that partially defective signals can cooperate highlights a potential difficulty in interpretation in studies using multiple copies of sequences related to nuclear location signals covalently attached to cytoplasmic carrier proteins. Since in the latter studies it is difficult to control both the location and number of attached peptides, differences in context and cooperative effects might make interpretation of data difficult. Competition between Targeting Signals We have shown here that the SV40 large T nuclear signal was unable to target to the nucleus of microinjected Vero cells a middle T mutant possessing the carboxy-terminal hydrophobic domain. One explanation for this observation is that the nuclear signal could not be recognized when inserted into this particular site within middle T, just as it was unable to function when inserted into the sequences corresponding to the Sacl site of the pyruvate kinase cDNA. However, the same nuclear signal could function when inserted into the same internal site of a variant middle T lacking the hydrophobic tail. This suggests that the nuclear location signal was accessible and potentially active in the full-length middle T molecule, but that under circumstances where two different and competing intracellular signals were present, the membrane-binding signal was dominant. Only when that signal was removed could the nuclear signal function. Molecules containing two intracellular signals have previously been studied in a different context. An influenza virus hemagglutinin-SV40 large T fusion protein (Sharma et al., 1985) and the 78 kilodalton glucose-regulated protein (Munro and Pelham, 1988) both contain an aminoterminal hydrophobic signal sequence that targets the proteins to the lumen of the endoplasmic reticulum as well as a nuclear location signal. Related protein species lacking the respective amino-terminal hydrophobic signal peptides are targeted to the nucleus. In these cases, however, competition between the two signals is not direct in that the amino-terminal signal targets the nascent proteins into the lumen prior to translation of the nuclear location signal. By the time the latter signal is synthesized, the protein is already committed to enter another compartment and is unavailable for movement to the nucleus. Another example is a large Tmiddle T hybrid protein arising from a frameshift mutation in the polyoma viral genome. This protein is localized to cellular membranes in

certain transformed cells (Wilson et al., 1986), consistent with the observation that the hybrid protein contains the hydrophobic carboxy-terminal domain of middle T The hybrid protein is also predicted to contain one of the two nuclear location signal sequences of polyomavirus large T (Richardson et al., 1986). This signal sequence, including Lysine-192 of polyomavirus large T, is less potent than the SV40 large T prototypic signal studied here. Moreover, the potential contribution of this sequence to the nuclear localization of truncated large T species is not established. However, in the context of the hybrid protein, the membrane-binding domain dominates the activity of the potential nuclear signal. Multiple targeting signals appear to be required for the localization of some proteins to subcompartments of chloroplasts and mitochondria (reviewed by Colman and Robinson, 1986). These intracellular targeting signals may function sequentially rather than simultaneously to ensure correct localization. The data presented here suggest that one intracellular targeting signal may dominate another and consequently, otherwise active targeting signals may fail to function. One exciting possibility that this suggests is that a single polypeptide could be modified within the cell to remove or inactivate one signal, whereupon a second might become evident. In this way, for example, a membrane-bound receptor protein could be processed to remove a membrane-anchoring domain, after which a nuclear location signal could become active. This would enable a regulatory molecule to shuttle between the plasma membrane and the nucleus. Alternatively, an otherwise active nuclear location signal may fail to function when situated within a polypeptide that binds to a membrane-associated protein. The nuclear signal could become active upon the dissociation of the protein complex in response to cellular signals. An example of this type of phenomenon may be the rapid translocation of the catalytic subunit of CAMP-dependent protein kinase from the Golgi complex to the nucleus in response to increased intracellular CAMP levels (Nigg et al., 1985). However, it has yet to be shown that the catalytic subunit contains a nuclear location signal. We have demonstrated a direct approach toward comparing the efficacy of intracellular targeting signals within the same protein. This approach will be of value in determining the hierarchy of intracellular targeting signals within any given protein and for testing the dual signal: dual location model. Experimental

Procedures

Plasmid Constructions Oligonucleotide-Directed Mutagenesis Minimal SV40 large T nuclear location signals were mutated using the gapped heteroduplex method essentially as described by Kalderon et al. (1984a). A synthetic oligonucleotide (CTTCTACCTTGCGCTTCGTTTTTG) partly complementary to nucleotides 4441 to 4419 of SV40 DNA was used to convert the Lysine-129 equivalent codon to a Threonine codon while creating a novel HinPl site at nucleotide 4429. Mutant plasmids were characterized by restriction enzyme mapping using HinPl and were sequenced using the oligonucleotide primed dideoxy method. Plasmids XR35-PK and XR45-PK are predicted to encode proteins that are identical to those encoded by pXR30-PK and pXR40PK, respectively (Kalderon et al., 1994b), with the exception of the sub-

The Protein 473

Context

of Nuclear

Table 1. Linker Insertion Pyruvate Kinase cDNA

Mutants

Signals

of the Chicken

Muscle

Plasmid

Lanker

5’ Nucleotide

3’ Nucleotide

BseRlO BseR12 stuFt10 slux10 SacX8 SacR.12 BSXXB BsxR12

CGGAATTCCG CCGGAATTCCGG CGGAATTCCG GGCTCGAGCC CCTCGAGG CCGGAATTCCGG CCTCGAGG CCGGAATTCCGG

964 964 1332 1332 457

980 980 1333 1333 489

1651 1846

The sequences of EcoRl and Xhol linkers inserted into four restriction enzyme sites withrn the PK cDNA are shown. The first nucleotide of the pyruvate kinase cDNA on either the 5’ or 3’side of the linker is also shown (numbering system of Lonberg and Gilbert, 1983). Linkers were inserted into the unique BstEll site of plasmid RL142PK8F (Richardson et al., 1986) following filling in with E. coli DNA polymerase (Klenow fragment) rn the presence of dNTPs to generate pBseRl0 and pBseR12. Similarly, linkers were inserted into the unique Stul site of plasmid RL18PK8 (Kalderon et al., 1984b) to generate plasmids StuRlO and StuXlO. Lrnkers were also inserted into the unique Sacl and BstXI sites of RL142PK8F following removal of overhanging B’ends by T4 DNA polymerase in the presence of dNTPs and filling in with Klenow to qenerate SacX8, SacR12, BsxX8. and BsxRl2, respectively.

stitution of Threonine for Lysme at the 128 equivalent position of the SV40 large T nuclear location signal. Generation of Linker Insertion Mutants The chicken muscle pyruvate kinase cDNA mutated in this study was contained within plasmid RL142PK8F. described previously (Richardson et al., 1986). Table 1 lists the linker insertion mutants generated In our study. An Xhol linker (CCTCGAGG) was inserted into each of the EcoRl sites of plasmids BseRlO and BseRl2 to create plasmids BseRlOX and BseRlPX, respectively. Similarly, an Xhol linker of 12 base parrs (GGGCTCGAGCCC) was inserted into the EcoRl site of pBsxRl2 to create plasmid BsxR12X12. The predicted structures of these plasmids were confirmed by DNA sequencing and extensive restrIctron enzyme analysis. Generation of Signal insertion Mutants The XR30 signal (Kalderon et al., 1984b) was inserted into the Sacl site of the PK cDNA by combining the large Xhol-BamHI fragment of pSacX8. the small EcoRI-BamHI fragment of pSacR12, and the small Xhol-EcoRI XR30 DNA fragment in a three-fragment ligation. The ligation product, pSacXR3OA, was opened at the unique EcoRl site, filled In with Klenow, and self-ligated in the presence of excess Smal linkers (CCCGGG). Plasmid SacXRBOB contained a single copy of a Smal lanker inserted into the EcoRl site of pSacXRSOA, as determined by DNA sequencing. Plasmid SacXR35B is identical to pSacXR30B, except that a small Xhol-EcoRI fragment encoding the XR35 signal was substituted for the XR30 signal. Signals were inserted into the BstEll site of the PK cDNA as follows. The XR30 srgnal was kgated together with the large Xhol-BamHI of pBseR12X and the small EcoRI-BamHI fragment of pBseRl0 to create pBseXR30. Plasmid BseXR35 IS Identical to pBseXR30, except that a small Xhol-EcoRI fragment encoding the XR35 signal was substituted for the XR30 signal. Similarly, the XR12 signal (Kalderon et al., 1984b) was inserted separately into the BstEll site by combination with the large Xhol-BamHI fragment of pBseRlOX and the small EcoRl-BamHI fragment of pBseRl0 to create plasmid BseXR12. Plasmids StuXR30A and StuXR35A were created by ligating together the large Xhol-BamHI fragment of pStuX10, the small EcoRIBamHl fragment of pStuR10. and the small Xhol-EcoRI fragment encoding either the XR30 or XR35 signals, respectively. The products of these two ligations were subcloned into an appropriate expression vector by separately ligatmg the large Sacl-BamHI fragment of pRL142PK8F to each of the small Sacl-BamHI fragments of pStuXR3OA and pStuXR35A to create plasmids StuXR30B and StuXR35B, respectively. The XR30 and XR35 srgnals were separately ligated to the large

Xhol-BamHI fragment of pBsxX8 and the small EcoRI-BamHI of pBsxRl2 to generate pBsxXR3OA and pBsxXR35A, respectively. Similarly, the XR12 signal was ligated together with the large Xhol-BamHI fragment of pBsxR12XiP and the small EcoRI-BamHI fragment of pBsxRl2 to generate plasmid BsxXRlPA. An m-frame termination codon was introduced into the immediate 3’side of the signal insertion sites by cutting with EcoRI, filling in with Klenow, and self-ligating to generate plasmids BsxXR30B, BsxXRSBB, and BsxXRlPB. The predicted structures of the above signal insertion mutants were confirmed by extensive restriction enzyme analysis, sizing of small DNA fragments on NuSieve agarose gels, and DNA sequencing. In every case, only one copy of a small Xhol-EcoRI fragment encoding a signal sequence was inserted into defined sites within the PK cDNA. Combination of Signal fnsertion Mutants Pyruvate kinase mutants containing two copies of the XRt2 signal were created as follows. The large Sacl-BamHI fragment of pXRlP-PK was ligated to the small Sacl-BamHI fragments of either pBseXR12 or pBsxXR12B to generate plasmids XR12[1+2] and XR12[1+3], respectively. Similarly, the large Kpnl-BamHI fraqment of pBseXR12 was ligated to the small Kpnl-BamHI fragment of pBsxXR12B to generate plasmid XR12[2+3]. A pyruvate kinase mutant containing three copies of the XR12 signal was created by ligating the large Sacl-BamHI fragment of pXRIP-PK to the small Sacl-BamHI fragment of pXR12[2+3)] to generate pXR12[1+2+3]. The predicted structuresof these plasmids were confirmed by extensive restriction enzyme analysis. fnsertion of Nuclear Signals into Pofyomavirus Middle T Small DNA fragments encoding either a wild-type (XR40) or defective (XR45) SV40 large T nuclear location signal sequence were inserted into the coding sequence of polyomavirus middle T to create plasmids SVXD57-XR40-101 and SVXD57-XR45-101, as shown in Figure 9. Similarly, signal sequences were inserted into the coding sequence of a carboxy-terminal truncation mutant of middle T (RXT; Markland et al., 1986) to generate SVXD57-XR40-RXT and SVXD57-XR45-RXT. Subcellular Localization of Proteins Plasmid microinjection into Vero cells and indirect immunofluorescence were performed as described previously (Kalderon et al., 1984b). The specificity of rabbit anti-chicken muscle type Ml pyruvate kinase polyclonal antibody (Lonberg and Gilbert, 1983) has been demonstrated previously(Kalderon et al., 1984b). Polyomavirus middle T species were detected using a rabbit anti-peptide antibody (anticarboxy peptide C) prevrously characterized by Harvey et al. (1984). RITC-conjugated goat anti-rabbit IgG was obtained from Cooper Biomedical (Malvern, Pennsylvania). Generation of Cell Lines Expressing Pyruvate Kinase Mutants Plasmid 82-2 (Hsiung et al., 1984). which contains the entire genome of bovine papilloma virus type 1 cloned between the Sall and BamHl sites of pBR322, was cut with Hindlll, filled m with Klenow. and selfligated with T4 DNA ligase. The large Sall-BamHI fragment of the resultant plasmid B2-2F was ligated to the large Sall-BamHI fragment of plasmid RL66 (Kalderon and Smith, 1984) to generate pRL66BPV. The large Hindlll-BamHI fragment of pRL66BPV was ligated to each of the small Hindlll-BamHI fragments encoding the PK signal insertion mutants. BPV shuttle vectors were transfected onto mouse Cl27 cells as described by Hsiung et al. (1984). Two weeks later, foci of transformed cells were picked, expanded, and analyzed for the presence of pyruvate kinase proteins by indirect immunofluorescence. Isotopic labeling of cells, immunoprectpitation with antt-pyruvate kinase antibody, and SDS-polyacrylamide gel electrophoresis were carried out essentrally according to methods described previously (Paucha et al.. 1984). Polyomavrrus mrddle T mutants were assayed for their ability to transform by focus formation on a monolayer of Rat-l cells (Markland et al., 1986). For those middle T mutants that were able to transform, foci were picked, expanded, and analyzed for the expression of middle T proteins by indirect immunofluorescence Acknowledgments We thank Robert Harvey for technical assistance, Doug Lovern for the synthesis of oligonucleotides, and William Markland for cntical reading

Cell 474

0 p101

1 Cut with EmRl 2 Fill in with Klercw 3 Ligate to Bglll linkers (GAAGATATCTTC)

1 XR40SIGNAL(TPPKKKRKVEDP) 1

Pvull

XR45SIGNAL(TPPKTKRKVEDP) MIDDLE T CODING SEQUENCE

0

Pvull

PYRUVATE KINASE CODING SEQUENCE

bglll 1 Cut with Pvull

2 Ligate to EccRl linkers (GGAATTCC) 3 Cut with EmRI/Bglll 4 ISolate 416 bp fragment Ligate to the large EcoRllBglll fragment of pSacXR40

Ligate to large EcoRllBglll fragment of pSacXR45

Xhol EcoRl

1 Cut with Xhol/BamHI

1 Cut with Xhol/BamHI 2 Isolate 600bp fragment

2 Isolate 6oObp fragment 3 Ligate to large XhoVBamHl fragment of pSVXD57-PK12X

3 Ligate to large Xhol/BamHI fragment of pSVXD57-PK12X

1 Bglll

Boll1

pSVXD57. XR45-101

Xhol EcoRl

Xhol

EC&

0

Band+ 8g11

Figure

9. Insertion

of Nuclear

Signals

into Polyomavirus

Middle

T

Small DNAfragments encoding either a wild-type (XR40) or defective (XR45) SV40 large T nuclear signal were inserted into a plasmid (~101; Markland et al., 1986) encoding wild-type polyomavirus middle T as shown here. Plasmids SacXR40 and SacXR45 are identical to plasmids SacXRdOA and SacXR35A, respectively, except that the XR40 and XR45 nuclear signals are substituted for the XR30 and XR35 signals, respectively. Similarly, nuclear signals were inserted into the coding sequence of the carboxy-terminal truncation mutant RXT (Markland et al., 1986) to generate pSVXD57XR40-RXT and pSVXD57-XR45-RXT.

of the manuscript. We also thank Daniel Kalderon for stimulating discussions in the early stages of this work, and Scott Decker for assistance in the preparation of laser-printed figures. This work was supported in part by a fellowship from Integrated Genetics to B. L. R. and a grant from the National Cancer Institute (ROI CA 43186-01). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Received

March

30, 1987; revised

May 26, 1987.

Colman, A., and Robinson, C. (1986). Protein import hierarchical targeting signals. Cell 46, 321-322.

Dilworth, S. M., Hansson, H.-A., Darnfors, C., Bjursell, G., Streulr, C. H., and Griffin, B. E. (1986). Subcellular localisation of the middle and large T-antigens of polyoma virus. EMBO J. 5, 491-499. Ding, D., Dilworth, S. M., and Griffin, 8. E. (1982). Mlt mutants oma virus. J. Virology 44, 1080-1083.

of poly-

Dingwall, C., Sharnick, S. V., and Laskey, R. A. (1982). A polypeptide domain that specifies migration of nucleoplasmin into the nucleus. Cell 30, 449-458. Goldfarb, Synthetic

References

into organelles:

D. S., Gariepy, J., Schoolnik, G., and Kornberg, R. D. (1986). peptides as nuclear localization signals. Nature 322, 641-

644.

Blobel, G. (1980). Intracellular Sci. USA 77, 14961500.

protein

topogenesis.

Carmichael, G. G., Schaffhausen, B. S., Dorsky, D. I., Benjamin, T. L. (1982). Carboxyl terminus of polyoma mor antigen is required for attachment to membranes, tein kinase activities and cell transformation. Proc. USA 79, 3579-3583.

Proc. Natl. Acad. Oliver, D. B., and middle-sized tuassociated proNatl. Acad. Sci.

Harvey, R., Oostra, 8. A.. Belsham, G. J., Gillett, P., and Smith, A. E. (1984). An antibody to a synthetic peptide recognizes polyomavirus middle T anhgen and reveals multiple in vitro tyrosine phosphorylation sites. Mol. Cell. Biol. 4, 1334-1342. Hsiung, N., Fitts, R., Wilson, S., Milne. A., and Hamer, D. (1984). Efficient production of hepatitis B surface antigen using a bovine papilloma viru-metallothionein vector. J. Mol. Appl. Genet. 2. 497-506.

The Protein 475

Context

of Nuclear

Signals

Kalderon, D., and Smith, A. E. (1984). In vitro mutagenesis of a putative DNA binding domain of SV40 large T. Virology 139, 109-137. Kalderon, D., Richardson, W. D., Markham, A. F., and Smith, A. E. (1984a). Sequence requirements for nuclear localisation of SV40 largeT antigen. Nature 377, 33-38 Kalderon, D., Roberts, B. L., Richardson, W. D., and Smith, A. E. (1984b). A short amino acid sequence able to specify nuclear location. Cell 39, 499-509. Lanford, R. E., and Butel, J. S. (1984). Construction tion of an SV40 mutant defective in nuclear transport 37, 801-813.

and characterizaof T antigen. Cell

Lanford, R. E., Kanda, t?, and Kennedy, R. C. (1986). Induction of nuclear transport with a synthetic peptide homologous to the SV40 T antigen transport signal. Cell 46, 575-582. Lonberg, N., and Gilbert, W. (1983). Primary structure of chicken muscle pyruvate kinase mRNA. Proc. Natl. Acad. Sci. USA 80, 3661-3665. Lonberg, N., and Gilbert. W. (1985). Intron/exon chicken pyruvate kinase gene. Cell 40, 81-90.

structure

of the

Markland, W., Cheng, S. H., Oostra, 8. A., and Smith, A. E. (1986). In vitro mutagenesrs of the putative membrane-binding domain of polyomavirus middle-T anttgen. J Virology 59, 82-89. Muirhead, H., Clayden. D. A., Barford, D.. Lorimer, C. G., FothergillGilmore, L. A., Schiltz, E., and Schmitt, W. (1986). The structure of cat muscle pyruvate kinase. EMBO J. 5, 475-481. Munro, S., and Pelham, H. R. B. (1986). An Hsp70-like protein rn the ER: identity with the 78 kd glucose-regulated protein and immunoglobulin heavy chain binding protein. Cell 46, 291-300. Ntgg, E. A., Hilz, H., Eppenberger, H. M., and Dutly, F, (1985). Rapid and reversible translocation of the catalytic subunit of CAMP-dependent protein kinase type II from the Golgi complex to the nucleus. EMBO J. 4, 2801-2806. Nilsson, S. V., Tyndall, C., and Magnusson, G. (1983). Deletion mapprng of a short polyoma vrrus middle-T segment important for transformation. J. Virology 46, 284-287. Novak, U.. and Griffin, B. E. (1981). Requirement forthe C-terminal regron of middle-T antigen in cellular transformation by polyoma virus. Nucl. Acids Res. 9, 2055-2073. Paucha, E., Harvey, R., and Smith, A. E. (1984). of some forms of stmran virus 40 large T antigen thetrc peptides. J. Virology 57, 670-681.

lmmunoprecipitahon by antibodies to syn-

Richardson, W. D., Roberts, B. L., and Smith, A. E. (1986). cation srgnals In polyoma vrrus large-T Cell 44, 77-85

Nuclear

lo-

Roberts, B. L., Richardson, W. D., Kalderon, 0. D., Cheng, S. H., Markland, W., and Smith, A. E. (1986). Amino acid sequences able to confer nuclear location. In Nucleocytoplasmic Transport, R. Peters and M. Trendelenburg, eds. (Berlin: Springer-Verlag), pp. 185-198. Sharma, S., Rodgers, L., Brandsma, J., Gething, M.-J., and Sambrook, J. (1985) SV40 T antigen and the exocytotic pathway. EMBO J. 4, 1479-1489 Smith. A. E., Kalderon, D., Roberts, B. L., Colledge, W. D., Edge, M., Gillett. P., Markham, A. F., Paucha, E., and Richardson, W. D. (1985). The nuclear locatton signal. Proc. Trans. Roy. Sot. Lond. B 226, 43-58. Templeton, D., and Eckhart, W. (1982). Mutation causing prematuretermination of polyoma virus mIddIeT antigen blocks cell transformation, J Vrrology 47, 1014-1024. Watson, M E. E. (1984) Comprlation Nucl Acids Res. 72. 5145-5164

of publrshed

signal sequences.

Wilson. J. B, Hayday, A., Courtnerdge, S., and Frred, M. (1986). A frameshrft at a mutattonal hotspot in the polyoma virus early regron generates two new proteins that define T-antigen functional domains. Cell 44. 477-487. Wychowski, C., Benrchou, D., and Girard, M. (1986). A domain of SV40 capsid polypeptide VP1 that specrfies migratron into the cell nucleus. EMBO J 5. 25692576. Zhu. Z Veldman, G. M., Cowre, A., Carr, A., Schaffhausen, B., and Kamen. R. (1984). Constructron and functional characterrzatron of polyomavrrus genomes that separately encode the three early proteins. J. /irology 51, 170-180