(Received for publication, August 19, 1996)
From the Women's Cancers Section, Laboratory of Pathology, Division of Clinical Sciences, NCI, National Institutes of Health, Bethesda, Maryland 20892
We previously compared the structure and motility suppressive capacity of nm23-H1 by transfection of wild type and site-directed mutant forms into breast carcinoma cells. Wild type nm23-H1 and an nm23-H1S44A (serine 44 to alanine) mutant suppressed motility, whereas the nm23-H1P96S, nm23-H1S120G, and to a lesser extent, nm23-H1S120A mutant forms failed to do so. In the present study wild type and mutant recombinant Nm23-H1 proteins have been produced, purified, and assayed for phosphorylation and phosphotransfer activities. We report the first association of Nm23-H1 mutations lacking motility suppressive capacity with decreased in vitro activity in histidine-dependent protein phosphotransferase assays. Nm23-H1P96S, a Drosophila developmental mutation homolog, exhibited normal autophosphorylation and nucleoside-diphosphate kinase (NDPK) characteristics but deficient phosphotransfer activity in three histidine protein kinase assays, using succinic thiokinase, Nm23-H2, and GST-Nm23-H1 as substrates. Nm23-H1S120G, found in advanced human neuroblastomas, exhibited deficient activity in several histidine-dependent protein phosphotransfer reactions, including histidine autophosphorylation, downstream phosphorylation on serines, and slightly decreased histidine protein kinase activity; significant NDPK activity was observed. The Nm23-H1S120A mutant was deficient in only histidine-dependent serine autophosphorylation. Nm23-H1 and Nm23-H1S44A exhibited normal activity in all assays conducted. Based on this correlation, we hypothesize that a histidine-dependent protein phosphotransfer activity of Nm23-H1 may be responsible for its biological suppressive effects.
The nm23 family of genes has been characterized on the basis of its reduced expression in certain highly metastatic cell lines (1) and tumors (reviewed in Ref. 2). In human tissues, three members have been identified, nm23-H1 (3), nm23-H2 (4), and nm23-DR (5). Transfection of nm23 cDNAs into murine K-1735 TK (6, 7), B16F10 (8), or B16FE7 (9) melanoma cells, rat MTLn3 mammary adenocarcinoma cells (10), or human MDA-MB-435 breast carcinoma cells (7, 11, 12) resulted in a significant reduction in tumor metastatic potential in vivo, providing evidence that nm23 can have a metastasis suppressor function.
Nm23 proteins also participate in the development and differentiation process. The most thoroughly studied model system is Drosophila, in which reduced expression of the abnormal wing discs (awd)1 gene, which encodes a protein approximately 77% identical to Nm23, causes widespread aberrant differentiation post-metamorphosis leading to lethality (3, 13, 14). Additionally, mutation of proline 96, the killer-of-prune mutation (awd k-pn), causes developmental abnormalities when expressed in the context of prune (pn) eye color mutations (reviewed in Ref. 15). Associations of nm23 and development have been reported in higher organisms. In murine embryogenesis, elevated Nm23 expression has been correlated with the functional differentiation of most epithelial tissues (16). Human MDA-MB 435 breast carcinoma cells transfected with nm23-H1 showed morphological and biosynthetic evidence of differentiation in vitro (12). The phenotypic similarities of the tumor metastatic and developmental processes have been long recognized, leading to the hypothesis that they may involve similar regulatory factors, such as nm23.
A long list of known and postulated biochemical activities and protein:protein associations have been reported for Nm23, which could serve as candidate biochemical mechanisms for its metastasis suppressive effects. Nm23 proteins exhibit nucleotide-diphosphate kinase (NDPK) activity (EC 2.7.4.6), a nonspecific phosphotransferase that mediates the transfer of a terminal phosphate from a NTP to a NDP via a Nm23-phosphohistidine 118 intermediate (17-23). Nm23 proteins have been reported to associate with microtubules (17, 24-26) and small or heterotrimeric G proteins (27-33), and its NDPK activity was postulated to directly transphosphorylate or indirectly modulate nucleotide pool sizes to affect their activity; however, each of these hypotheses remains controversial. In Drosophila, the developmental effects of awd can be dissociated from its NDPK activity: (a) Awdk-pn protein retained significant NDPK activity (34); (b) transformation of the null germ line with human nm23 cDNAs restored NDPK activity but not the fully differentiated phenotype (35). In transfection model systems, neither total nor subcellular NDPK activities correlated with Nm23 expression and suppression of metastasis (36).
Nm23 proteins have recently been reported to exhibit histidine protein kinase activity (37), presumably using the same histidine 118 phosphorylated intermediate involved in the NDPK activity. Histidine protein kinases form the "two component" signal transduction system in bacteria (reviewed in Ref. 38) and have recently been described in eukaryotic systems, including human platelet activation (39, 40).
A lower energy serine autophosphorylation of Nm23 was described in vitro and in vivo (41-44); levels of Nm23-phosphoserine were directly correlated with Nm23 expression and metastasis suppression in model systems (41). Recombinant Nm23-H1 exhibited autophosphorylation on two proteolytic fragments containing serine 44 and serines 120, 122, and 125, of which position 120 remains the most conserved evolutionarily (41).
Nm23-H2 has been identified as a transcription factor for the PuF site/CT element found in the c-myc promoter (45-48). Other groups have found nonspecific binding of Nm23-H2 to single-stranded polypyrimidines and RNA (49), and we have found no evidence in vivo for a direct transcriptional activity.2
We recently reported a different strategy to investigate the functional biochemical activity of Nm23 in metastasis suppression using site-directed mutagenesis (51). Human MDA-MB-435 breast carcinoma cells were transfected with wild type or mutated forms of nm23-H1, and their phenotype was examined using an in vitro assay for one component of the metastatic process, motility in Boyden chamber assays. Tumor cell motility to either fetal calf serum or partially purified autotaxin (ATX) was reduced by transfection with wild type or serine 44 mutated forms of nm23-H1 but not by transfected nm23-H1 cDNAs mutated at either serine 120 or at the homolog of the killer-of-prune position (proline 96).
The present report describes in vitro biochemical activities of recombinant Nm23-H1 wild type and site-directed mutated proteins with and without motility suppressive capacity. We report a stunning correlation in that those mutant proteins incapable of motility suppression exhibited reduced activity in biochemical assays of histidine-dependent protein phosphotransferases in vitro. The data permit the development of the hypothesis that a histidine-dependent protein phosphotransferase activity of Nm23-H1 may mediate its biological motility suppressive activity.
Site-directed
mutagenesis of nm23-H1 was carried out by PCR as described
previously (51), using the oligonucleotide pET5 (5
-GGGGGGATCCATATGGCCAACTGTGAG-3
) instead of CMV5
. The PCR products
were double-digested with NdeI and BamHI,
gel-purified, and subcloned into the pET3c expression vector (52). The
resulting plasmids were analyzed by double-stranded sequencing and
transformed into Escherichia coli strain BL21(DE3).
Transformed BL21(DE3) bacteria were grown overnight in LB broth
containing 100 µg/ml AMP at 37 °C, diluted 1/20 in the same
medium, and grown under the same conditions until the culture reached
an A600 of 1.0. Expression of the
nm23 cDNAs was induced by addition of 1 mM
isopropyl-1-thio-
-D-galactopyranoside. After a 2-h
induction, bacteria were collected by centrifugation, washed twice with
1/5 volume of TBS (20 mM Tris-HCl, pH 8.0, 150 mM NaCl), resuspended in 1/50 volume of the same buffer,
and disrupted by sonication. The bacterial lysates were centrifuged at
35,000 × g for 20 min at 4 °C, and the soluble
extracts were stored at
20 °C. Alternatively, in the preparation
of bacterial lysates for HPLC purification, the above buffer was
replaced with 20 mM Tris-HCl, pH 8.0, 100 mM
NaCl (rNm23-H1) or 20 mM Tris-HCl, pH 8.3 (rNm23-H2).
Lysates from
bacteria expressing wild type or site-directed mutant rNm23-H1 proteins
were loaded in a DEAE-Sepharose CL-6B column (Pharmacia Biotech Inc.),
equilibrated with 20 mM Tris-HCl, pH 8.0, 100 mM NaCl (0.5-ml bed volume/ml lysate). The column was
washed with 20 column volumes of the same buffer and finally eluted
with 20 mM Tris-HCl, pH 8.0, 200 mM NaCl.
rNm23-H1S44A was eluted using 250 mM NaCl
instead of 200 mM. The fractions from the eluate containing
significant amounts of protein were pooled and dialyzed against 50 mM sodium acetate buffer, pH 5.0, and applied to a Mono S
HR 5/5 (Pharmacia), equilibrated with the same buffer. After washing
the column with 10 volumes of the starting buffer, proteins were eluted
with a linear gradient (20 ml) of 0-1 M NaCl in the same
buffer at a flow rate of 1 ml/min. 0.5-ml fractions were collected,
analyzed by SDS-PAGE, and those containing rNm23 proteins were pooled,
concentrated, and applied to a Superdex 75 HR 10/30 size exclusion
column (Pharmacia), equilibrated, and eluted with 20 mM
Tris-HCl, pH 8.0, 150 mM NaCl at a flow rate of 0.5 ml/min.
Fractions of 0.25 ml were collected, and those containing pure
recombinant proteins, as judged by silver-stained SDS-PAGE, were stored
at 4 °C or at 20 °C for long-term storage. HPLC purification of
rNm23-H2 was carried out as described previously (53). Both cation
exchange and size exclusion chromatography were performed using a
Dionex DX 500 chromatography system, equipped with a GP40 gradient pump
and an AD20 absorbance detector (Dionex Corp., Sunnyvale, CA). The
absorbance of the eluate was monitored at 280 nm in the sensitivity
range of 2.0.
Bacterial lysates containing wild type or mutant
rNm23 proteins were loaded into Cibacron Blue 3GA-agarose columns, type
1000 (Sigma) (0.25-ml column volume/ml lysate), equilibrated with 20 mM Tris-HCl, pH 8.0, 150 mM NaCl. After washing
the columns with 20 column volumes of the same buffer, Nm23 proteins
were eluted with 20 mM Tris-HCl, pH 8.0, 10 mM
ADP, 1 M NaCl. Eluate containing affinity purified rNm23-H1
proteins was diluted 10-fold with 20 mM Tris-HCl, pH 8.0, and applied to a DEAE-Sepharose CL-6B column (Pharmacia), equilibrated
with 20 mM Tris-HCl, pH 8.0, 100 mM NaCl
(0.1-ml bed volume/ml lysate), washed, and eluted as indicated above.
Fractions containing purified recombinant Nm23-H1 proteins were stored
at 4 °C or at 20 °C. Affinity purified rNm23-H2 was dialyzed
against 20 mM Tris-HCl, pH 8.3, and loaded onto a
DEAE-Sepharose column equilibrated with the same buffer. Fractions of
the flow-through containing purified rNm23-H2 protein were concentrated
and stored as indicated above.
Two oligonucleotides containing EcoRI
restriction sites were synthesized (5Nm23-H1+RI,
5
-CCCGAATTCATGGCCAACTGTGAGCG-3
; 3
Nm23-H1+RI,
5
-GCCGAATTCATTCATAGATCCAGTTCTG-3
). These two primers were used to
amplify the Nm23-H1 coding sequence, using 1 ng of pNm23-H1 (3) as
template. The PCR reaction mixture, overlaid with 75 µl of mineral
oil to prevent evaporation, was heated at 94 °C for 3 min and then
35 cycles were run as follows: 94 °C, 1 min; 55 °C, 1 min,
72 °C, 3 min. The PCR product was digested with EcoRI,
and the resulting 459-base pair fragment was subcloned into the
EcoRI site of pGEX-2T(128/129) (gift of Dr. M. A. Blanar, Hormone Research Institute, UCSF). The recombinant plasmid containing the insert in the correct orientation, named pGEX-H1, was analyzed by
double-stranded sequencing and transformed into E. coli
DH5
bacteria. Lysates from pGEX-H1-transformed bacteria were
prepared as indicated above for BL21(DE3) cells. Recombinant
GST-Nm23-H1 fusion protein was purified from the lysates by affinity
chromatography on glutathione-Sepharose 4B columns (Pharmacia Biotech,
Uppsala, Sweden), following the manufacturer's instructions. The
glutathione used in the elution of the fusion protein was removed by
gel filtration on PD-10 columns (Pharmacia) equilibrated and eluted
with 20 mM Tris, pH 8.0, 100 mM NaCl.
20 µg of purified
rNm23 proteins were incubated with 100 µM
[-32P]ATP (500 µCi) for 15 min at room temperature
in 100 µl of TMD buffer (20 mM Tris-HCl, pH 8.0, 5 mM MgCl2, 1 mM dithiothreitol). Free [
-32P]ATP was removed by successive rounds of
dilution with 20 mM Tris-HCl, pH 8.0, 100 mM
NaCl, and concentration with Centricon-30 tubes (Amicon Inc., Beverly,
MA 01915) until concentration of [
-32P]ATP, estimated
by counting of the filtrate, was less than 5 nM
(approximately 0.1% of the total counts in the phosphorylated proteins). The absence of detectable amounts of free
[
-32P]ATP was checked by thin layer
chromatography.
NDPK activity of
purified rNm23 proteins was assayed by thin layer chromatography (53).
Briefly, 50 ng of purified rNm23 proteins were incubated with 1 µM [-32P]ATP (0.2 mCi/ml) and 100 µM GDP in TMD buffer (20 mM Tris-HCl, pH 8.0, 5 mM MgCl2, 1 mM dithiothreitol) in
a 10-µl total volume at room temperature for 10 min. The reaction was
stopped by addition of 10 µl of 50 mM EDTA, and 2 µl
were applied to a 20 × 20 cm PEI-cellulose TLC plate (J. T. Baker
Inc.) and resolved by capillary action with 0.75 M
KH2PO4, pH 3.65. When the chromatography was complete, the plate was dried and exposed to autoradiography.
106 cpm of purified 32P-rNm23 proteins (approximately 200 ng) were incubated with 2 µg of rNm23-H2 or GST-Nm23-H1 or 4 µg of porcine heart succinic thiokinase (Sigma, S-8420) for 10 min at room temperature in 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM MgCl2, 1 mM dithiothreitol in a final volume of 30 µl. The reaction was stopped by adding 10 µl of SDS sample buffer, pH 8.5, and the samples (20 µl) were loaded without boiling in 13% SDS-PAGE minigels. After electrophoresis, the gels were exposed to autoradiography without any further treatment. Since rNm23-H1H118F did not undergo histidine autophosphorylation, the amount of this protein used in the assays was normalized by protein concentration rather than scintillation counting.
Phosphoamino Acid Analysis4 µg of 32P-rNm23-H1 (approximately 106 cpm/µg) was incubated with 40 µg of succinic thiokinase in a final volume of 50 µl of TMD buffer for 30 min at room temperature. The phosphorylated proteins were resolved in a 13% SDS-PAGE minigel and electrophoretically transferred to a PVDF membrane (Immobilon-P, Millipore Co., Bedford, MA). The 32P-rNm23-H1 and 32P-STK bands were detected by autoradiography, and the corresponding PVDF fragments were excised, rewet with methanol, and rinsed twice with water. Alkaline hydrolysis of the phosphorylated proteins was carried out by placing the PVDF slices in 100 µl of 3 N KOH for 100 min at 120 °C in screw cap microcentrifuge tubes. After hydrolysis, potassium ions were precipitated by slow addition of 10% perchloric acid until a neutral pH was approached and removed by centrifugation (54). The supernatant was spotted onto a 20 × 20 cm silica gel thin layer chromatography plate (J. T. Baker, Inc.) along with phosphoamino acid standards and developed in t-butyl alcohol/methyl ethyl ketone/acetone/methanol/H2O/NH4OH (10:20:20:5:40:5, v/v) (54). Phosphotyrosine, phosphothreonine, and phosphoserine were purchased from Sigma. Phosphohistidine and phospholysine were synthesized from polyhistidine and polylysine (55). Phosphoamino acid standards were visualized by spraying the plate with 0.25% ninhydrin in n-butyl alcohol. 32P-Labeled phosphoamino acids were detected by autoradiography.
Wild
type nm23-H1 and nm23-H2 cDNAs were cloned
into a pET3c bacterial expression vector. Site-directed mutants of
nm23-H1, previously described (51) and listed in Table
I, were cloned into the same vector. Recombinant wild
type and mutated Nm23 proteins produced by
isopropyl-1-thio--D-galactopyranoside-induced BL21(DE3) E. coli transformed with the pET3c constructs were purified
by two protocols. Protocol 1 involved sequential DEAE-Sepharose, Mono S
cationic exchange, and Superdex 75 gel filtration chromatography steps,
resulting in >95% purity. Protocol 2 used a 3GA Cibacron Blue agarose
column and DEAE-Sepharose chromatography, with essentially equivalent
results. Fig. 1A shows a Coomassie
Blue-stained gel of representative preparations. As shown, rNm23-H1 and
its site-directed mutant forms migrated as a single band of
approximately 20,000, while rNm23-H2 exhibited a molecular weight of
approximately 18,000. Silver-stained gels of each protein purification
were devoid of detectable contaminating bands (data not shown).
|
Also listed in Table I is a synopsis of the biological relevance of each nm23-H1 site-directed mutant to motility suppression upon transfection into human MDA-MB-435 breast carcinoma cells, previously reported (51). Representative motility experiments using fetal calf serum or partially purified autotaxin (ATX) as chemoattractants are shown as indicators of relative motile behavior. Wild type nm23-H1 consistently inhibited motility, as compared with control transfectants (empty constructs); the nm23-H1S44A (serine 44 to alanine) transfectants exhibited variable levels of motility that were never significantly different from those of wild type. Both the nm23-H1P96S (proline 96 to serine, the k-pn homolog) and nm23-H1S120G (serine 120 to glycine, neuroblastoma mutation) transfectants consistently moved toward chemoattractants at levels at or higher than the control constructs. Another mutation at serine 120, nm23-H1S120A (serine 120 to alanine) which maintained the position's "R" group and hydrogen bonding characteristics, exhibited deficient motility to serum in the 8-h assays shown in Table I but was motile at or above the levels of the control transfectants in both longer assays to serum and in all ATX assays. The motility suppressive characteristics of the nm23-H1H118F (histidine 118 to phenylalanine) construct could not be determined with confidence due to poor expression characteristics.
Autophosphorylation CharacteristicsApproximately 2 µg of
each rNm23 protein was incubated with [-32P]ATP for 10 min. Half of the sample was electrophoresed in a basic sample buffer
without boiling, and the gel was subsequently exposed for
autoradiography without further fixation to observe histidine
autophosphorylation levels. The remaining half of the sample was boiled
in traditional sample buffer and electrophoresed, and the gel was
subsequently fixed in 20% (w/v) trichloroacetic acid,
Coomassie-stained and destained in methanol/acetic acid to permit
autoradiography of serine autophosphorylation. A radioautograph of
histidine autophosphorylation of the rNm23 proteins is shown in Fig.
1B. The Nm23-H1H118F site-directed mutant
protein, lacking the histidine previously shown to form the
phosphorylated intermediate of the Nm23 NDPK activity, remained
unphosphorylated. This protein thus provided a control for the absence
of significant contamination by wild type bacterial Nm23, which should
autophosphorylate. Histidine autophosphorylation levels exhibited by
the other rNm23 proteins were relatively uniform, with the exception of
Nm23-H1S120G. Experiments using three preparations showed a
2-5-fold decrease in histidine autophosphorylation levels, as compared
with wild type, determined by densitometry of autoradiographs. We have
noted that, upon storage at 4 °C, the S120G mutant rNm23-H1 lost
histidine autophosphorylation capacity at a greater rate than wild type Nm23 (data not shown), suggesting a defect in stability. Fig. 2 shows a comparison of wild type and freshly purified
rNm23-H1S120G histidine autophosphorylation at varying
concentrations of [
-32P]ATP. Autophosphorylation
levels for the S120G mutant failed to equal those of the wild type
protein at any [
-32P]ATP concentration tested.
Acid-stable serine autophosphorylation of Nm23 proteins was demonstrated in vitro (41-44) and upon [32P]orthophosphate labeling of MCF-7 breast carcinoma cells (41), although it represents only a minor proportion of total phosphorylation (44, 56). Serine autophosphorylation levels among the rNm23 proteins are shown on Fig. 1C. While we previously reported that purified, recombinant murine Nm23-1 protein, in which the histidine 118 was mutated to a glycine, was capable of serine autophosphorylation, the corresponding human protein, Nm23-H1H118F is not. Thus, serine phosphorylation in this system likely represents a transfer from phosphohistidine. Among the other wild type and mutant rNm23 proteins, serine autophosphorylation levels were quantitatively decreased in the S120G (1.5-2-fold) and S120A (2-3-fold) mutants, determined by densitometry of radioautographs representing three independent experiments. That serine autophosphorylation remained on both of these proteins confirms the existence of multiple serine autophosphorylation sites.
NDPK ActivityNm23 proteins and homologs identified in many
species possess a nonspecific NDP phosphotransferase activity, in which
the terminal phosphate of a NTP is removed, transferred to
Nm23-histidine 118, and subsequently transferred to a NDP (reviewed in
Ref. 2). The NDPK activity of the rNm23 proteins is shown in Fig.
3, in which the formation of [-32P]GTP
from [
-32P]ATP and cold GDP is determined by
autoradiography of TLC separations. With the exception of
Nm23-H1H118F, all of the rNm23 proteins demonstrated
significant NDPK activity. The reduced amount of
[
-32P]GTP product formed by the S44A and P96S mutants
represented slightly decreased rNm23-H1 protein input into the
reaction, as all of the available [
-32P]ATP was
consumed. Under conditions where [
-32P]ATP was not
rate-limiting, significant NDPK activity was also exhibited by each of
the site-directed mutant Nm23-H1 proteins, with the exception of H118F
(data not shown).
Histidine Protein Kinase Activity
Nm23 protein was recently
reported to exhibit histidine protein kinase activity in
vitro (37), in which the terminal phosphate of
[-32P]ATP was transferred to a histidine residue of
Nm23, which then kinased ATP citrate lyase on a histidine residue. Fig.
4A presents a radioautograph of the histidine
protein kinase activity of the wild type and mutant rNm23 proteins
in vitro. Briefly, Nm23 proteins were autophosphorylated
with [
-32P]ATP, and the unbound label was removed;
each recombinant protein, normalized by cpm of autophosphorylation, was
subsequently incubated with a novel histidine protein kinase substrate,
porcine succinyl thiokinase (STK, also known as succinyl-CoA
synthetase, EC 6.2.1.4), selected on the basis of its phosphorylatable
histidine residue. Reaction substrates and products were visualized by
electrophoresis and autoradiography. Both wild type rNm23-H1 and
rNm23-H2 kinased the
-subunit of STK; the lack of STK
phosphorylation by incubation with identically treated
Nm23-H1H118F protein confirms the lack of
[
-32P]ATP contamination in this series of experiments
and indicates histidine 118 as the Nm23-phosphorylated intermediate.
Histidine kinase activities of the rNm23 proteins were relatively
equivalent with the notable exception of the rNm23-H1P96S
protein, representing the Drosophila k-pn developmental
mutation. By densitometry of radioautographs, the STK kinase level was
6-10-fold lower when incubated with the k-pn, as compared with wild
type Nm23-H1 protein, over three independent experiments. A
corresponding greater percentage of histidine phosphorylation remained
on the Nm23-H1P96S, indicating that the deficit was due to
lack of transfer to the STK protein. Fig. 5 presents
time course data for a STK histidine kinase reaction using
autophosphorylated wild type and P96S mutant rNm23-H1 proteins. As
shown, the killer of prune rNm23-H1 failed to histidine kinase STK over
an extended period of reaction times.
Phosphoamino acid analysis of the wild type rNm23-H1 and STK were
performed to verify that histidines were involved in the reactions
previously described. Autophosphorylated rNm23-H1 was incubated with
STK; the two proteins were resolved by SDS-PAGE and transferred to a
PVDF membrane. Phosphorylated rNm23-H1 and STK were excised and
subjected to alkaline hydrolysis in 3 M KOH. Phosphoamino
acids were separated by thin layer chromatography. As shown in Fig.
6, N1-phosphohistidine was
detected in the Nm23-H1 hydrolysate, as expected (37), whereas the
product of STK alkaline hydrolysis was
N3-phosphohistidine.
To determine the generality of these data, additional histidine protein
kinase substrates were tested, including cold rNm23-H2 protein. Fig.
4B shows a radioautograph of an experiment in which rNm23
proteins were autophosphorylated with [-32P]ATP, and
excess label was removed; the autophosphorylated proteins were
subsequently incubated with cold, purified rNm23-H2 and
transphosphorylation of rNm23-H2 detected by electrophoresis and
autoradiography. Histidine autophosphorylation of Nm23-H2 occurred in
the presence of autophosphorylated wild type rNm23-H1 but not
rNm23-H1H118F. The histidine kinase activity of the various
rNm23-H1 proteins were approximately equivalent, with the exception of
decreased phosphotransferase activity of rNm23-H1P96S. A
corresponding increase in the amount of residual autophosphorylated rNm23-H1 protein was observed. The histidine protein kinase activity of
the k-pn rNm23-H1 was 4-6 times lower than that of wild type protein,
determined from densitometry of autoradiographs representing three
independent experiments. A slight but reproducible decrease in rNm23-H2
histidine phosphorylation was observed using rNm23-H1S120G
as the kinase.
Fig. 4C provides a third test of the in vitro histidine protein kinase activity of the rNm23 proteins, using the experimental design previously described and a GST-Nm23-H1 fusion protein as a substrate. Transphosphorylation of GST-Nm23-H1 by autophosphorylated wild type and mutated rNm23 proteins was dependent on the presence of a histidine 118 in the rNm23-H1, in agreement with previous results. Transphosphorylation was negligible using rNm23-H1P96S as the kinase and slightly but reproducibly reduced using rNm23-H1S120G.
Additional aliquots of the histidine protein kinase reactions were electrophoresed under conditions to permit visualization of acid-stable phosphorylation. For GST-Nm23-H1, this phosphorylation is expected to be serine and represent downstream transfer from phosphohistidine; it is unclear if acid-stable phosphorylation of the remaining proteins represents independent serine or other kinase events or a downstream transfer from phosphohistidine. A very minor proportion of acid-resistant protein phosphorylation was observed using each of the rNm23 substrates described above and exhibited autoradiographic patterns among the various mutant donor proteins similar to that of histidine phosphorylation (data not shown).
The long-term goal of the research described herein is to identify
the biochemical pathway whereby increased Nm23 expression suppresses
tumor metastatic potential in vivo in melanoma and breast
carcinoma cells (6, 8-11). Previous approaches have identified
numerous biochemical activities and/or associations for Nm23 proteins.
However, many of these activities remain debated, and few have been
linked to biological changes in development or tumor metastasis. An
alternate approach using site-directed mutagenesis was therefore
adopted, in order to more closely link biology and biochemistry. We
have previously reported the transfection of wild type
nm23-H1 and several site-directed mutant constructs into
human MDA-MB-435 breast carcinoma cells, using in vitro
motility in Boyden chamber assays as an indicator of aggressiveness
(51). Mutation of proline 96 to serine (P96S), the killer or prune
(k-pn) developmental mutation in the Drosophila homolog of
nm23, awd, or mutation of serine 120 to glycine
(S120G), a mutation of a phosphorylatable serine residue (41) found in
human stage IV neuroblastomas (57), failed to inhibit the motility of
MDA-MB-435 breast carcinoma cells to either serum or partially purified
ATX. A serine 120 to alanine (S120A) substitution, which preserved "R" group and hydrogen bonding characteristics at this position, lacked motility inhibitory activity in some, but not all assays performed. In contrast, wild type Nm23-H1 protein and a serine 44 to
alanine (S44A) mutation produced transfectants with suppressed motility
characteristics. In the present report purified rNm23-H1 and its
site-directed mutants have been surveyed for activity in several assays
involving phosphorylation and phosphotransfer, the results of which are
summarized on Fig. 7. The question asked by these
experiments is whether those mutant proteins lacking motility
suppressive capacity in vivo exhibit altered activity in a
particular biochemical pathway.
The most striking observation reported herein concerns Nm23-H1P96S. This mutation represents the homolog of a P97S mutation in Drosophila awd, which causes no phenotype alone but widespread aberrant development leading to lethality in the third instar stage when coexpressed with a pn mutation (58), most likely the result of an almost complete loss of Pn product (reviewed in Ref. 15). X-ray crystallography studies in several species have localized the k-pn mutation to a k-pn loop, whose function changes depending on the oligomeric structure of Nm23/Awd (59). We report that purified rNm23-H1P96S is deficient in the phosphotransferase portion of a protein histidine kinase reaction in vitro. While the autophosphorylation portion of this pathway remained functional, transfer of phosphate from rNm23-H1 to a histidine residue on three substrates, STK, Nm23-H2 and GST-Nm23-H1, was significantly reduced. In contrast, no significant diminution of Nm23-H1P96S histidine phosphotransfer was noted to NDPs in the NDPK reaction, or internally to Nm23-H1 serine residues.
The coordinate abrogation of motility suppressive capacity in vivo and histidine protein kinase activity in vitro by Nm23-H1P96S suggests the intriguing hypothesis that the two may be functionally linked. If Nm23 functions biologically via a protein histidine phosphotransferase pathway, it can be hypothesized that this activity normally favors the nonmetastatic cellular state. Reduced expression of Nm23 or its k-pn mutation could diminish levels of the kinased product, resulting in increased metastatic potential. Further experimentation to test this hypothesis will necessarily address the following: (a) the contribution of small amounts of wild type Nm23-H1 and Nm23-H2, as found in transfection systems; (b) possible contributions by relevant Nm23 binding proteins; (c) the histidine protein kinase activity of these wild type and mutant rNm23 proteins using as yet unidentified physiologically relevant substrates; (d) the potential contribution of pn and other interacting genes that may be discovered. It also remains possible that loss of Nm23 histidine protein kinase activity and/or mutation of Nm23 at the k-pn or other sites exerts effects on protein kinase activity not observed in the present series of analytical reactions, such as alterations in substrate specificity.
Histidine protein kinases have been thoroughly characterized in prokaryotes, where they form the "two-component" regulatory system, (reviewed in Ref. 38). This is signal transduction system, linking extracellular changes in chemoattractant, nutrient, osmotic, or other gradients to transcriptional alterations. The two-component system has also been linked to differentiation in Caulobacter crescentus (60). The prototypic elements of this system include a protein with a periplasmic sensor; when an external event is detected it modulates the activity of a protein kinase catalytic domain, which in turn phosphorylates a histidine residue in the substrate domain. The phosphate on the histidine is typically transferred to an aspartate on a second protein, which activates a coupled effector domain. Phosphohistidine is undetected in many studies due to its lability, but recent studies in eukaryotes have estimated that 6% of protein phosphorylation is histidine, a level higher than phosphotyrosine (50). Two-component systems have been described in yeast, where they are thought to regulate mitogen-activated protein kinase activity (62). Recently histidine phosphorylation of P-selectin was reported upon activation of human platelets, providing a first example of histidine phosphate involvement in mammalian signaling (39). Comparison to Nm23 function is noteworthy on several points. First, transfection of nm23 cDNAs has been reported to inhibit tumor cell signal responsiveness in colonization, motility, invasion, etc., and a role in limiting the signal transduction process has been previously proposed (6). Second, in prokaryotic histidine protein kinase systems phosphate is transferred from histidine to another amino acid residue, whereas in the assays reported herein transfer is from one histidine to another. This activity may be better termed a histidine protein phosphotransferase. However, using GST-Nm23-H1 and rNm23-H2 as a substrate we noted subsequent transfer to serine residues in vitro, which would represent an example of a downstream phosphorylation. The identification of physiologically relevant substrates will enable further definition of this activity.
The serine 120 to glycine mutation of Nm23-H1 demonstrated altered activity in multiple phosphorylation and phosphotransferase assays. In transfection experiments, nm23-H1S120G failed to inhibit tumor cell motility (51); cohort sequence data in childhood neuroblastoma identified this mutation in 6/28 aggressive (stage IV) tumors (57). Serine 120 is localized two amino acids carboxyl to the histidine phosphorylation site, is a potential site of phosphorylation (41), and is thought to contribute to active pocket conformation by hydrogen bonding to a Glu residue.3 rNm23-H1S120G exhibited deficient histidine and downstream serine rNm23-H1 autophosphorylation and, to a minor extent, deficient histidine protein kinase activity in vitro. Each of these reactions, however, represents a segment of a histidine-dependent protein phosphotransferase pathway, in general agreement with the hypothesis put forth based on the P96-S data. No difference in NDPK activity was noted. It remains possible that more significant alterations in histidine protein kinase activity for Nm23-H1S120G will emerge when physiologically relevant substrates and other variables are tested.
Mutation of Nm23-H1 serine 44, which resulted in generally inhibited motility upon transfection into breast carcinoma cells, did not demonstrate altered activity in any of the biochemical assays conducted. These data provide an important control for the specificity of the altered activities noted previously. A second mutation at serine 120, to alanine, resulted in motility inhibition under some but not all assay conditions (51). In the current series of experiments rNm23-H1S120A exhibited reduced serine autophosphorylation levels. However, this represents another histidine-dependent protein phosphotransferase activity, consistent with the hypotheses previously developed. Mutation of histidine 118 abrogated activity in all of the biochemical phosphorylation assays conducted; comparison to transfection data, however, is prevented by its poor expression characteristics.
In conclusion, assays of phosphorylation and phosphotransferase activity indicate a correlation of Nm23-H1 mutations lacking motility suppressive capacity upon transfection and exhibiting deficient activity in histidine-dependent protein phosphotransferase pathways. It is recognized that the mutant proteins may also differ in additional biochemical activities and associations untested in the present series of experiments, which could be biologically relevant. However, the data permit the development of the hypothesis that a histidine-dependent protein phosphotransferase pathway is functionally involved in the biological metastasis suppressive effect of Nm23.