(Received for publication, December 23, 1996, and in revised form, March 31, 1997)
From the Department of Anatomy and Neurobiology,
University of California at Irvine, Irvine, California 92697 and the
Departments of ¶ Biochemistry and
Physiology, Dartmouth Medical School,
Hanover, New Hampshire 03755
The extracellular matrix protein agrin plays an important role in the formation and maintenance of the neuromuscular junction. However, regulation of agrin gene expression and pre-mRNA splicing, important in determining the biological actions of agrin, is not well understood. To begin to identify mechanisms controlling agrin expression, quantitative polymerase chain reaction techniques were used to analyze the effect of growth factors on the expression of agrin mRNA isoforms in rat pheochromocytoma (PC12) cells. Agrin transcripts in untreated cells lacked inserts in the Y and Z sites (agriny0z0), encoding agrin isoforms with low acetylcholine receptor aggregating activity and a primarily non-neuronal tissue distribution. Transcripts encoding isoforms with high aggregating activity and neuronal tissue distribution (agriny4z8, agriny4z11, and agriny4z19) were not detected. Treatment of PC12 cells with nerve growth factor (NGF) caused a significant increase in total agrin mRNA. In contrast, exposure to epidermal growth factor had no effect. Analysis of alternative splicing of agrin mRNA revealed that NGF elicited a specific increase in agriny4 and agrinz8 mRNAs that did not occur in the presence of epidermal growth factor, insulin, dexamethasone, or retinoic acid. Analysis of PC12 sublines stably overexpressing a dominant inhibitory form of p21 Ras indicated that NGF induced changes in levels of agrin mRNA and alternative splicing required Ras activity. The results show that NGF can influence important aspects of neuronal differentiation by regulating alternative splicing. Furthermore, these data provide insight into the mechanisms governing agrin gene expression and suggest that neurotrophic factors may play a role in regulating agrin expression in vivo.
Agrin is an extracellular matrix protein that plays a key role in directing the formation and maintenance of the postsynaptic apparatus of the neuromuscular junction. Consistent with this hypothesis, agrin is expressed by embryonic and adult motor neurons in vivo and induces clustering of AChR1 on muscle fibers in cell culture (reviewed in Ref. 1). In addition, antibodies against agrin block motor nerve-induced clustering of AChR on cultured muscle cells (2), and the accumulation of AChR at developing neuromuscular junctions is disrupted in agrin-deficient mice (3). Agrin is also expressed in many populations of neurons throughout the central and peripheral nervous systems (4-6), raising the possibility that agrin has similar functions at other chemical synapses. However, agrin mRNA and/or protein has been detected in non-neuronal cells within the nervous system (7, 8) as well as in muscle and epithelia and other tissues (9-11). This complex tissue distribution, combined with agrin's multidomain structure (1), suggests that agrin may serve functions in addition to those defined at the neuromuscular junction. The results of studies showing that agrin is a potent protease inhibitor (12), binds to neural cell adhesion molecules (13), and can influence neurite outgrowth of cultured neurons (14) further support this possibility.
Agrin mRNA levels in brain, spinal cord, parasympathetic ganglia, and skeletal muscle undergo dramatic changes during development (4, 15, 16). Subsequent levels of expression are not necessarily static because agrin mRNA levels in several regions of the brain are altered by seizures (17). In addition, although agrin is encoded by a single gene, alternative splicing of agrin pre-mRNA gives rise to multiple protein isoforms that differ in their tissue distribution and biological activity. In particular, two sites of alternative splicing in the carboxyl-terminal half of the protein, referred to in the rat gene as Y and Z (15, 18) and in chicken as A and B (7), respectively, have been identified. Optional exclusion of sequences encoding 4 amino acids at the Y site and 8 and 11 amino acids at the Z site result in mRNAs encoding agrin proteins with unique functional properties and tissue-specific patterns of expression. For example, forms of agrin lacking inserts at the both Y and Z sites (agriny0z0) are unable to induce clustering of AChR (7, 19) and are most highly expressed by non-neuronal cells (7, 8). In contrast, agrin proteins containing the 4-amino acid insert at the Y site and either one or both of the 8 and 11 amino acid inserts at the Z site (agriny4z8, agriny4z11, and agriny4z19) are potent inducers of AChR aggregates (7, 19) and are specifically expressed by neurons (7, 8). Despite the importance of agrin gene expression as a determinant of agrin function during development and in different tissues, little is known about the mechanisms that regulate it.
Changes in gene expression underlying the formation and maintenance of chemical synapses are coordinated through signals exchanged between neurons and the structures they innervate. Within this scheme, agrin acts as an anterograde signal that triggers changes in the postsynaptic cell. Several lines of evidence suggest that neuronal agrin gene expression may in turn be influenced by factors produced by or associated with target tissues. For example, total agrin mRNA levels in the ciliary ganglion increase dramatically following contact between ganglionic neurons and their targets in the eye and decrease when the connection with their target tissues is interrupted by axotomy (4, 20). Furthermore, changes in the pattern of alternatively spliced agrin mRNAs in ciliary ganglion neurons during development and following axotomy suggest that alternative splicing of agrin mRNA may be similarly target-dependent (4, 20). Although similar developmental changes in agrin expression have been reported throughout the nervous system (5, 15, 16, 21), the mechanisms underlying these changes are not known.
Among the target-associated factors having an important influence on neuronal development are soluble growth factors. In particular, the neurotrophins, a family of neurotrophic factors that includes NGF, influence neuronal survival, growth, and differentiation in vitro and in vivo (for review see Refs. 22-25). Neurotrophins also affect the development of synaptic connections and the expression of synaptic proteins and modulate synaptic function (reviewed in Ref. 26). Whether neurotrophins have effects on agrin gene expression is unknown, although it is intriguing that the seizure-induced changes in agrin mRNA in rat hippocampus (17) follow seizure-induced changes in neurotrophin gene expression that also occur there (27, 28).
To begin to identify mechanisms regulating agrin gene expression, the ability of various growth factors to influence agrin mRNA levels was analyzed in the PC12 rat pheochromocytoma cell line. PC12 cells have been widely used to investigate the cellular and molecular mechanisms underlying neuronal differentiation and the actions of growth factors. For example, upon activation of the TrkA receptor by NGF, PC12 cells adopt a sympathetic neuron-like phenotype that results from the activation of a variety of signaling pathways and a spectrum of changes in protein phosphorylation and gene expression (for review see Refs. 29-32). The results presented here demonstrate that NGF selectively regulates both the level of total agrin mRNA and the pattern of alternative splicing of agrin mRNA in PC12 cells and does so in a manner that requires Ras activity, providing the first insight into the mechanisms governing agrin gene expression.
Cells were maintained in a humidified CO2 environment in 100-mm tissue culture dishes (Falcon Labware, Becton Dickinson, Lincoln Park, NJ) containing Dulbecco's modified Eagle's medium with 0.45% glucose, 10% fetal bovine serum, 5% heat-inactivated horse serum, 100 units/ml penicillin, and 100 µg/ml streptomycin (Life Technologies, Inc.). Culture medium for the 17.2 and 17.26 sublines (kindly supplied by G. Cooper, Dana Farber Institute, Boston, MA) and the protein kinase A-deficient 123.7 subline (kindly supplied by J. Wagner, Cornell Medical Center, New York, NY) also contained 500 µg/ml G418 (Life Technologies, Inc.). In cases where 100 ng/ml 7S NGF (Upstate Biotechnology, Inc., Plattsburg, NY), 100 ng/ml EGF (Upstate Biotechnology), or 300 nM insulin (Sigma) were included, fresh growth factor was added every other day when the medium was changed.
RNA Isolation and PCRTotal cellular RNA was isolated from dishes containing approximately 106 PC12 cells, using the method of Chirgwin et al. (33). The molar concentration of agrin mRNA in the RNA samples was determined by competitive PCR, as in our previous studies (4, 20). Briefly, a competing template was generated by subcloning a PCR fragment corresponding to nucleotides 4516-4820 of the published rat agrin gene sequence (34) and common to all agrin isoforms into the plasmid vector pGEM-T (Promega Corporation, Madison, WI). This agrin DNA fragment contained a unique KpnI restriction site at position 4739 that was subsequently mutated to a BamHI site by recombinant PCR (35) and subcloned into pGEM-T to yield the competing template pRABam. For competitive PCR, cDNA from 50 ng of total RNA was mixed with different amounts (0.05-10 pg) of pRABam (diluted in Tris/EDTA buffer containing 0.5 µg/ml of yeast tRNA) and coamplified in a thermal cycler (Perkin-Elmer) for 40 cycles using an annealing temperature of 60 °C and the F25/B12 primer pair (see list below). Reaction products were labeled by inclusion of ~2 × 105 cpm of 32P-labeled forward primer. Subsequently, 2 µl of each reaction was subject to overdigestion with BamHI, and the products were separated by electrophoresis on a 6% polyacrylamide gel. Product yields from endogenous and competing templates were determined by analysis on a PhosphorImager (Molecular Dynamics, Sunnyvale, CA), and the point of equivalence was determined by linear estimation.
The relative abundance of alternatively spliced agrin mRNAs was
determined by PCR using aliquots of cDNA synthesized from 100-200
ng of total RNA. Reaction conditions were as described in our previous
studies (6). PCR products were radiolabeled by inclusion of ~2 × 105 cpm of 32P-labeled forward primer in the
reaction mixture, separated by electrophoresis on 8% polyacrylamide
gels, and the relative abundance determined by analysis of the gel with
a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). To investigate
the pattern of alternative splicing at either the Y or Z sites, samples
were subjected to a single round of amplification (35 cycles) using
oligonucleotide primers F111/B112 or F24/B2 flanking the Y and Z sites,
respectively (18, 19). To examine alternative splicing in the Z region of agrin transcripts containing the 4-amino acid insert at the Y site,
two rounds of amplification were performed using nested primers. During
the first round of amplification the forward primer F110, whose 3-end
consisted of the 12-base pair sequence of the 4-amino acid insert, was
used in combination with the reverse primer B6, to specifically amplify
agriny4 mRNAs. Subsequently, first round PCR products
were diluted 1000-fold and reamplified using the F24/B2 primer
combination flanking the Z region.
Primers used in the study are
listed below with the corresponding location in the published rat agrin
cDNA (34) shown in parentheses: F25,
5-TGTGGAGACCATCCGTGCTTAC-3
(4516-4537); B12, 5
-TTCAGTTCCAGGTAGGAGAAGCC-3
(4820-4798); F111,
5
-TGGGTCAGGGTATTCCTGGAA-3
(5044- 5064); B112,
5
-AGGTTGAGCATGGTGTGCGG-3
(5143-5162); F24, 5
-AGTCAGTGGGGGACCTAGAAACAC-3
(5462-5485); B2,
5
-AAGCTCAGTTCAAAGTGGTTGCTC-3
(5576-5553); F110,
5
-GAATCTCCGAAATCCCGCAAG-3
(5119-5139); and B6,
5
-TGGTGTTGACCTTCACAGTGGAG3
(5791-5769).
The statistical significance of the differences in samples from different experimental conditions was determined using ANOVA. Differences between control untreated cells and cells treated with growth factors was determined using Student's t test. All statistical analyses were performed on the raw data. Data presented in the text and figures represent the means ± S.E.
In response to NGF, PC12 cells stop dividing and
differentiate, acquiring characteristics of sympathetic neurons. To
determine whether this neuronal differentiation included changes in
agrin gene expression, competitive PCR was used to compare the molar concentration of total agrin mRNA in PC12 cells maintained in the
presence or the absence of 100 ng/ml NGF. Analysis of two different
sublines of wild type PC12 cells revealed that NGF caused a significant
increase (p < 0.01, ANOVA) in total agrin mRNA, with the average level of expression in PC12 cells treated with NGF for
7 days more than 2.5 times that in untreated cells (Fig. 1; p < 0.02, t test).
To determine whether the increase in agrin mRNA was a general response to growth factors simply reflecting hypertrophy of the cells or instead was associated with their neuronal differentiation and specific actions of NGF, the influence of EGF on agrin mRNA expression in PC12 cells was also determined. Although EGF is similar to NGF in that both exert their effects through specific transmembrane receptor tyrosine kinases, their biological effects are in sharp contrast, because EGF promotes the growth and proliferation of PC12 cells (36). As evidenced by increases in c-fos mRNA, ornithine decarboxylase activity, and protein phosphorylation detected in parallel studies of these cells (37), the two independent sublines of wild type PC12 cells analyzed responded to EGF. However, EGF had no detectable effect on the level of agrin mRNA in these sublines, even after treatment for 7 days (Fig. 1). This suggests that the increase in agrin mRNA is among the specific responses to NGF associated with the neuronal differentiation of these cells.
Alternative Splicing of Agrin RNA Is Specifically Regulated by NGFAlternative splicing, particularly the optional inclusion of
8- and 11-amino acid exons in the Z site, is an important mechanism regulating both cell-specific and developmental patterns of agrin gene
expression. Accordingly, oligonucleotide primers flanking the Z site
were used to examine the pattern of expression of alternatively spliced
agrin transcripts in wild type PC12 cells maintained in the presence or
the absence of growth factors. As shown in an autoradiogram from a
representative experiment (Fig. 2A), based on
their electrophoretic mobility and confirmed by sequencing, PCR
products representing two different agrin isoforms were amplified from
treated and untreated cells. Consistent with previous analyses of agrin
gene expression in vivo (6, 15, 16), transcripts lacking
both exons at the Z site (agrinz0) were always the most abundant isoform detected. More striking, however, was the increase in
the relative abundance of agrinz8 transcripts in response
to NGF. This increase appeared to be selective, because
agrinz11 and agrinz19 mRNA remained below
the limit of detection (Fig. 2A). Quantitative analysis of
two different sublines of wild type PC12 cells confirmed that there was
a significant increase in agrinz8 mRNA
(p < 0.001, ANOVA), with the relative abundance more than 12-fold greater (p < 0.03, t test) in
NGF-treated versus untreated cells (Fig. 2B). The
increase in agrinz8 mRNA was a relatively late response
to NGF, most noticeable in cells that had been treated for 5-7 days.
Similar to the change in total agrin mRNA, the change in the
pattern of agrin mRNA isoforms appeared to be specific to NGF. For
example, less than a 2-fold increase in the relative abundance of
agrinz8 mRNA was observed following treatment with a
saturating concentration of EGF, and no effect was produced by insulin
(Fig. 3), despite the fact that in a parallel study (37)
these two factors had clearly detectable effects on the cells.
Dexamethasone and retinoic acid also failed to change agrinz8 mRNA levels (data not shown).
To compare the effects of NGF treatment on the absolute levels of
agrinz0 and agrinz8 mRNA, data for total
agrin mRNA (Fig. 1) were transformed using the measurements of
relative abundance of agrinz0 and agrinz8
mRNA obtained for each RNA sample. As shown (Fig.
4), there was a significant increase
(agrinz8, p < 0.001; agrinz0,
p < 0.01, ANOVA) in the molar concentrations of both transcripts. However, the level of agrinz0 mRNA
increased approximately 2.5-fold (p < 0.02, t test) upon treatment of the cells with NGF for 7 days
(Fig. 4A), similar to the increase in total agrin mRNA levels. In contrast, the increase in the level of agrinz8
mRNA was much more dramatic, increasing 35-fold (p < 0.003, t test) during the same time period (Fig.
4B).
Inclusion of a 4-amino acid exon at the Y site is necessary for normal
AChR aggregating activity of agrin and occurs only in neural tissues.
Therefore it was of particular interest to analyze the pattern of agrin
mRNA splicing at this site. Oligonucleotide primers flanking the Y
site were used to compare the pattern of expression of alternatively
spliced agrin transcripts in PC12 cells maintained in the presence or
the absence of growth factors. In the absence of added growth factors
(Fig. 5A), only agriny0 mRNA
was detected, with agriny4 mRNA at or below the level
of detection. However, in response to NGF the relative abundance of
agriny4 transcripts increased 11-fold over background
(p < 0.001, t test), comparable with the
increase in relative abundance of agrinz8 mRNA. Again,
this induction was specific to NGF, because treatment with EGF for 7 days failed to induce the appearance of agriny4 mRNA
(Fig. 5A).
Previous studies have shown that alternative splicing at the Y and Z
sites are coordinated such that only mRNAs containing the 4-amino
acid insert at the Y site can include the 8- and/or 11-amino acid
inserts at the Z site (15). To examine the influence of NGF on the
coordinate splicing of the Y and Z sites directly, two rounds of
amplification were performed using nested primers flanking the Z site,
in which the 3-end of the forward primer used for the first round of
PCR was coextensive with the y4 exon sequence. In response to NGF, the
relative abundance of agriny4z8 mRNA increased
dramatically (Fig. 5B, p < 0.001, t test). However, even at the higher sensitivity with which
agriny4 transcripts are amplified using this approach,
agriny4z8 mRNA was not detected in untreated or
EGF-treated PC12 cells (Fig. 5B). Together these data show
that when analyzed separately or in combination, the effects of NGF on
the pattern of agrin mRNA splicing at the Y and Z sites in PC12
cells is consistent with its effects on neuron-specific gene expression
and neuronal differentiation.
An important element in the transmission of signals from
the Trk receptor for NGF is the activation of Ras, which is
necessary for many of the responses to NGF to occur (for review see
Ref. 32). However, not all responses of PC12 cells to NGF, including the induction of sodium channel expression, require Ras activity (38,
39). Thus the extent to which Ras is involved in neuronal differentiation has not been fully defined. To determine if Ras activity is necessary for the NGF-induced changes in agrin mRNA expression, we analyzed two independent sublines of PC12 cells (17.2 and 17.26) that stably overexpress the dominant inhibitory ras
mutant c-Ha-ras (Asn-17). Previous analysis of these
sublines has shown that the absence of Ras activity blocks NGF-induced neurite outgrowth and certain changes in gene expression, while leaving
other functional responses and changes in gene expression intact (38,
40). The levels of total agrin mRNA in untreated 17.2 and 17.26 cells were similar to the levels detected in wild type PC12 cells when
maintained in the absence of NGF. However, the induction of total agrin
mRNA normally observed in response to NGF was completely blocked in
the 17.2 and 17.26 cells (Fig. 6). Furthermore, whereas
the relative abundance of the agrin mRNA isoforms in untreated 17.2 and 17.26 cells was similar to that found in the parental subline of
wild type cells in the absence of added growth factors, the induction
of agrinz8 mRNA by NGF was inhibited in the 17.2 and
17.26 cells (Fig. 7), as was the induction in
agriny4 mRNA (data not shown). The small, though not
significant, increase in agrinz8 mRNA observed in the
17.2 subline is consistent with previous reports demonstrating lower levels of expression of the mutant Ras protein in this subline compared
with 17.26 cells (40). Thus, the NGF-induced changes in total agrin
mRNA expression and the pattern of alternative splicing of agrin
mRNA both occur via Ras-dependent mechanisms in PC12
cells.
Previous studies have shown that the levels of agrin mRNA expression and the pattern of alternative splicing of agrin mRNA in the nervous system are cell-specific (7, 8, 15), are regulated during development (4, 15, 16), and can be influenced experimentally by postganglionic axotomy (20) or elevated levels of neuronal activity that accompany seizures (17). Together these results suggest that agrin gene expression in the nervous system is influenced by environmental cues. Here we show that NGF alters both the level of expression and pattern of alternative splicing of agrin mRNA in PC12 cells, whereas other growth factors had little or no effect. These results extend our understanding of both the actions of neurotrophins and the mechanisms governing agrin gene expression and suggest that similar mechanisms may be important in regulating agrin gene expression in vivo.
Neurotrophins have dramatic effects on neuronal differentiation and phenotype through effects on gene expression (reviewed in Refs. 23 and 31). In PC12 cells, which adopt characteristics of sympathetic neurons in response to NGF, this is evidenced by changes in gene expression associated with neurite outgrowth, neurotransmitter synthesis, neurotransmitter receptor expression, and the development of electrical excitability. These changes often reflect alterations in the transcriptional activity of various genes, as well as effects on mRNA stability (41-44). Consistent with the effect of NGF on neuronal differentiation, our studies show that NGF not only causes an increase in total agrin mRNA but also induces the expression of agrin mRNA splice variants with inserts at the Y and Z sites (agriny4z8, agriny4z11, and agriny4z19), which are specifically expressed by neurons (7, 8, 15). The increases in agrinz8 mRNA were detected in several independent sublines of wild type PC12 cells, with the levels in NGF-treated cells similar to those found in adult rat olfactory bulb and cerebellum (6). Our results are in apparent contrast to a previous report that PC12 cells only express agrin splice variants lacking inserts at the Z site, even in the presence of NGF (15). However, in this earlier study the expression of agrin mRNA was analyzed in cells treated with NGF for only 2 days, which according to our data is before significant increases in agrinz8 mRNA occur. The increase in agrin mRNA and induction of the "neuronal" isoform in PC12 cells appear to be a late response to NGF, similar to the induction of other genes associated with a neuronal phenotype in these cells. For example, induction of genes encoding neurofilament-L, acetylcholinesterase, GAP-43, and nicotinic AChR, is first seen at 24-48 h, whereas muscarinic and enkephalin receptors, synapsin, Na/K ATPase, and the neurofilament protein peripherin occurs 48-96 h following exposure to NGF (for review see Ref. 31). We conclude, therefore, that the level of expression and pattern of alternative splicing of agrin mRNA parallels other aspects of NGF-mediated neuronal differentiation in PC12 cells and that this represents yet another mechanism by which neurotrophins affect an important aspect of neuronal gene expression and differentiation.
The regulation of agrin gene expression by NGF also contributes to the emerging evidence that neurotrophins have important effects on synapses. Neurotrophins influence the development of synaptic connections, affect the expression and modification of synaptic proteins, and modulate synaptic function at both pre- and post-synaptic levels (45-49). Our results indicate that agrin gene expression is also a synapse-related target of neurotrophin action. Our observation that differentiated PC12 cells contain mRNA encoding an agrin isoform with high AChR aggregating activity may also be relevant to the reported ability of PC12 cells to form functional synapses when co-cultured with L6 skeletal muscle cells (50). Given that growth factors present in muscle, including members of the neurotrophin family, influence motor neurons (51), it will be important in future studies to determine if the patterns of agrin gene expression in motor neurons and other primary neuronal populations are the consequence of the growth factor receptors they express and the soluble growth factors to which they are exposed. Indeed, in separate studies we have found that the neurotrophins can cause increases in the abundance of neuron-specific agrin mRNA isoforms in primary cultures of embryonic rat brain neurons.2
Alternative splicing is a ubiquitous mechanism that allows diverse
protein isoforms to be encoded by a single gene (for review see Refs.
52 and 53). It plays a critical role in the expression of a wide
variety of proteins in the nervous system, including agrin,
acetylcholine receptor inducing activity,
N-methyl-D-aspartate receptors, Trk receptors,
calcitonin gene-related protein, and neural cell adhesion molecule (7,
54-58). Despite this, the mechanisms that regulate patterns of
alternative splicing are not well understood. Previous studies have
shown that growth factors influence the pattern of alternative splicing
in non-neuronal cells. For example, serum growth factors and EGF induce
alternative splicing of the phosphotyrosine phosphatase PTP-1B
pre-mRNA in HeLa cells, whereas transforming growth factor-
modulates alternative splicing of fibronectin and tenascin pre-mRNA
in fibroblasts (59-62). In PC12 cells, NGF affects the alternative
splicing of the pre-mRNAs encoding Trk and the
calmodulin-dependent plasma membrane
Ca+2-ATPase (56, 63). However, to our knowledge the results
presented here are the first example of NGF influencing the pattern of
alternative splicing in which the functional importance and biological
implications of the splicing are known.
The exact mechanism(s) by which growth factors cause differential accumulation of alternatively spliced mRNAs is not known, and our results, like those described above, do not distinguish among different possibilities such as selective stabilization of a specific mRNA isoform or the regulation of specific splicing factors. However, in cases where NGF either affects mRNA stability (41) or alternative splicing (53, 63), there is no evidence to suggest that these changes involve selective stabilization of specific mRNA isoforms. On the other hand, tissue-specific splicing factors have been identified (53), and a possible role for such factors in regulating alternative splicing in PC12 cells has been raised in a previous study (56). It will be important in future studies to determine more precisely the mechanism by which NGF triggers changes in agrin expression and learn whether alternative splicing or expression of other neuronal genes associated with synaptic differentiation, such as acetylcholine receptor inducing activity (64), are similarly regulated by neurotrophic factors.
Previous studies describing the changes in relative abundance of agrin mRNA isoforms during development have generally relied upon quantitative PCR analysis of total RNA isolated from brain, spinal cord, or peripheral ganglia (4, 15). As a result, it has been unclear whether changes in the relative abundance of different agrin mRNA species reflect changes in agrin gene expression within individual neurons, the proliferation of non-neuronal cells that express different agrin isoforms, or naturally occurring cell death that preferentially affects a subpopulation of cells, thereby sculpting the pattern of agrin gene expression observed for the population as a whole. Single cell RT-PCR analysis of chick ciliary ganglion cells is also consistent with more than one of these possibilities, revealing considerable heterogeneity in the agrin mRNA profiles of individual cells at any given embryonic stage, as well as a decline in the percentage of cells that express agrinz11 mRNA during development (8). In comparison, PC12 cells are a relatively homogeneous population. Thus, the NGF-mediated increase in the level of agrinz8 mRNA that occurred consistently in several independent sublines of wild type PC12 cells suggests that changes in the pattern of alternative splicing of agrin mRNA can occur within an individual cell type and be regulated by environmental signals.
In developing and adult nervous tissue, agrinz0 mRNAs are the most abundant agrin transcripts, and generally account for more than 50% of the total agrin mRNA (4, 6, 15). However, the relative abundance of agrin Z region variants changes during development, with the levels of agrinz11, agrinz19, and agrinz8 mRNA peaking at early, intermediate, and late stages, respectively, in a pattern that appears to be reiterated throughout the nervous system (4, 15, 21). In the results presented here, the agrinz0 transcript was also the most abundant isoform expressed in both untreated and NGF-treated PC12 cells. However, NGF-induced differentiation of PC12 cells increased the expression of agrinz8 mRNAs without expression of detectable levels of agrinz11 or agrinz19 mRNAs. These data suggest that the temporal sequence of expression of Z region variants observed in vivo is not obligatory and that specific cell-cell signals may be responsible for triggering expression of each isoform or combination of isoforms.
Our data provide the first information about the molecular mechanisms
regulating the pattern of agrin mRNA isoform expression. NGF
activates the TrkA receptor tyrosine kinase, which results in the
activation of a number of cellular signaling pathways, including those
involving phosphatidyinositol-3-kinase, phospholipase C, and the Ras
GTPase-activated cascade of kinases that includes MAPK. Of these, the
activation of Ras and the signaling events subsequent to it have been
shown to play a major role in the response to NGF (for review see Refs.
31 and 32). In response to EGF and insulin, which cause proliferation
of PC12 cells rather than differentiation, there are also increases in
receptor-mediated tyrosine kinase activity. In fact, comparison of the
cellular signals generated in PC12 by NGF, EGF, or insulin indicates
that many, but not all, of the same signals are generated by these factors in PC12 cells (for review see Refs. 36, 65, and 66). However,
in contrast to the sustained increase in MAPK activity in PC12 cells in
response to NGF, both EGF and insulin elicit only transient increases
in these cells (65), a difference that has been proposed as the
distinction between those factors that are proliferative signals and
those that lead to neuronal differentiation (67). In the present study,
we show in two independent sublines of PC12 cells expressing a dominant
inhibitory mutant form of Ras that the NGF-mediated increases in total
agrin mRNA, including the specific induction of agrinz8
mRNA and agriny4 mRNA, are abolished. This
inhibition was specific and not simply due to interfering with
signaling molecules in these cells, because the NGF-induced changes in
agrin gene expression still occurred in sublines of PC12 cells rendered
protein kinase A-deficient via the overexpression of a dominant
inhibitory mutant form of the protein kinase A regulatory subunit (data
not shown). Furthermore, in the Ras-deficient cells several
Ras-independent responses to NGF still occurred, including the
induction of sodium channel expression (38). As a
Ras-dependent response to NGF, the changes in agrin gene
expression are similar to other neuronal-specific responses to NGF in
PC12 cells, including the induction of transin mRNA, SCG10
mRNA, and neurite outgrowth (39, 40). The involvement of Ras
activity in the regulation of agrin expression is also similar to other
signaling events associated with the formation of synapses.
Specifically, the effect of acetylcholine receptor inducing activity on
AChR gene expression involves Ras/MAPK signaling and can be mimicked by
activated Ras and inhibited by dominant inhibitory forms of Ras (68).
Finally, the involvement of the Ras/MAPK pathway in the regulation of
agrin mRNA expression may also explain the effect of EGF on agrin
mRNA expression. The small induction in agrin mRNA expression
elicited by EGF might reflect the transient activation of the Ras/MAPK pathway by EGF, whereas the sustained elevation of MAPK activity may be
the basis for the larger changes in agrin mRNA observed in response
to NGF. However, EGF also elicits a small increase in sodium channel
expression in PC12 cells, despite the fact that sodium channel
induction in PC12 cells is a Ras-independent response to NGF (38).
Therefore, other mechanisms may also be responsible for the difference
in the effectiveness of various growth factors to elicit changes in
agrin gene expression in PC12 cells. Analysis of additional PC12
sublines with alterations in signaling molecules should allow further
identification of the signaling pathways involved in the regulation of
agrin gene expression by NGF.
In summary, the results presented here extend our understanding of both the actions of neurotrophins and the mechanisms governing agrin gene expression and suggest that the NGF-induced changes in agrin pre-mRNA splicing in PC12 cells may be advantageous for future studies. For example, PC12 cells may be a good candidate system for identifying factors that regulate agrin pre-mRNA splicing events. Several tissue-specific splicing factors have been identified (for review see Ref. 53), and it is possible that the regulation of agrin pre-mRNA splicing by NGF is due to its effects on the expression or activity of factors like these. In addition, the availability of a wide variety of PC12 sublines with experimentally generated alterations in growth factor signaling components provides a means of delineating the cellular signals that regulate the pattern of agrin pre-mRNA splicing and will increase our understanding of the actions of growth factors in the nervous system.
We thank J. A. Wagner (Cornell University Medical Center) for the PKA-deficient 123.7 cells, G. M. Cooper (Dana Farber Cancer Institute) for the Ras-deficient 17.2 and 17.26 cells, and L. P. Henderson and D. K. O'Dowd for comments on the manuscript.