Characterization of posttranslational modifications in neuron-specific ...

3 downloads 0 Views 1MB Size Report
neuron-specific class III /8-tubulin by mass spectrometry .... 1 Left was stained with Coomassie blue and shows clearly the high degree of protein heterogeneity ...
Proc. Nadl. Acad. Sci. USA Vol. 88, pp. 4685-4689, June 1991 Chemistry

Characterization of posttranslational modifications in neuron-specific class III /8-tubulin by mass spectrometry (tubulin heterogeneity/isoelectric focusing/tandem mass spectrometry)

JANICE E. ALEXANDER*, DONALD F. HUNT*t, MICHAEL K. LEEO, JEFFREY SHABANOWITZ*, HANSPETER MICHEL*, SUNETARY C. BERLIN*, TIMOTHY L. MACDONALD*, RICHARD J. SUNDBERG*, LIONEL I. REBHUN*, AND ANTHONY FRANKFURTER: Departments of *Chemistry and tBiology, University of Virginia, Charlottesville, VA 22903

Communicated by J. L. Beauchamp, December 3, 1990 (received for review September 5, 1990)

ABSTRACT Class III fi-tubulin, isolated from adult bovine brain, is resolved into at least seven charge variants on isoelectric focusing gels. To identify the posttranslational modifications responsible for this heterogeneity, a mixture of brain tubulins was treated with cyanogen bromide and the C-terminal fragments from the class III I3-tubulin isoforms were then isolated by binding them to the monoclonal antibody TuJ1. Combined use of tandem mass spectrometry and both subtractive and automated Edman degradation chemistry on the isolated peptides indicates that many of the isoforms differ by phosphorylation at Ser-444 plus attachment of one to six glutamic acid molecules to the side chain of the rst glutamate residue, Glu-438, in the C-terminal sequence Tyr-Glu-AspAsp-Glu-Glu-Glu-Ser-Glu-Ala-Gln-Gly-Pro-Lys.

attachment of at least three and perhaps as many as six glutamic acid molecules to the side chain of the first glutamate residue, Glu-438, in the C-terminal sequence Tyr-GluAsp-Asp-Glu-Glu-Glu-Ser-Glu-Ala-Gln-Gly-Pro-Lys (YEDDEEESEAQGPK).

Microtubules, assembled from two similar 50-kDa proteins designated a- and f3-tubulin, are involved in a number of important biological processes including segregation of chromosomes during cell division, cell motility, organelle transport, and maintenance of cell shape (1). Brain tubulin exhibits a high degree of heterogeneity. Up to 21 charge variants have been observed by isoelectric focusing (IEF) techniques (2-4). Both a- and 83-tubulins are encoded by small multigene families (5). Since the number of charge variants exceeds the number of tubulin genes by approximately a factor of 2, the remaining isoforms are assumed to result from posttranslational modifications. a-Tubulin is reported to undergo acetylation (6-8), glutamylation (9), and the addition and removal of a C-terminal tyrosine residue (10, 11). Phosphorylation of a single seine residue in one of the five P-tubulin isotypes expressed in brain has also been described (12-16). Class III p8-tubulin, a vertebrate isotype found only in neurons and cells in the mammalian testis (17), undergoes a developmentally regulated increase in heterogeneity in the former but not the latter tissue (18, 19). The testis and earliest embryonic rat brain isoforms cofocus on an IEF gel (19). At least six additional isoforms of neuron-specific class III ,B-tubulin are detected in the adult rat or bovine brain. Previous work localized the site of heterogeneity to the C-terminal 20 amino acids (19), a highly acidic region of the protein thought to be involved in the interaction of microtubules with both calcium ions (20) and microtubule-associated proteins (21-23). To further characterize both the nature and location of structural modifications responsible for the observed heterogeneity, class III /3-tubulin from bovine brain was subjected to amino acid sequence analysis by tandem mass spectrometry. We report that the neuron-specific class III /3-tubulin isoforms result in part from phosphorylation at Ser-444 and

MATERIALS AND METHODS Isolation of Brain Tubulin. Bovine brain tubulin was isolated and purified by the modified procedure of Shelanski and Weingarten (24). High-resolution IEF was performed with flat-bed IEF gels as described (18). CNBr Cleavage. Bovine brain tubulin (a 2-mg mixture of isotypes and isoforms of a- and 1-tubulin), -25% of which corresponds to the class III -tubulin isotype, was dissolved in 2 ml of 70% (vol/vol) trifluoroacetic acid (TFA) and treated with excess CNBr, and the resulting mixture was allowed to stand at 37°C for 18 h in the dark under argon. The reaction was stopped by the addition of 3 ml of distilled water and by lyophilization of the resulting mixture. Isolation of the C-terminal CNBr fragments derived from the class III f3-tubulin isoforms was accomplished by incubating the reaction mixture with a 4-8 molar excess of the monoclonal antibody TuJ1 (18) in phosphate-buffered saline (pH 7.4) for 5 h. Unbound peptides were removed by filtration through a Centricon filter (Amicon) with a 30-kDa cutoff. Bound fragments in the retentate were liberated on treatment of the antibody with 100 Al of 0.1% TFA. Solubilized peptides passed through the filter and were then fractionated further by reverse-phase HPLC on an Applied Biosystems model 130A separation system. Sample dissolved in 0.1% aqueous TFA (100 ,ul) was injected onto a narrow-bore RP300 Aquapore column (2.1 mm x 3 cm) and material was eluted with a 40-min linear gradient of 0%-60% (vol/vol) acetonitrile (0.085% TFA) in 0.1% TFA. Column effluent was monitored at 214 nm and fractions were collected by hand. Solvent was then removed by lyophilization. Early eluting fractions were rechromatographed on a narrow-bore RP-18 column (2.1 mm x 10 cm) and material was eluted with a 40-min linear gradient of 0%-20% acetonitrile (0.085% TFA) in 0.1% TFA. Preparation of Peptide Methyl Esters. Peptide methyl and ethyl esters were prepared as described (26). Mass Spectrometry. Mass spectra were recorded on a triple-quadrupole mass spectrometer (TSQ-70; FinniganMAT, San Jose, CA) (25-27) and a Fourier transform mass spectrometer (28). Operation of these instruments has been described (25-28). Subtractive Edman Degradation. Sample dissolved in 50% (vol/vol) aqueous pyridine (5 ,l) was mixed with 5% (wt/vol) phenylisothiocyanate in pyridine (5 ,l) and the resulting

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Abbreviations: IEF, isoelectric focusing; TFA, trifluoroacetic acid; CAD, collision-activated dissociation; PTH, phenylthiohydantoin. tTo whom reprint requests should be addressed.

4685

446%PWR6

_

Chemistry: Alexander et al.

Proc. Natl. Acad. Sci. USA 88 (1991)

mixture was allowed to stand under argon at 50'C for 1 h. Solvent was removed by lyophilization. The residue was dissolved in anhydrous TFA (10,ul) and the resulting solution allowed to stand under argon at 370C for 5 min. TFA was removed by lyophilization and the sample was redissolved in distilled water (10/l4). This solution was extracted twice with 30,ul of butyl acetate, the aqueous layer was lyophilized, and the resulting residue was redissolved in glacial acetic acid. A sample was removed for analysis by mass spectrometry and the remaining solution was lyophilized. Additional cycles of Edman degradation were performed on the residue as necessary. Automated Edman Degradation. Automated Edman degradation was performed by standard methods on an Applied Biosystems model 473 protein sequencer.

RESULTS An IEF gel from the analysis of a typical sample of brain tubulin is displayed in Fig. 1. Lanes 1 and 2 contain tubulin obtained from bovine and rat adult brains, respectively. The gel in Fig. 1 Left was stained with Coomassie blue and shows clearly the high degree of protein heterogeneity present in each sample. Protein in Fig. 1 Right was transferred by capillary action to a poly(vinylidene difluoride) membrane and visualized with an anti-class IIIf,-tubulin antibody (TuJ1) that recognizes an epitope within the last 20 amino acids of this particular isotype (18). Bands corresponding to at least seven isoforms of class III 13-tubulin were observed in samples from rat and bovine brain. During neural development in the rat, the number of class III /3-tubulin isoforms increases from one to seven over a period of 44 days (19). The appearance of an identical or highly similar set of seven bands in the sample from bovine brain suggests strongly that these isoforms also result from a developmentally regulated posttranslational process. To isolate posttranslationally modified peptides from the class III -tubulin, the complete mixture of tubulin isotypes and isoforms from bovine brain was treated with CNBr. The resulting peptide fragments were then incubated with monoclonal antibody TuJ1, which binds selectively to an epitope on the C terminus of the class III f-tubulin isotype. Unbound peptides were removed from the solution by filtration. Bound peptides were liberated on treatment of the antigen-antibody complex with dilute acid and fractionated further by reversephase HPLC. Peptides were characterized initially by recording mass spectra on samples from each of the HPLC fractions. Bombardment of an acidic matrix containing sample plus thioglycerol with 6-keV Cs+ ions was employed to desorb (M+H)+ ions from each peptide into the gas phase. This latter process is particularly efficient for hydrophobic peptides that reside preferentially on the surface of the matrix and much less efficient for hydrophilic peptides that are highly solvated by the thioglycerol. Accordingly, to maximize the observed signal, it is often advantageous to convert

___

1 2 CB

2

1

J

polar peptides to the more hydrophobic methyl or ethyl esters prior to mass analysis. This tactic was employed because previous work suggested that the posttranslationally modified residues in class III P3-tubulin would be found at the C terminus of the protein, a highly polar region of the molecule. Shown in Fig. 2 are the mass spectra recorded on the peptide methyl esters found in two of the early eluting fractions (fractions 9 and 12) from a typical HPLC separation (data not shown). In Fig. 2B, five (M+H)+ ions were observed at m/z (average mass values) 1881.8, 2024.7, 2167.9, 2311.3, and 2454.7. Each of these signals is separated by 143 Da, the incremental value expected if the peptides corresponding to a pair of adjacent ions differ in composition by a single glutamic acid residue in the form of its methyl ester. When the peptides in another sample of this same HPLC fraction were converted to ethyl esters, the resulting mass spectrum contained (M+H)+ ions shifted to higher mass by 126, 140, 154, 168, and 182 Da, respectively. Since conversion of a methyl ester to ethyl ester increases the mass of the molecule by 14 Da per carboxylic acid group, the above results indicate that the five peptides contained 8, 9, 10, 11, and 12 acidic residues, respectively, plus the usual C-terminal carboxyl group. In the ethyl ester spectrum, the incremental mass shift observed for a pair of adjacent signals was 157 Da, the expected result for two peptides differing in composition by a glutamic acid residue in the form of its ethyl ester. Fig. 2A shows the mass spectrum recorded on the earlier HPLC fraction, fraction 9. The (M+H)+ ion at m/z (average mass value) 1738.6 differs in mass from 1881.8 by 143 Da and thus corresponds to the next lower homolog of the ion series observed in Fig. 2B. Signals in Fig. 2A at m/z (average mass) 1%1.8, 2105.1, 2248.4, and 2390.2 are related to each other by an incremental mass shift of 143 Da and to the corresponding first four members of the family of ions observed in Fig. 2B by the addition of 80 Da. The latter incremental mass shift is that expected for a peptide that has been phosphorylated. The appearance offragment ions in the spectrum at a position 98 Da below that of the corresponding (M+H)+ ions (loss of (MP + H)++ Glu

(MW+H)++2Glu

1961.8

2105.1

CO) z uJ

I-

o-)

uJ

(M +H)+

(MP + H)~+ 3GIu (Mp+H)++4GIu -' ''X''2390-2

®24. 224.4 °("+H)+ @ Al e 1738.6

w

2000

1800

2200

2400

M/Z

1881.8

z

(M+H)++3GIu

(M+H)++Ghu

o B 0

w

~~~~~~~~2167.9

(+H~4~

(M+ H)++2GIu 2024.7

~~~~~~~~~~~~2311.3

0 pH -5.6-

-5.5-

_5A-5-

>

FIG. 1. IEF gel of bovine brain tubulin (lane 1) and rat

brain tubulin (lane 2). (Left) Gel is stained with Coomassie blue (CB) and illustrates the heterogeneity observed in brain tubulin samples. (Right) Gel is stained with monoclonal antibody TuJ1 (J1) to visualize isoforms of the class III

isotype.

3-tubulin

+ H)+5GMu ~~~~~~~~~~~~~~~(M

2454.7

wl 0--

1800

2000

2200 M/Z

2400

FIG. 2. Mass spectra recorded on two HPLC fractions containing methyl esters of the posttranslationally modified C-terminal CNBr fragment derived from neuronal class III (3-tubulin. Signals labeled P are due to the loss of phosphoric acid from (M+H)+ ions and occur at m/z 1863.6, 2107.0, and 2250.2. (A) Fraction 9. (B) Fraction 12.

Proc. Natl. Acad. Sci. USA 88 (1991)

Chemistry: Alexander et al. H3PO4) supports this conclusion and sugg(ests that the phosphate groups are attached to serine or thrceonine rather than to tyrosine. To deduce the amino acid sequences of tl he above peptides, additional samples of the HPLC fractions were subjected to analysis by the process of collision-activrated dissociation (CAD) on the triple quadrupole mass spe(ctrometer. In this experiment the (M+H)+ ions corresponiding to a single peptide in the mixture were selected with t]he first mass filter of the instrument and transmitted to a secorid chamber where they underwent multiple collisions with a g.;aseous mixture of argon and xenon atoms and suffered fragn nentation more or less randomly at the various amide bonds irn the peptide. The resulting collection of fragment ions, many of which differ in length by a single amino acid residue, were then subjected to mass analysis in a second mass filter. 0O ily seconds were required to generate this type of spectrum frrom each peptide in a particular HPLC fraction. Shown in Fig. 3 is the CAD mass spectrurn recorded on the (M+H)+ ions at m/z 1738.8 (average mass)). The amino acid sequence YEDDEEESEAQGPK was de duced from this spectrum. Predicted monoisotopic m/z val ues, for fragment ions of type y (28, 29), all of which contaiin the C-terminal residue plus 1, 2, 3, etc., additional resin dues, are shown below the sequence in Fig. 3. Those observe d in the spectrum are underlined. Subtraction of m/z values Jfor any two fragments that differ by a single amino acid1, NHCH(R)CO, generates a value that specifies the mass andIthustheidentity of the extra residue in the larger fragment. Fragment ions of type y allow all 14 residues in the peptide Ito be specified. Shown above the peptide sequence in| Fig. 3 are the predicted monoisotopic m/z values for fragrment ions of type b (28, 29), all of which contain the N termin ius of the peptide plus 1, 2, 3, etc., additional residues. Those observed in the spectrum are underlined. Subtraction of m/z values for

1Ser

l40 50

'X20b3

IC

w~~~~~~Y

30a

XIO

Y4

~~~Y2

650.

a

W

10

4

LL. -A

c

a lo'O.

c

._

IX -15

YB

10-

.0


~ 41 6

a)

CAD spectrum (Fig. 4 A

the above data provide strong evidence that the peptides

66......

I I.........

200

100

Unaccounted for by the above data are amino acid residues having masses of 286 and 429 units in the peptides showing (M+H)+ ions at m/z 1881.8 and 2024.7, respectively. The missing masses correspond to that expected for the methyl esters of two and three glutamic acid residues. Since neither

Of type b then accounts for series of ions labeled

2

W,

spectrum.

chain.

j/jbj

i0-

fragments of type b that differ by a single amino acid facilitates assignment of residues 3-14 and provides confirmation of the sequence deduced from fragment ions of type y. Most of the unlabeled ions in the spectrum result from the loss of small molecules such as H20 and NH3 from ions of either type b or y. A multistep fragmentation process leads to the formation of ions of type b derived from the peptide fragment containing residues 3-14, DDEEESEAQGPK (28). Ions of this type are labeled in the spectrum with the appropriate amino acid sequence contained in each fragment. The primary structure deduced from Fig. 3 corresponds to the C terminus of class III 13-tubulin. Residue 10 in this peptide is not conserved. Class III 8-tubulins from other sources contain arginine, serine, or a deletion at this position (30). Our results indicate that residue 10 in the C-terminal CNBr fragment from the bovine brain isoform is alanine. Fig. 4 contains CAD spectra recorded on the first two members, m/z 1881.8 and 2024.7, of the series of (M+H)+ ions observed in Fig. 2B. Common to both spectra is a set of signals with m/z values that are identical to the first 12 ions of type y found in Fig. 3. Accordingly, we conclude that all three peptides contain the C-terminal sequence DDEEESEAQGPK. Presence of the series of type b ions derived from the internal peptide fragment DDEEESEAQGPK in all three spectra provides additional support for this assignment. That the three peptides contain tyrosine as the N-terminal residue is indicated by the appearance of abundant fragment ions (type Y13) resulting from the loss of 163 mass units from the (M+H)+ ions in each

b or y formed by cleavage of the amide bond between these extra residues, it seemed likely that the additional glutamic acid residues were not part of the linear peptide chain. Appearance of a series of fragment ions labeled as bn-E in Fig. 4B provides additional support for this conclusion. Loss of a single glutamic acid residue as the intact dimethyl ester from the (M+H)+ ion explains the formation of m/z 1849 in Fig. 4B. This could only occur if this residue were part of a side

Ml b 1A24 1401 11738 07 i asp Asp As Gu Glu Glu Glu Ser GluIu As CGin Gly Pro Lys Tyr Glu Asp 1571432=1174103 74 E All I444 = 20 Al y

164T

4687

second

cycle of Edman degradation showed only

ions at

m/z

1567 and 1432. These

masses

two

(M+H)+

correspond to the

sequence DDEEESEAQGPK with and without addition of phenylisothiocyanate to the side chain of the C-terminal lysine residue. Note that the three (M+H) ions observed after the first Edman cycle were shifted to lower mass by 286, 429, and 572 Da. These are the values expected for loss of two, three, and

4688

Chemistry: Alexander et al.

Proc. Natl. Acad. Sci. USA 88 (1991)

A ; AD:''9 7Q5 7Q 51 lg ~d 5 54 29.4 '172a 1.42 1.624 .88 AS,'. ASD Glu GIL; Glu Ser GITJ Ala G,!n Gly Pro Lvs 13 r, 11741 1 7 i.. 30 117 744 65 514 444 ;31 2556 I.

l-3618

Ix40

100

1X7Z

/Da

x1O

Y|

80

40

E-

Y

LU_

O

20

0-oi 100 200 300 C-o

_

xIO

100 -Q ct

400 x30

500

600

Y,~~~~cu

80

LUL

60 LL

40

If)

0

LU

uiul 8 j Lj

LU

-0

LU 0

0

700

800

a

900

b

y-r

1000

1100

1200 AlO

100

-

4.0

(M+H)+

8 0-

18 8 08

60'

bb i

12.0

1 6.0 MIN.

20.0

24.0

28.0

FIG. 5. Data were obtained from the first five cycles of automated Edman degradation performed on a mixture of mono-, di-, and tri-glutamylated C-terminal peptides obtained from /3-tubulin.

Y 3 b ,.

b,,.

8.0

LAA.

FIG. 4. CAD mass spectra recorded on the (M+H)+ ions at m/z 1881.8 (A) and 2024.7 (B) in Fig. 2B.

in cycle 2. This is the expected result for a glutamate residue that is glutamylated on its side chain since the resulting PTH derivative is known to be sufficiently polar to remain bound to the filter of the sequenator (9). Cycles 3-14 release a single PTH derivative corresponding to the amino acids DDEEESEAQGPK from each of the three peptides present in the mixture. All of the above data support the conclusion that it is the glutamate residue at position two in the C-terminal CNBr fragment that is polyglutamylated in (3-tubulin. Location of the phosphorylated residue in the class III (3-tubulin, CNBr fragments that are both glutamylated and phosphorylated (Fig. 2A), was accomplished by using the technique of electrospray ionization (31). Peptide ions generated by this latter technique usually carry a proton on each basic residue in the sample. As a result, the C-terminal fragment of class III /3-tubulin is expected to appear as a doubly charged ion corresponding to (M+2H)2+. Shown in Fig. 6 is the CAD spectrum recorded on the (M+2H)2+ ions (m/z 918) generated from the free acid corresponding to the methyl ester of the monoglutamylated monophosphorylated species shown in Fig. 2A [(M+H)+ = 1%1.8]. Fragmentation observed in this spectrum requires that the phosphate be present on Ser-444. Predicted monoisotopic m/z values for fragment ions of type b and y for this structure are shown at the top of Fig. 6. Those observed are underlined. Note that fragments of type Y1-Y6 define the partial sequence EAQGPK. The next ion in the series y7 is found at m/z 796, an increment of 167 Da. This is the expected value for a phosphoserine residue. Consistent with this assignment is the appearance of an ion at m/z 698 formed by the loss of phosphoric acid (98 Da) from y7. Residues 8-12 are assigned from ions of type y or y-H3PO4. These all appear at the expected mass increments. We conclude that the monoglutamylated C-terminal CNBr fragment is phosphorylated at Ser-444.

four glutamate methyl ester residues, respectively, from the three peptides in the second Edman cycle. Data obtained from the first 5 of 14 cycles of automated Edman degradation performed on the above mixture of three glutamylated peptides (underivatized) are displayed in Fig. 5. Only a tyrosine phenylthiohydantoin (PTH) derivative is released in cycle 1. No PTH amino acid derivative is detected

DISCUSSION Class III 83-tubulin in brain acquires increasing amounts of negative charge near the C terminus during neural differentiation. We demonstrate that polyglutamylation and phosphorylation account for most, if not all, of this heterogeneity. Molecular mass measurements indicate that up to six gluta-

13 00

1400

1600

1500

1700

!900

1800

M/Z 28 3 iiaP ; - C; r a; .L". 1K30

dxO

;xi5

100

80'Y

60 5 Y2

Ys

Yyfl

Y

40~~~~ 20

l_0

2061D30 2ac 30 '43504 Q -4 Q

-5 51

x010 V 0

}

4~~~~~ 0~4I~

C: 3 _0 C1)

__-

60>0

6 700_7 [x20~

1008 0 '!

75

800 -

Q

900

L

950 1000 1050 ix 50

~:,2

Y

LU

LL .iU

60

240 110_o 1150 1200 1250 1300 1350 1400 1450 1500 1550 100-

W+

20 2

b2

(M+H)

2 D

0

600 1650 1700

750 1800 1850 1900 1950 2000 2050

M/Z

Chemistry: Alexander et al. -q

Proc. Nadl. Acad. Sci. USA 88 (1991)

This research was supported by National Science Foundation Grants CHE8710416 (to R.J.S., D.F.H., T.L.M., L.I.R., and A.F.) and CHE-8618780 (to D.F.H.) and National Institutes of Health Grants NS21142 (to A.F.) and GM37537 (to D.F.H.).

7 .e A2.39, 11206 _ 1335 4065-33:3 19 .F-92 6. C.-i ,-;;., 23.r_ Se-P Czi.. Ac (3:.> Kit P.: ,, As.,s

-

;;z Ask.

, 1

'x4

xlO

100

Y

_

60

600-

w wY5 5

M

AL LL

b3

0

b2

Y6

40

[x220i6o

360

26 0

X5

43

M

zi100-

400

5u00

600

Y

y'

iMH*+1xI0

+O

Lii a)

Yq-o

60-

8: 0 40

hQ7

Y

-I

b5

_

20

1200

00

80-

4689

)I2

Y12

40

20

1300

1400

1500

1600

1700

J800

M/Z FIG. 6. Electrospray ionization/CAD mass spectrum recorded ion derived from the free acid corresponding to the methyl ester of the monoglutamylated and monophosphorylated C-terminal peptide in Fig. 2A; (M+H)+ = 1961.8.

on the (M+2H)2+

mate residues can be added to the C-terminal 14 amino acids

in (-tubulin. All of the polyglutamylated species can also exist with phosphate attached to the same C-terminal 14 residue peptide. IEF gels and mass spectrometry indicate that >85% of class III 3-tubullin in adult bovine brain is modified by posttranslational glutamylation. Approximately 10% of the glutamylated isoforms are also phosphorylated. Within the family of glutamylated isoforms, the abundance of a particular species decreases with increasing degree of glutamylation. CAD mass spectrometry and subtractive- and automatedEdman degradation experiments confirm that the site of attachment for the first three glutamates is at Glu-438. The monoglutamylated species is also found with phosphate attached to Ser444. Available sample quantities have so far precluded complete characterization of the higher homologs within the phosphorylated and nonphosphorylated isoform's. Preliminary experiments on additional HPLC fractions containing C-terminal CNBr fragments from bovine (-tubulin indicate that polyglutamylated species can also be phosphorylated at Tyr-437 or Ser-444 or both (data not shown). The complete structural characterization of these isoforms will appear elsewhere. Yet to be defined is the nature of the linkage connecting the residues in the polyglutamyl side chains. The-first glutamate residue must be'linked to the 'y-carboxyl group of Glu-438. Additional glutamic acid residues could then be attached through amide bonds to their a- or ycarboxyl groups. Enzymes exist for the polyglutamylation of tetrahydrofolate in both mammalian liver (32) and in Escherichia coli (33). In the latter organism, Glu-2 and -3 are attached through the -carboxyl group of the preceding residue and Glu-4 to -8 are attached through the a-carboxyl group of the preceding residue (33).

1. Dustin, P. (1984) Microtubules (Springer, New York). 2. Field, D. J., Collins, R. A. & Lee, J. C. (1984) Proc. Natl. Acad. Sci. USA 81, 4041-4045. 3. Wolff, A., Denoulet, P. & Jeantet, C. (1982) Neurosci. Lett. 31, 323-328. 4. Lee, J. C., Field, D. J., George, H. J. & Head, J. (1986) Ann. N. Y. Acad. Sci. 466, 111-128. 5. Sullivan, K. F. (1988) Annu. Rev. Cell Biol. 4, 687-716. 6. L'Hernault, S. W. & Rosenbaum, J. L. (1985) Biochemistry 24, 473-478. 7. Piperno, G. & Fuller, M. (1985) J. Cell Biol. 101, 2085-2094. 8. Black, M. M. & Keyser, P. (1986) J. Neurosci. 7, 1833-1842. 9. Edde, B., Rossier, J., LeCaer, J. P., Desbruyeres, E., Gros, F. & Denoulet, P. (1990) Science 247, 83-85. 10. Barra, H. S., Arce, C. A., Rodriquez, J. A. & Caputto, R. (1974) Biochem. Biophys. Res. Comnmun. 60, 1384-1390. 11. Raybin, D. & Flavin, M. (1975) Biochem. Biophys. Res. Commun. 65, 1088-1095. 12. Edde, B., Denoulet, P., De Nechaud, B., Koulakoff, A., Verwald-Netter, Y. & Gros, F. (1989) Biol. Cell 65, 109-117. 13. Luduena, R. F., Zimmermann, H. & Little, M. (1988) FEBS Lett. 230, 142-146. 14. Gard, D. J. & Kirschner, M. W. (1985) J. Cell Biol. 100, 746-774. 15. Eipper, B. A. (1974) J. Biol. Chem. 249, 1407-1416. 16. Forgue, S. T. & Dahl, J. L. (1979) J. Neurochem. 32, 10151025. i7. Lewis, S. A. & Cowan, N. J. (1988) J. Cell Biol. 106, 20232033. 18. Lee, M. K., Tuttle, J. B., Rebhun, L. I., Cleveland, D. W. & Franffiurter, A. (1990) Cell Motil. Cytoskel. 17, 118-132. 19. Lee, M. K., Rebhun, L. I. & Frankfurter, A. (1990) Proc. Natl. Acad. Sci. USA 87, 7195-7199. 20. Vera, J. C., Rivas, C. L. & Maccioni, R. B. (1989) Biochemistry 23, 333-339. 21. Littauer, U. Z., Giveon, D., Thieraur, M., Ginsburg, I. & Posting, H. (1986) Proc. Natl. Acad. Sci. USA 83, 1716217166. 22. Serrano, L., de la Torre, J., Maccioni, R. B. & Avila, J. (1984) Proc. Natl. Acad. Sci. USA 81, 5989-5993. 23. Serrano, L., Avila, J. & Maccioni, R. B. (1984) Biochemistry 23, 4675-4681. 24. Williams, R. C., Jr., & Lee, J. C. (1982) Methods Enzymol. 85,

376-385. 25. Griffin, P. R., Kumar, S., Shabanowitz, J., Charbonneau, H, Nanikung, P. C., Walsh, K. A., Hunt, D. F. & Petra, P. H. (1989) J. Biol. -Chem. 264, 19066-19075. 26. Hunt, D. F., Yates, J. R., III, Shabanowitz, J., Winston, S., & Hauer, C. R. (1986) Proc. Natl. Acad. Sci. USA 83, 6233-6237. 27. Erickson, A. K., Payne, D. M., Martino, P. A., Rossomando, A. J., Shabanowitz, J., Weber, M. J., Hunt, D. F. & Sturgill, T. W.. (1990) J. Biol. Chem. 265, 19728-19735.. 28. Hunt, D. F., Shabanowitz, J., Yates, J. k., III, Zhu, N. Z., Russell, D. H. & Castro, M. E. (1987) Proc. Natl. Acad. Sci. USA 84, 620-623. 29. Biemann, K. (1988) Biomed. Environ. Mass Spectrom. 16, 99-111. 30. Sullivan, K. F. & Cleveland, D. W. (1986) Proc. Natl. Acad. Sci. USA 83, 4327-4331. 31. Smith, R. D., Loo, J. A., Edmonds, C. G., Baringa, C. J. & Udseth, H. R. (1990) Anal. Chem. 62, 882-899. 32. Cichowicz, D. J. & Shane, B. (1987) Biochemistry 26,,504-512. 33. Ferone, R., Hanlon, M. H., Singer, S. C. & Hunt, D. F. (1986) J. Biol. Chem. 261, 16356-16362.