The Four Terminal Components of the Complement System AreC-Mannosylated on Multiple Tryptophan Residues*

C-Mannosylation is a unique form of protein glycosylation, involving the C-glycosidic attachment of a mannosyl residue to the indole moiety of Trp. In the two examples found so far, human RNase 2 and interleukin-12, only the first Trp in the recognition motif WXXW is specificallyC-mannosylated. To establish the generality of proteinC-mannosylation, and to learn more about its mechanism, the terminal components of the human complement system (C6, C7, C8,and C9), which contain multiple and complex recognition motifs, were examined. Together with C5b they form the cytolytic agent, the membrane attack complex. These are the first proteins that areC-mannosylated on more than one Trp residue as follows: six in C6, four in C7, C8α, and C8β, and two in C9. Thus, from the 113 Trp residues in the complete membrane attack complex, 50 were found to undergo C-mannosylation. The other important finding is that in C6, C7, C8, and C9 Trp residues without a second Trp (or another aromatic residue) at the +3 position can beC-mannosylated. This shows that they must contain an additional C-mannosylation signal. Whether this is encoded in the primary or tertiary structure is presently unknown. Finally, all modified Trp residues are part of the highly conserved core of the thrombospondin type 1 repeats present in these proteins. Since this module has been found in a large number of other proteins, the results suggest further candidates for C-mannosylation.

The modification of proteins by covalent attachment of carbohydrate is a common feature of both intra-and extracellular proteins (1,2). In particular proteins that contain N-or Olinked oligosaccharides have been known for a relatively long time (3,4), and many examples are known (1). Recently, a new type of glycosylation has been discovered in human RNase 2, which differs fundamentally from N-and O-glycosylation with respect to the protein carbohydrate linkage. It involves the C-glycosidic attachment of an ␣-mannopyranosyl residue to the C-2 atom of the tryptophan side chain (5-7) (Scheme 1). This modification has been shown to be catalyzed by a microsomeassociated transferase, which C-mannosylates the first Trp residue in the recognition sequence -WXXW- (8,9). The enzyme uses dolichylphosphate mannose as the sugar donor, and its activity has been found in mammals, birds, amphibians, and fish. Furthermore, it could also be detected in most mouse organs that were examined (10,11). In addition to the widespread distribution of the C-mannosyltransferase, its recognition motif, WXXW, has been found in over 300 secreted mammalian proteins (9). This strongly suggests that Cmannosylation is a more common form of glycosylation than indicated by the fact that only two C-mannosylated proteins have been characterized so far, i.e. human RNase 2 (5, 6) and IL-12 1 (12). These two proteins contain a single C-mannosylation motif of the simple structure WXXW. To further our insight into the process of protein C-mannosylation, we examine here proteins that contain multiple and more complex Cmannosylation motifs.
The terminal complement proteins C6, C7, C8, and C9 are attractive candidates for studying protein C-mannosylation because of the following: (i) they contain complex C-mannosylation motifs (WXXWXXW) as part of their thrombospondin type 1 repeats (TSR modules); (ii) they possess up to three TSR modules; and (iii) they have a clearly defined physiological function (see below). The latter is important, since so far a function for C-mannosylation of Trp residues has not been established.
The complement, which forms a first line of defense against microbial infections, can be activated by three different pathways as follows: the alternative, the classical, and the mannose-binding lectin-mediated pathway (13,14). Each of these results in the formation of the membrane attack complex (MAC), the actual lytic component of complement (15). The assembly of MAC begins with the formation of C5b by C5 convertase at the target membrane, followed by the sequential addition of the plasma proteins C6, C7, C8, and C9 (see Reaction 1).
C5b-9 n forms a trans-membrane channel, whose exact size can vary, depending on the value of n. The association process has been examined in detail both on natural as well as artificial membranes (16 -19). The value of n for the complete MAC ranges from 12 to 18 with an average of approximately 16 (see Ref. 19 for a discussion), which yields a calculated mass of 1.4 ϫ 10 6 Da.
Here we have examined the C-mannosylation of Trp residues in the TSR modules of C6, C7, C8␣, C8␤, and C9 by analyzing * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  peptides by ESIMS, Edman degradation, and NMR. It was found that C-mannosylation can occur at 50 out of 113 Trp residues in the MAC and that also Trp residues without a second Trp at position ϩ3 can undergo this modification.

EXPERIMENTAL PROCEDURES
Materials-Human complement component C6 was purchased from Sigma, and components C7, C8, and C9 were from Advanced Research Technologies, San Diego, CA. Component C9 used in the protein chemical analyses was a generous gift of Dr. J. Tschopp, University of Lausanne, Switzerland.
Protein Chemistry-Proteins were reduced and carboxamidomethylated as described previously (20). Tryptic digestion of the proteins was performed twice in 50 mM Tris-HCl, pH 8.0, for 2 h at 37°C at an enzyme to substrate ratio of 1/50 (C7 and C8) or 1/100 (C6 and C9). Subsequently, C7 and C8 were digested for 4 h with endoproteinase Asp-N (1/50; Roche Molecular Biochemicals) at 37°C. The amount of protein per digest was 50 -75 g for C6, C7, and C8, and two digests were made. The total amount of C9 digested was 7 mg. Cleavage of isolated peptides with 200 ng of chymotrypsin was performed for 2 h at 37°C.
Digests were fractionated by C 8 reversed phase LC-ESIMS as described (10). If necessary final purification was achieved by LC-ESIMS, using 25 mM NH 4 acetate, pH 6.0, containing 2 or 80% CH 3 CN, as buffers A and B, respectively.
NMR Spectroscopy-NMR spectra were measured on a Varian Unity Plus spectrometer operating at a 1 H frequency of 600 MHz. Susceptibility matched NMR tubes (Shygemi) with sample volumes of 220 l were used. Clean TOCSY spectra were recorded with a mixing time of 40 and 80 ms. The spectra were folded once in t 1 to increase the resolution. The spectra were recorded with 192 increments and 160 transients. Experiments of samples in 90% H 2 O, 10% D 2 O were recorded using the Watergate method. Spectra were referenced against acetone (␦ ϭ 2.225 ppm (23)).
Peptide Nomenclature-The numbering of the mature protein has been used here as follows: C6 (24), C7 (25), and C8␣ (26) as corrected in Swiss-Prot entry P07357 and C8␤ (27) and C9 (28). Tryptic peptides have been numbered according to their occurrence in the mature polypeptide. Peptides that resulted from tryptic digestion have been labeled with the prefix "T," whereas subsequent digestion with endoproteinase Asp-N and chymotrypsin have been indicated with the affix "D" and "Ch," respectively.
In Vitro C-Mannosylation-The peptide Ac-WAKW-NH 2 , the C6derived peptide Ac-LLGDFGPWSDCD-NH 2 , and the C9-derived peptide Ac-RMSPWSEWSQCD-NH 2 were C-mannosylated in vitro using salt-washed porcine liver microsomes as the source of transferase (12). The reaction mixture contained in a total volume of 24 l the following: 0.9 mM peptide, 22.5 pmol of Dol-P-[2-3 H]Man (5.6 Ci/mmol), porcine liver microsomes (77 g of protein), 20 mM Hepes-NaOH, pH 7.2, 0.2% (w/v) Triton X-100, 2 g/ml benzamidine, 5 g/ml pepstatin, 5 g/ml leupeptin, 2 mM EGTA. The reaction was performed for 30 min at 37°C and stopped by the addition of 2 ml chloroform/methanol 3:2 (v/v) and 0.48 ml of water. After centrifugation the upper, aqueous, phase contained the peptide. The radioactivity in 0.2 ml of the upper phase was determined by scintillation counting (8). For preparative purposes, 10 reactions were performed for 42 h at room temperature (12). The radiolabeled peptide was reduced, carboxamidomethylated, and isolated by C 18 reversed phase HPLC as described (5). Final purification was achieved after tryptic cleavage and fractionation by C 8 reversed phase LC-MS (12).

RESULTS
Component C9 -The fractionation of the tryptic peptides from carboxamidomethylated C9 by reversed phase LC-ESIMS is shown in Fig. 1A. The MS data were extracted for the theoretical mass of the peptide(s) containing a C-mannosylation motif, with or without the mass of a hexosyl residue (162 Da) added. Two forms of peptide T2, comprising residues 24 -38 of the mature protein (for nomenclature of peptides and numbering system see "Experimental Procedures"), were detected. Peptide T2-1, eluting at 41.1 min, had a molecular mass of 2101 Da (Fig. 1B, upper panel), which was 162 Da heavier than expected from the cDNA sequence (28,29). This suggested the presence of one hexosyl residue. ESIMSMS of this fraction yielded a nearly complete series of y ions, confirming its sequence and localizing the hexosyl residue to Trp-27 ( Fig. 2A and Table I). The spectrum showed secondary neutral losses of 120 Da from [Mϩ2H] 2ϩ and y13 2ϩ . This fragmentation has typically been observed with aromatic C-glycosides (30,31) and has previously been found with (C 2 -Man-)Trp in RNase 2 (5) and IL-12 (12). Edman degradation of this peptide yielded at the fourth cycle a PTH-derivative that comigrated with PTH-(C 2 -Man-)Trp obtained from human RNase 2 (Fig. 1C), a protein whose C-mannosylation has previously been extensively characterized (5)(6)(7).
To confirm the identity of the mannosyl residue, the peptide was analyzed by NMR spectroscopy. The region of the spectrum that is just upfield from the water resonance is expected to be empty for a random coil 15-mer peptide. It would only show signals of upfield shifted H␣ resonances in case the peptide would obtain stable turns or a ␤-sheet structure. However, the amide resonances exchange very fast with D 2 O, and the observed J HNH␣ couplings are all around the random coil value of 6 -7 Hz, indicating that the peptide was not structured (data not shown). This allows the unequivocal identification of anomeric sugar protons, which resonate in this region. The anomeric part of the spectrum of the pentadecapeptide T2-1 is shown in Fig. 3B and is compared with that of the C-mannosylated peptide isolated from IL-12 ( Fig. 3A (12)). It is evident from this comparison that the doublet signal that resonates at 5.18 ppm and that was split by a J coupling of 7.4 Hz is the anomeric H1Ј resonance of (C 2 -Man-)Trp. This interpretation was confirmed by the disappearance of one of the H2 resonances of Trp and by the assignments of the other resonances of the sugar moiety and of the Trp residue with the help of TOCSY experiments (Table II). An additional doublet signal at 5.27 ppm which was split by a J coupling of 6.5 Hz was observed. This signal has an intensity of only 15%. Minor components of the same magnitude were also observed for the aromatic protons of (C 2 -Man-)Trp. We therefore concluded that the extra anomeric resonances were caused by the cis-transisomerization of the peptide bond between Ser and Pro preceding the mannosylated Trp (32), rather than by a heterogeneity in the sugar moiety.
The identification of a second form of peptide T2 (T2-2 in  Table I). The PTH-derivatives observed at cycles 4 and 7 during Edman degradation of this peptide comigrated exactly with each other and with PTH-(C 2 -Man-)Trp from RNase 2 (Fig. 1D).
is shown in Fig. 3C. Interestingly, in comparison with sample T2-1, an additional doublet resonance was observed. The resonances at 5.20 and 5.18 ppm were split by J-couplings of 7.1 and 7.4 Hz, respectively. Fig. 3D shows part of the TOCSY spectrum, recorded with a mixing time of 80 ms. It is evident from this spectrum that these anomeric resonances are part of similar spin systems with approximately the same chemical shifts (Table II). In fact the TOCSY spectra are similar to those recorded of the peptides with only one glycosylated Trp. In the case of peptide T2-2 two indole H2 resonances were missing. Thus, it is concluded that in this particular case both Trp residues in the WXXW motif were C 2 -mannosylated. The signal at 5.27 ppm with an intensity of about 15%, caused by the cis-trans-isomerization of the Ser-Pro bond appeared in this spectrum as well.
Examination of the LC-ESIMS data did not reveal the presence of unmodified peptide T2. Furthermore, the other two Trp residues in component C9, at positions 176 and 436, were observed in the unmodified form only.
Component C8 -Component C8 is a heterotrimeric protein consisting of a disulfide-linked ␣-␥ dimer and a non-covalently associated ␤ subunit. Both the ␣ and ␤ subunit contain two TSR modules. To obtain peptides suitable for MS and sequence analysis, carboxamidomethylated C8 was digested first with trypsin and subsequently with endoproteinase Asp-N. Fractionation by reversed phase LC-ESIMS yielded two C-glycosylated peptides originating from the ␣-subunit, ␣T1-D (2155 Da; residues 1-26 with one hexosyl residue) and ␣T57 (2317 Da; residues 508 -522 with three hexosyl residues). The fraction containing ␣T1-D was further cleaved with chymotrypsin, yielding pure ␣T1-D-Ch (residues 10 -18) upon LC-ESIMS purification. Nanospray ESIMSMS of this peptide established that the sugar moiety was attached to Trp-14, whereas Trp-17 occurred in the unmodified form (Table I). The peptide was sequenced by Edman degradation, showing exact comigration of the PTH-derivative at cycle 5 with PTH-(C 2 -Man-)Trp from RNase 2 (Table I).
Peptide ␣T57 contained three hexosyl residues and was purified to apparent homogeneity by LC-ESIMS at pH 6.0 (data not shown). Its nanospray ESIMSMS spectrum and Edman degradation ( Table I) showed unequivocally that Trp-512, -515, and -518 were C-mannosylated. Consistent with this, no loss of 162 Da was observed in MSMS experiments with either ␣T1-D-Ch or ␣T57, which would have been evidence for O-linked hexosyl residues. Since only C-mannosylated forms of ␣T1-D and ␣T57 were detected in the chromatogram of the starting digest, it is concluded that these four Trp residues were fully modified. From the C8␤ chain two peptides, ␤T1-D (1942 Da, residues 10 -22 with two hexosyl residues) and ␤T59 (2010 Da, residues 494 -506 with two hexosyl residues) were detected and purified further by LC-ESIMS at pH 6.0 (data not shown). Both Trp-16 and -19 in ␤T1-D were modified, as was demonstrated by nano-ESIMSMS, which showed a continuous series of b and y ions covering both Trp residues, as well as 120-Da loss from a number of ions (Table I). These results were confirmed by Edman degradation (Table I). Peptide ␤T59 still contained a contaminating peptide (YILNTR), even after rechromatography. Its presence did not, however, interfere with further analysis. ESIMSMS localized the hexosyl residues on Trp-497 and -500, and 120-Da losses were observed from the parent as well as the y7 and y9 ions (Table I). Edman degradation confirmed Trp-494 to be unmodified and yielded PTH-(C 2 -Man-)Trp in cycles 4 and 7 (Table I). With neither of these peptides was any evidence for O-linked hexosyl residues obtained.
These results extend the unexpected finding in C9 that Trp residues without a Trp or another aromatic residue at position ϩ3 can be C-mannosylated.
Component C7-Of the two TSR modules in component C7 only the second one contains WXXW motifs. Nevertheless, a tryptic/Asp-N peptide from the first repeat was found that had a mass that was 162 Da larger than expected from the cDNA sequence (T1-D, 1897 Da, residues 9 -22). ESIMSMS of this peptide (Fig. 4A) positioned the substituent at Trp-14, and a C-glycosidic linkage was concluded from the 120-Da loss from the parent, the y10, and y11 ions. Further evidence was obtained by Edman degradation of the peptide that had been purified by reversed phase LC-ESIMS at pH 6.0. The PTHderivative produced in cycle 6 comigrated with PTH-(C 2 -Man-)-Trp from RNase 2 (Table I).
From the second repeat four different C-glycosylated peptides were obtained. Approximately equal amounts of unmodified peptide (T50-D-1, 3193 Da, residues 465-494) and monohexosylated peptide (T50-D-2, 3355 Da) were detected. A nearly complete y and b ion series in the ESIMSMS analysis of T50-D-1 confirmed its sequence and the absence of modification (Table I). Similarly, the spectrum of T50-D-2 confirmed the structure and established the site of modification at Trp-484, with 120-Da losses from the parent and the doubly charged y13, y17, and y23 and y24 ions (Table I). Two peptides containing two modified Trp residues were found as minor fractions, T50-D-3 and T50-D-4 with a mass of 3517 Da. ESIMSMS analysis established that in T50-D-3 Trp-481 and -484 were modified, whereas T50-D-4 contained modified Trp-484 and -487 (Table I). The amount of these two peptides was too small for further purification and analysis by Edman degradation.
Component C6 -C6 contains three TSRs, with WXXW motifs present in the first and third repeat (residues 8 -11 and 547-553). In the second repeat a single Trp is present but no C-mannosylation motif. The C-mannosylation of Trp-14 in C7, however, warranted the analysis of this region of C6.
Fractionation of the tryptic digest of C6 yielded a variety of peptides that originated from the TSR modules. This complexity was caused by the partial modification of most Trp residues. Peptide T1 (residues 1-16) from the first repeat was detected in two chromatographically separated fractions. Each of these was digested with chymotrypsin and purified to apparent homogeneity by reversed phase LC-ESIMS. The ions observed in the collision-induced dissociation spectrum of T1-Ch-1 (853 Da; residues 7-11 containing 1 hexosyl residue) demonstrated Trp-8 to contain the hexosyl residue. Since the parent, and the b3 and b4 ions showed the typical 120 Da loss, the presence of a C-glycosidic linkage was concluded. These results were confirmed by Edman degradation (Table I). The ESIMSMS analysis of T1-Ch-2 (1578 Da, residues 7-16 containing 2 hexosyl residues) is shown in Fig. 4B. The spectrum illustrates the complexity that may arise due to the presence of more than one (C 2 -Man)-Trp residue. Loss of 120 Da, as well as multiple losses of H 2 O, were observed twice from the doubly charged parent ion. Furthermore, the data clearly show that both Trp-8 and -11 were modified with a hexosyl residue. Edman degradation confirmed the sequence of the peptide and provided chromatographic evidence for the presence of (C 2 -Man-)Trp at cycles 2 and 5 (Table I).
Peptide T8, which contains Trp-69 but no recognition motif, was detected mainly in the unmodified form (2653 Da, residues 57-78). Since the mass of this peptide as well as its ESIMSMS spectrum fully established its structure, no Edman degradation was carried out. The peptide with modified Trp-69 was a minor fraction. ESIMSMS of the unpurified chymotryptic fragment (T8-Ch, 2058 Da, residues 63-78) confirmed its identity and provided clear evidence for its C-hexosylation. The observations that its mass and that of the y11 and y12 ions was 162 Da higher than expected from the cDNA showed that Trp-69 contains the hexosyl residue (Table I). The loss of 120 Da typical for aromatic C-glycosides was observed for the parent, the y11, and y12 ions. Because the peptide was obtained in very low amounts and impure, no attempt was made to analyze it by Edman degradation.
Peptide T69, which contains Trp-547, -550, and -553, was     d y and b ions that were observed have been indicated with an arrow above and below the sequence, respectively. The ions from which the 120-Da loss (that defines the presence of the C-mannose linkage) has been observed are indicated by identifying the fragment (e.g. y7, y8, b3, b4, and p ϭ parental) above or below the sequence. W*, (C 2 -Man-)Trp. Cysteine residues have been carboxamidomethylated. e The elution time of the PTH-derivative of a modified Trp residue obtained by Edman degradation was compared to that of PTH-(C 2 -Man-)Trp from the peptide FT(C 2 -Man-)WAQW from RNase 2 (5), which was sequenced immediately before the sample. The differences in elution time between samples are due to aging or replacement of the column and to adjustments in the buffer compositions to optimize the separation of the other PTH-derivatives.
f ND, not determined. detected in four separate chromatographic fractions. The major one contained modified Trp-547 and -550, as shown by ESIMSMS analysis and Edman degradation (Table I)  2890 Da) or all three Trp residues modified (T69 -3, 3052 Da) were detected. The amount of these peptides did not allow further purification and characterization by Edman degradation. Nevertheless, their analysis by ESIMSMS was consistent with C-hexosylation as summarized in Table I. A small amount of completely unmodified T69 was detected as well (T69 -1, 2566 Da). Since its primary structure was fully established by ESIMSMS, no Edman degradation was done.
Component C6 contains six further Trp residues. The LC-ESIMS data of the first fractionation of the digest was extracted for the masses of each of the expected peptides assuming either the presence of Trp or (C 2 -Man-)Trp. All peptides were found in the unmodified form only.
A summary of the data obtained for all four terminal components, including an estimate of the stoichiometry of modification, is shown in Fig. 5.
In Vitro C-Mannosylation-C-Mannosylation in human RNase 2 and IL-12 can be reproduced in vitro using peptides containing the motif WXXW and a microsome-associated transferase (8,12). The modification in these cases is completely restricted to the first Trp, and replacement of the second by Ala completely abolishes the modification in vitro and in vivo (8,9). In contrast, the results summarized in Fig. 5 show that all terminal components of complement contain one or more Cmannosylated Trp residues with a Cys rather than Trp or another aromatic residue at the ϩ3 position. It was therefore of interest to examine the in vitro C-mannosylation of such peptides.
The peptide Ac-LLGDFGPWSDCD-NH 2 derived from C6 and Ac-RMSPWSEWSQCD-NH 2 derived from C9 were incubated with porcine liver microsomes and Dol-P-[2-3 H]Man. The general acceptor peptide Ac-WAKW-NH 2 was used as a positive control (12). As shown in Fig. 6A no C-mannosylation could be detected in the peptide from C6, whereas the one from C9 was nearly as effective an acceptor as the control peptide. To determine the site(s) of modification the radiolabeled C9 peptide was isolated from the reaction mixture by C 18 reversed phase HPLC (Fig. 6B). This yielded a major peak of radioactivity eluting at 19.5 min which contained a peptide with a mass expected for the mono-hexosylated peptide from C9 (1771 Da). The minor peak eluting at 18.5 min contained 8% of the recovered radioactivity. ESIMSMS analysis demonstrated it to be the same peptide, but with oxidized Met at position 2. The major peptide was cleaved with trypsin to remove Ac-Arg and yielded a single radioactive peak upon purification by rechromatography on a C 8 HPLC column (data not shown). The purified peptide had a mass of 1574 Da, corresponding to MSP-   (12), and RNase 2 (5) The notation of the atoms is as depicted in Scheme 1.  WSEWSQ(carboxamidomethyl)CD-NH 2 plus an additional hexosyl residue. The ESIMSMS spectrum of this peptide is depicted in Fig. 6C. The m/z values of the b4, b5, and b6 ions identified the first Trp residue as the attachment site of the mannosyl residue. Furthermore, the values of the y5, y6, and y7 ions provided evidence for the unmodified second Trp. Importantly, y ions at m/z 856, 985, and 1072, which would have indicated C-mannosylation of the second Trp, were conspicuously absent in the spectrum (arrows in Fig. 6C). These results were confirmed by Edman degradation of the radiolabeled peptide. A peak of radioactivity was observed in the 4th cycle (the position of the first Trp) but not in cycle 7 (data not shown). Thus, in vitro C-mannosylation of the peptide from C9 was restricted to the first Trp in the WXXW motif.

DISCUSSION
The results presented here demonstrate that C-mannosylation of Trp residues is not a rare event restricted to RNase 2 and IL-12. Nearly all Trp residues in the TSR modules of the four terminal components of complement can be C-mannosylated (Fig. 5). The most complete analysis was carried out on component C9, where the presence of the (C 2 -Man-)Trp residues at positions 27 and 30 was established by MS, Edman degradation, and NMR. The identity of the hexosyl residue was established by comparing chemical shifts and J-coupling data of the peptides from C9 with those of peptides obtained from RNase 2 and IL-12. The close agreement of these parameters showed that the chemical structure of residues 27 and 30 of C9 is identical to that of (C 2 -Man-)Trp present in RNase 2 and IL-12 (5,6,12). No NMR data are available for C-linked hexoses other than mannose. However, they are expected to have significantly different chemical shifts, connectivity patterns, and coupling constants (6). For the other three terminal components, (C 2 -Man-)Trp in the major peptides (modified fraction Ͼ0.1 mol/mol, Table I) was established by MS and comigration of its PTH-derivative with PTH-(C 2 -Man-)Trp from human RNase 2. No further NMR analysis was done for these, because of lack of sufficient amounts of protein. Furthermore, we feel confident that the MS and Edman degradation studies are sufficiently supported by NMR in the past (5-7, 12) and by the analysis of C9 (this paper). Therefore, NMR is only required in unusual new cases. Exact comigration of the PTH-derivatives of the modified Trp residues with authentic PTH-(C 2 -Man-)Trp was always observed (Table I). Due to the lack of synthetic standards and natural examples, the elution position of PTH-(C 2 -hexose)-Trp with hexoses other than mannose has not been established. It seems unlikely, however, that these would comigrate with PTH-(C 2 -Man-)Trp. The small peak eluting shortly after the main peak of PTH-(C 2 -Man-)Trp (* in Fig. 1, C and D) is a diastereomer (5). The good separation between these peaks suggests that the resolution of the system is sufficient to separate other hexose derivatives as well. The recent publication of a synthetic route for (C 2 -Man-)Trp will possibly also lead to the synthesis of the other hexosyl derivatives and provide a solution to this matter (33).
In Fig. 5 an estimate of the stoichiometry of modification has been indicated. These values were calculated as a fraction of the total area of the [M ϩ nH] nϩ ions (n ϭ 2 and 3) of a particular peptide in the original chromatogram of a digest. The lower panel shows the radioactivity in each fraction. The total recovery of radioactivity was 48%. C, the ESIMSMS spectrum of the purified radiolabeled C9-derived peptide. The loss of 120 Da from fragment ions and [M ϩ 2H] 2ϩ has been indicated with 120 and 60, respectively. The arrows indicate the position of y ions that would have appeared if the second Trp, rather than the first one, were C-mannosylated. amu, atomic mass units.
Differences in the efficiency of ionization of the differently modified forms of a peptide could affect the accuracy of intermediate values. The extreme values of 0 (no modified peptide found) and 1 (no unmodified peptide found) are unequivocal, however. Furthermore, in the case of C9 shown in Fig. 1A, where the two modified peptides were obtained in pure form from the original chromatogram, a good correlation between integration of the UV and MS data was observed.
C-Mannosylation is a significant feature of the MAC. The mature complex, C5b-C9 16 (19), has a calculated mass of approximately 1.4 ϫ 10 6 Da and contains a total of 113 Trp residues. Of these, 50 have been found to undergo C-mannosylation. Taking into account the degree of modification of each of them, and assuming a random distribution in MAC particles, we calculated that on average a MAC contains 34 (C 2 -Man-)Trp residues. It has been hypothesized, but not proven, that the TSR modules in the complement proteins are involved in adhesion to each other and perhaps in the assembly of the MAC (34). Some support for this has been obtained from the inhibition of the assembly process by a monoclonal antibody specific for the third TSR module in C6 (35). Whether the (C 2 -Man-)Trp residues play a role in facilitating complex formation on the surface of the pathogen remains to be determined. Although interaction of (C 2 -Man-)Trp in the MAC with the multivalent mannose-binding lectin from serum has not been demonstrated yet, it could play a role, since this lectin has been shown to bind to a variety of microorganisms (for an overview see Ref. 14). Furthermore, many pathogenic microorganisms are known to carry mannose-binding proteins on their surface (36).
The C-mannosylation of the terminal components of complement reveals a number of novel features of protein C-mannosylation and the properties of TSR modules in general.
A new finding is that the TSR modules in the terminal complement components contain more than one modified Trp residue. Unexpectedly, Trp residues that do not have a Trp or another aromatic residue at the ϩ3 position were found to be C-mannosylated. The most extreme examples being Trp-69 in C6 and Trp-14 in C7, which are not even part of a WXXW motif (Fig. 5). This is in contrast with human RNase 2 and IL-12 (5,12). Reexamination of their LC-ESIMS data, in the light of these new findings, confirmed that in these proteins C-mannosylation takes place exclusively on the first Trp. The case of RNase 2 is particularly compelling, because it holds for the enzyme from urine and HL-60 cells (5, 10), as well as for recombinant RNase 2 and the hybrid RNase 2.4 from HEK-293 cells (9). Furthermore, in vitro studies with the microsomeassociated transferase from liver and IL-12-secreting B-lymphoblastoid cells confirm these findings (12). The most likely explanation for the C-mannosylation of Trp residues that do not have a Trp or another aromatic residue at the ϩ3 position is that the terminal complement components contain, in addition to the WXXW motif, another signal for C-mannosylation of these Trp residues. In all cases such Trp residues are part of the sequence W(S/T)XC (Fig. 5). If this sequence, or a segment of it, would form the recognition motif, it seems likely that another transferase than the one recognizing WXXW is involved. Support for this was provided by the in vitro C-mannosylation experiments (Fig. 6). No C-mannosylation of the peptide from C6 was observed, and C-mannosylation of the one from C9 was restricted to the first Trp residue. An alternative explanation would be that the extra signal is located outside this sequence or that it is formed by a three-dimensional "signal patch" akin to substrates for UDP-GlcNAc:lysosomal enzyme N-acetylglucosamine-1-phosphotransferase. These proteins contain a conformation-dependent protein determinant for interaction with the transferase that is formed by amino acid residues that are separated in the primary structure (37).
The data obtained here (Fig. 6), together with those obtained previously (5,12), allow a partial definition of interspersing residues that are compatible with C-mannosylation of the WXXW motif. Whereas Thr, Ala, Gly, and Ser may occur at position ϩ1, Cys, Ala, Asn, Ser, Glu, and Gln are allowed at position ϩ2. Clearly, further studies are needed to complete this analysis.
There is no three-dimensional structure available for a TSR module, but the results presented here allow the definition of some of its spatial properties in the complement components. The hydrophilic nature of a mannosyl residue makes it unlikely that it is buried in the molecule. Consequently, the Trp residues that become C-mannosylated probably occur on the surface of the module. This proposal is consistent with the position of Trp-7 in the three-dimensional structure of recombinant human RNase 2 (38) and model building of the mannosyl residue (39). Furthermore, the occurrence of more than one (C 2 -Man-)Trp residue within a motif poses sterical restrictions on possible backbone conformations. Model building shows that in such a case the polypeptide chain cannot adopt a helical conformation like it does in RNase 2, which contains only a single (C 2 -Man-)Trp residue.
In addition to TSP-1 and the terminal components of complement, TSR modules have been found in a number of other proteins as follows: properdin (40), F-, M-, and SCO-spondin (41)(42)(43)(44), semaphorin F and -G (45), brain-specific angiogenin inhibitor-1, -2, and -3 (46, 47), a disintegrin and metalloproteinase with thrombospondin repeat-1 (48), aggrecanase (49), UNC-5 (50), and thrombspondin-related anonymous protein (51). Since all of them contain at least one WXXW motif, it will be of interest to examine these proteins for C-mannosylation as well. TSR modules have been classified into three groups based on length, position of cysteine residues, and net charge (52). The terminal components contain examples of each of these groups, and the present results show that all of them can be C-mannosylated (Fig. 5). The exact function of the TSR module, and therefore that of (C 2 -Man-)Trp, remains unknown. By far the most information has been obtained for TSP-1. TSP is a modular cellular adhesion molecule that has been proposed to interact with a variety of cells and macromolecules (for an overview see Refs. 53 and 54). Its TSR modules, and more importantly the WXXW motif and the neighboring sequence CSVTC, have been implicated in the adhesive process with a number of cells (55) and in protein-protein (56 -58) and proteinglycosaminoglycan interactions (59). It is of interest to note that C6, C7, C8, and C9, as well as most of the above-mentioned proteins, act at the cell surface through protein-protein and protein-membrane interaction.
The results obtained here may also bear upon proteins that are not related to TSP. One of the best conserved portions of the sequence of the TSR module, WSXWS (52), occurs in nearly all type 1 cytokine receptors (60). Their analysis for C-mannosylation would be of interest because mutagenesis studies have revealed the functional importance of this so-called "WSXWS box" (61,62).
In conclusion, we have shown that C-mannosylation can occur on multiple Trp residues in a protein. As exemplified by the membrane attack complex, this modification may alter a large proportion of the Trp residues. The results further demonstrate that this post-translational modification is more widespread than initially seemed to be the case. In particular proteins that contain TSR modules, and the class 1 cytokine receptors appear to be highly probable candidates for C-mannosylation. New mechanistic insights were obtained by the discovery of C-mannosylation of Trp residues outside or at the C-terminal end of WXXW motifs. Finally, the C-mannosylation of four complement proteins with clearly defined physiological roles provides the structural basis for studies on the function of (C 2 -Man-)Trp residues.