./. Mol. Biol. (1992) 227, 532-543
Loop Mutations Can Cause a Substantial Conformational Change in the Carboxy Terminus of the Ferritin Protein Roberto Jappelli’,
Alessandra Luzzago2 7, Paola Tataseo’, Ida Pernice’ and Gianni Cesareni’ $ 1Dipartimento di Biologia Universitd di Roma Tor Vergata Via Carnevale 00173 Roma, Italy ‘EMBL, Postfach 1022 09 6900 Heidelberg, Germany
(Received 10 January
1992; accepted 20 May 1992)
Although some protein folding theories sustain that the peptides (loops) that connect elements of more compact secondary structure may be important in the folding process, most of the data accumulated until now seemsto contradict this notion. To approach this problem we have isolated and characterized a number of mutants in which the amino acid sequence of the peptide that connects helix D and helix E in the H-chain of human ferritin has been randomized. Our results indicate that, though no single loop residue is absolutely required for ferritin to attain the native conformation, most of the mutants that we have obtained by random regional mutagenesis, affect its folding/assembly process. This conclusion was reached utilizing a sensitive test that associatesthe color formed by a colony synthesizing a hybrid ferritin-b-galactosidase protein to the ability of the ferritin domain to fold and assemble as the native protein. The characterization of the folding/assembly properties of our collection of mutants and the comparison of the mutant loop sequences, have allowed us to draw the following conclusions. Mutants that have positively charged residues at position 159, 160 or 161 fail to assembleinto the native protein shell and form an insoluble aggregate. Interestingly some loop amino acid sequences cause the E-helix to reverse direction and to expose its COOH group, normally hidden inside the protein cavity, to the solvent. The propensity of a given ferritin mutant to fold into this “non-native” conformation can be attenuated by the introduction of Gly at position 159 and 164, as in the natural ferritin. Keywords: protein folding; loops; turns; ferritin; protein fusion
1. Introduction
ture (Lewis et aE., 1971; Zimmerman & Sheraga, 1977; Dyson et al., 1988). Alternatively they are Loops are peptide segments with non-repetitive considered to be essentially passive structures derivvalues of 4 and $ angles, that have a structural role ing from formative interactions in other parts of the in connecting elements of protein secondary strucprotein (Rose et aE.. 1976; Lesk & Rose, 1981). The ture, i.e. helices and sheets. They are located on the observation that, in structurally related proteins, surface of proteins and often play important funcloop residues are generally less conserved than the tional roles contributing to the formation of cataresidues that form ordered secondary structure lytic and binding sites (Rose et al., 1985). elements is consistent with the second hypothesis Some theories sustain that loops play an active role in protein folding, forming early in this process (Leszczynski & Rose, 1986). Site-directed mutagenesis and deletion analysis of a number of and pulling together elements of secondary strucdifferent loops also supports this hypothesis (Starzyk et aE., 1987; Castagnoli et al., 1989; Fetrow t Presentaddress: I.R.B.M., Via Pontina km 30.600, et al., 1989; L. Castagnoli et al., unpublished results). 00040 Pomezia, Roma, Italy. Furthermore, loop structures form late in the $ Author to whom all correspondence should be addressed. process of folding of the protein barnase 532 0022-2836/92/180532-12
$08.00/O
81:) 1992 Academic
P’rma Limited
Loop Xzitations
in the Fw-ritin
Protkn
/
533
4-FOLD AXIS
Figure 1. The xt~ructure of the D-E loop of the H-chain of human ferritin. Thin lines indicat,e the residue side-chains and the tracing of t,he ferritin fold in the region between the end of helix D and the start of the helix E. Only some of t,he side-chains have been drawn. The thick line is the a-carbon atom backbone of the protein. Broken lines represent hydrogen bonds. The .&fold axis of the assembled apoferritin molecule is represented by a t’hick arrow.
(Matouschek et nl.. 1990). To gain further insight into the role of loops and turns in determining the stable
structure
of a protein.
we have
performed
a
detailed genetic analysis of a loop connecting t,wo hrlices in the protein ferritin. The three-dimensional struct.ure of the H-chain of human ferritin was determined at high resolution (Lawson
it al..
1991) and is shown
to have a hack-
bone rssent,ially identical to that of horse spleen ferritin (Ford et tsl., 1983). The monomer is formed by a bundle
of four helices
connected
by a six amino
acid residue loop (D-E loop) to a fifth shorter helix (helix E) that forms an acme angle with the axis of the bundle. An apoferritin molecule is composed of 24 subunits. arranged in a 432 symmetry, that form a hollow shell perforated by six hydrophobic and eight hydrophilic channels. The inspection of the atomic model indicates that the residues that form
the
1)-I
terminus extending outside, is favoured in some mutant proeeins with amino acid substitutions in the E-helix. Flip conformation is determined by the formation of the hydrophobic channels that run parallel to the 4-fold symmetry axes and is desbabilized by subst,ituting the side-chains that form the channels with hydrophilic side-cha,ins. Here.
we ask whether
the amino
acid sequence
of
t,hr I)-);: loop is important in determining a c~wrc~ct folding of the ferrit in molewlc and t hr orient at ion of the E-helix.
rnutapenc% \Vith c,ligorlIcc.leotitir (:(‘T(:(‘T(‘(‘T’r(‘T(:(:TT”(:(:(’(:(:;\;\T.~l’
(c,) P~r.rifin
2. Materials (a)
Bacterial
strnina
DSd
methods
Ferritin expression plasmids used in this work are derivatives of p2HFT (Levi et al.. 1987). Two bacterial strains were used t)hroughout this work. The)- both contain pcI857. a plasmid compatible with p2HFT, which carries the temperature sensitive allele of the i repressor cI857 and confers resistance to the antibiotic kanamyrin (Rrmaut et al.. 1983). This plasmid is used to control the d pL promoter in the ferritin H-chain expression vector p2HFT (Levi pt al.. 1987). 71172 is a supE strain and was derived from ill18 (Alac-pro/F’[larIqlacZdF5115 proAB+]supE: Messing ef al.. 1977) by transformation with pcI8.57. When necessary. plasmids carrying amber mut,ations were transformed “into GC382. a derivative of GC76 (Alar-pro/F’[ZacTq lacZAM15 proAB’]lnup”) cnontaining the plasmid ~~1857. Microbiological techniques. recombinant DNA and D?jA sequencing were according to standard protocols (Sambrook ef al.. 1989).
(b) M&ant
and
:( :(:‘I’
pur~fimtio~c
ferritin exprrsaion. 0.1 ml of an owrnight wit wr of 71,‘72 (or (XW2) containing plasmid ~JPHFT (or afly of its mutant, derivatives) was inoculated into a IO0 ml flask containing IO ml of TJ-broth (100 pg ampic*illin,rnl and 50 /fg kanarnyin:‘ml) and incubated. with shaking, at 30°C until the c*ult,urc> reached an d,oo of’ approxirnatrl~ 0.7. The flask was then rapidly shifted to -I-“‘(’ by atItling 20 ml of prewarmed Lbroth and incubated. with shaking. at t)hat t’emprraturr for 3 h. The bacteria \vt’rtA c~ollrcted Ir~y rentrifupation. resuspended into I ml of 20 rnhl-Tris (pH 7.4) and tysed by sonication. (‘ell debris (IJellrt I ) were then pelletted by centrifuging for 10 min at 1 ‘(’ ill an Epprndorf ~entrlfuge. Heat labile proteins wrrt denatured by heating in the Eppendorf tube for 10 min at 70°C’. The precipitated heat labile proteins (1Jellet r’) wert* srparat)ed by centrifugat ion in an Eppendorf c.rntrifuyt~ for 15 min. In the case of wild-type f+rrit,in. at this st,agtL approximately iO”,, of’ thr tot,al prot,tsill i,r t hc2 s,rptxr natant was ti,rritin. Further puritication (up to 93”,,) was ac~hic~vtyl 1~) adding an equal \olumr of a solution oontainilry 200,, (I-:\-) I’IX: (fJolvc~th~lt~~~r-~l~(~ol) 6000 and 2.5 >I-Sac’\. leaving for I h ‘in ic*r and prt+Jitating thcl asstmblrd apoferritin in an Eppendorf cntsntrifuge (10 I,ritl). ‘rho pellet (IJellrl 3) \~as washtyl rapidly with 20 nIlI-Tris (IJH 7.4). This washing step van be omitted if c*olonies
formrd
t)v thr
corresponding mutants have varying intensities of blue color that can be easily identified after 20 h incubation at :37”‘(’ in the presence of BCTG (50 pg/ml) and of II’TC: (26 pg/ml). A pale blue color could be identified in the mutant,s that are classified as + only after a furt’her incubation of 36 h at 4”(1. while the colonies of - rnut,ants remained white even after prolonged incubation. The numbers in the 1’1, I?2 and 1’3 columns represent the percentage of frrritin that is rrc.overed in the 3 fractions as described in Materials and Methods. The mutants laheled with an asterisk were caonstrucatrtl for other purposes by site-directed mutagenesis
by the procedure described in &Iaterials and Methods. Three fractions corresponding to the insoluble (PI), thermo-sensitive (P2) and high molecular mass thermoresistant proteins (P3) were analyzed by denaturing PAGE and the percentage of ferritin in each fraction was scored by visual inspection. These results are presented in Figure 4. According to the results of this experiment, the mutated proteins could be grouped int,o three broad classes. To the first class belong mutant ferritins whose properties are indistinguishable from those of wild-type and that can be quantitatively recovered as a soluble high molecular mass thermoresistant protein. Proteins belonging to the second class were unable to fold correctly and could only be recovered as an insoluble aggregate. The third class of mutants were distinguished from the previous two because their products were present. in varying amounts, in each of the three pellets. Inspection of the sequence of the mutated loops identifies a striking correlation between mutants encoding a protein with folding defects and the presence of positively charged side-chains at positions 159, 160 and 161. Negatively charged residues
in these same positions do not, seem to cause equivalent, effects (see mut)ant’ f’fi4). (f) Mutations that favor the Flop co@mnation As already shown for some mutations that alter t,he amino acid sequence of the E-helix. mutations in the loop can result in a second, less dratnatk, f’olding defect, bv causing the E-helix to protrude outside the ferriiin shell. All the mutants that could be purified as soluble high molecular mass proteins were tested by incubating their protein product in the presence of trypsin. Under the c:onditJions used. wild-type ferritin (regardless of the presence of the a-peptlde tail) is totally resistant to proteolysis. By contrast, the carboxy-terminal trypsin fragment of the mutants isolated by random mutagenesis became sensitive to proteolysis in various degrees, and a new band consistent with an amino-terminal ferritin fragment terminating at Lys1’73 could be identified by denaturing PAGE (see for inst,ance fliX in Fig. 2(b)). The results of digestion experiments. carried out on the mutant,s that could be purified in a soluble
in thP i7writin
Loop Mutations
Mutant
539
Trypsin sensitivity
159
Colony color
Sol.
164
LeuAla LeuAla LeuAla ArgLysMet . AspAsnThrAspLeuSer. LeuAla ArgLysMet . GlySerLeuArgGluVal. LeuAla AsgLysMet . GlyArgAspSerArgLeu. LeuAla ArgLysMet . GlyValAsnTyrPheIle. LeuAla ArgLysMet . ArgThrSerGlyValIle. LeuAla ArgLysMet . GluSerValTrpAsnPro. LeuAla ArgLysMet ArglLysMet ArgLysMet
Protein
. GlyAlaProGluSerGly. . GlyGlyProGluSerGly. . GlyGlyProGlySerGly.
0 0 0
f 27Cwt) f39” f83* f64 f41 f69 f71
20 60 60 60 80 100
f70 f68
100 100 100 100 100 20 70 10 40
+++ ++ ++ ++ +i+++
Figure 5. Trypsin sensitivity of’ “ soluble” mutants. The numbersin the Trypsin sensitivity csolumnrepresentthe percentageof frrritin moleculeswhoserarboxy-terminal peptide is digestedby trypsin order the conditionsdescribedin Materials and Methods. Numbersin the Sol. column representthe percentageof ferritin that, in thr sprc~iticmutant, is recsovrrrdin the thrrmo-resistant high molecularmasssolubleform (P3). Seealsothe legendto Fig. 4.
form. is reportled in Figure 5, where “100” indicates complete loss of the full-length ferritin peptide, while “0” indicates absence of any proteolvsis. Three intermediate classescould also be identified by comparing the intensity of the bands corresponding to the full-length and the 173 amino acid residue-long proteolytic peptide. The characterization of the loop mutants that we have isolated led us to the following tentative conclusions. (I) The blue color of the colonies is associa,ted with mutant’s whose protein product is either recoverrd in the insoluble fraction or, though assembled in a ferritin-like shell, exposes the c*arboxy terminus to the solvent (Flop conformation). The only mutants that partially violate this rule arcAf43 and f38 that form colonies that are only pale blue (Fig. 4), in spite of synthesizing a protein product t)hat, is mostly insoluble, and f64 which forms intense blue colonies, though it directs the synthesis of a caompletety soluble and predominantl!
REVERSION
“Flip” ferritin product. These apparent incongruities could be explained by ad hoc hypotheses; f43 and f38 might form a less strucatured and more compa,ct aggregate of a different natlure to that formed bv. for instance, f14, and this could explain t,he diRerent levels of fl-galactosidase activity. The phenotype of f64, on the other hand. is consistent with a slower assembly process that might favor the int,rraction between the a-peptidr and the chromosomally encloded o-peptide of /?-galact,osidase. (2) Substit,uting residues 1.59. 160 and I61 with a positively charged side-chain causesferritin to form an aggregat,r. Again f43 seemsto violate this rule as it is the &lv mutant that forms a highly insoluble prot,ein desiite the absence of posit,ively charged residues. (3) Given the relatively small number of mutant prot)rins that) can be purified in the soluble form (Fig. 5) , the sensitivity of the carboxy terminus to proteotytic degradation ran not be correlated with substitution of specific residues in
OF MUTANT
Mutant
Blue
159 164 AngLysMet . GlyAlaProGluSerGly.
LeuAla f27(w.t.)
ArgLysMet
LeuAla
F45 % ferritin Pl P2 P3
-
0
+++
50
AngLysMet . TrpAlaAlaProSerLeu. LeuAla f63
+
0
ArglLysMet . GlyArgLysProSerGly. LeuAla f62
+
60
LeuAla f61
-
0
. TrpAugLysProSerLeu.
ArgEysMet . GlyAlaAlaProSerGly. Figure
6. Reversionof
mutant
fX5.
f45
SW
the legendsto Figs 4
and
Trypsin sensitivity
0 100
0
0
ND
0 100
50
50
0
ND
0 100
20
40
5. pill.
not’ detjerminrd
f80 mm
f64
f45 Pm
f27
I5x-z
Pl
Ferritin
a- pept ide
Ferritin
f27
f29
In
f68 -;
Y 000120001200012$ T012000120001200~
Carboxy
-peptidase
Ferritin
a-peptide
%rri+ln e Ferrltln
Gly 159-am
(b) Figure 7. l’henotvpic reversion of mutant f45. (a) Gel electrophoresis of the 3 protein fracations PI, 1’2 and 1’3 text) prepared from f&j and from the 3 mutants that have been c&onstrurted in an atkmpt to revert the phenatyl)r The sequence of the mutant loops is indicated in the top line w-ith the amino avid residues rrpresenbed in the c-ode. (h) Denat,uring gel electrophoresis of mut.ant ferritins f27, f29. f6l and f63 incubated in t,he presence of amounts of trypsin (see Materials and Methods). To identif?r the bands corresponding to t,rypsin and to soybean inhibitor. the last slot was loaded with a control reaction inrubat,ed with the maximum amount of t’rypsin and t.rypsin inhibitor in the absence of ferritin. f29 is a Flop mutation alread? described (Luzzago & Cksareni. c+orresponding to LrrrlSS-Arg. Above each mutant name is the loop amino acid residue sequence in the 1 Irtjt.rhr
(SW also of f45. 1 letkr varying trypsin soybeau 1989) IW~C~.
Loop Mutations
in
the loop. However, in the next section we test the hypothesis that Gly159 and Gly164. by increasing the flexibilitv of the loop, might favor the formation of the nat’ivr conformation with the carboxy t,erminus inside the cavity. (g)
Stepwisr
reversion
of mutand
f4.5
In order to t.est these hypotheses we set up an experiment aimed at reverting the mutant phenotype of f4.5. (loop sequence: Trp159Arg160Lys161Pro162 Serl63Leu164). f45 directs the synthesis of a ferritin protein t,hat can only be recovered either in t.he insoluble or in the thermo-sensitive fraction, at a ratio of approximately 1 : 1 (see Fig. 7(a)). Ry using synthet.ic oligonucleotides we constructed three new loop muta,tions and we analyzed the phenotypes of the corresponding proteins (Fig. 6). In agreement with hypothesis 2, above. f45 could be soluhilized hy removing the two positively charged residues at position 160 and 161. In fact, the protein synt)hrsized by mutant f63 with Ala. both at position 160 and 161, could now bta completeIS recovered in the soluble fraction, proving that’ hypothesis i+ is correct. By contrast, restoring positions 159 and 164 to Gly (mutant f62) did not have any substantial elect on the relative proportions of ferrit’in recovered in the three different fractions (Fig. 7(a)). The phenotype of f63, however, is not completely wild-type since the carboxy terminus of
the Fwritin
Protein
541
the assembled mutant ferritin is, to a certain degree, sensitive to proteolysis (Fig. 7(b)). The hypothesis that we have formulated in (3), above, predicts that complete reversion of the tnutant phenotype should be accomplished by reintroducing the two Gly residues flanking the loop. This was achieved by constructing mutant f6l. As predicted by hypotheses I and 3. this new mutant forms white colonies (Fig. 6), which can not be distinguished from wild-type, and the carboxy terminus of its protein product has acquired an increased resistance to proteolysis when compared to its parent f63, which lacks thr (‘11~ residues (Fig. i(b)).
4. Discussion Protein folding pathways can be envisaged as a series of exchanging conformers t,hat. are only transiently populated during the process that leads t,o the formation of thermodynamically stable conformations. Our work has demonstrated that the end point of ferritin folding and assembly can be any of three easily distinguishable stable structures (Fig. 8). One of these is the “native” Flip conformation whose structure has been det.ermined at, high resolution (Lawson et al.; 1991). As already demonstrated in the case of other proteins (Haase-Pettingel & King, 1988), “folding errors” can lead to the formation of an insoluble,
I
Figure 8. Shemat,ir and simplified model of the ferritin assembly pathway. Details of the transition indic*ated by I have been described by Gerl & Jaenicke (1988). The diagram is used to illustrate that the proportion of the 3 stable ferritin structures is determined by the relative abundance of 2 populations of folding intermediates whose hypothetical structures arc indicated by I and II. The Flip and the Flop conformations are depicted as an equatorial slip perpendicular t,o I of the 6 4-fold axes. Ko information is available on t,hr structure of the insoluble aggregat,r S.
st,rucQrall;y ill-defined. aggregate. TXfferently from most prot,ems, however! ferritin can assemble into a shell whose monomer conformation differs from the native one in the carboxy-terminal E-helix (Flop). The proportion of subunits that’ are either in the Flip or in the Flop conformation, in a given ferritin mutant, is det,ermined by the relative abundance of two hypothetical populations of assemhly-competent folding monomers. One of these populations is compatible with t,he formation of the hydrophobic interactions between the E-helix and the 4-helix bundle in a single ferritin subunit’ and among IChelices along the 4fold axis. while the other is not Whether the tjwo populations are in thermodynamic equilibrium depends on the relat’ivr kinetics of the two processes of folding and assembly. Once the shell is formed, however. no exchange is likely to occur between the t’wo populations of conformers. The proportion of molecules that ends up in one of the three stable structures can be altered by changing the ferritin primary sequence. Previously wt’ have shown that., hy introducing charged residues in the hydrophobic interface between the E-helices along the 4-fold axis, it is possible to favor the Flop conformation (Luzzago & Cesareni, 1989). Here we have asked whether the arnino acid sequenceof the short loop connecting helix D and E cva,ninfluence t,he path taken hy ferritin during t.he folding process. The ability to construct ferritin molecules with all thr loop residues substit’ut,ed that’ are mostly Flip (seefor instance f64 in Fig. 5) indicat’es t’hat none of them plays a crucial role in determining the native Flip (sonformation. Our results, however. suggest that most loop sequences interfere with the “normal” folding/assembly process as indicated by t#hr high percentage of mutant)s forming blurt colonies. This is rather surprising when compared with similar experiments on different proteins (Starzyk et nl., 1987; (lastagnoli et al., 1989: Fetrow et rrcl., 1989: Castagnoli et al.. unpublished results). Furthermore the residues that we altered do not seemto play any crucial role in stabilizing the native structure. We believe that in the ferritin molecule even minor changes that alter the kinetic constants of the t,ransitions illustrated in Figure 8 (or similar) are amplified hy the assembly process t,ha,t freezes the rxchange between different. pools of folding/ assembly intermediates before thermodynamic equilibrium is reached. In other words it is unlikely that the drastically different’ conformations of Flip and Flop ferritins reflect substant’ially different values of the AG of the different mutants in the two conformations. We suspect that, t,he Flip conformation is the most stable, even in those mutants that eventually end up in Flop. This statement is difficult to prove since, until now, we have not succeededin obtaining two homogeneouspreparations of one mutant in the two different conformations. However. qualitative considerat,ions about t’he number of favorable
csontactslost (between the N-helices along t,he &fold axis and hrtwern E-helices and t,he 4-helix-bundle) in going from Flip to Flop, and the experimental characterization of mutant proteins that, lacking the carboxyterminal E-helix, show a“thermo-instability” compa,rable to Flop mutants. are consistent with the previous st’atement (unpublished rrsults). Alt8ernativrly. some of the loop residues pls~ :t crucial role (not, easily identifiable in the native conformation) in stabilizing a “productive:” folding int’ermediate. These considerations underline the importance of the kinetic parameters, that govern t,he transitions along the folding pathway, in determining the conformation of a protein. The characterization of our collection of mutated proteins has allowed us to begin to understand how the sequence of the six amino acid residue loop infiuences t,he partitioning between the t,hree different, conformations of ferritin. The observation that the two Cly residues flanking the loop favour Flip is not surprising and can be ascribed to a necessity for loop flexibility in order to easily form t.he bend in the native protein. Lessexpected is t.hr negative effect of positively charged residues at positions 159. 160 and 161. Preliminary experiments point to an unfavorahlr interaction of positive charges in these positions with Lys157 at the carboxy terminus of helix 1) (set Fig. 1). as suggested by the abilit,y t,o suppress the ‘-aggregation phenotype” of some mutant proteins by scconci-sit,?mutants that map in the sequencers that, ent~odt~rrsidues 156 and 1.57. IVe thank I’. Harrison for information and advice in thr initial stage of this project and P. Arosio for critit,all> reading this manuscript. This work was supportt4 1)~ 1’. F. Tngegnrria Genetica. t’. F. Kiot~ecnologit~ (1 ~)iostrurnelltazioneand by the E.C. J3RIDGE ~~ograrn. .\.I, was supported
by an tC.(‘. training
fellowship
References Boyd. I).. Vrvoli. C’.. Belches. I). &I., ,lain, 8. Ii. k Drgsdale. *J. IV. (1985). Structural and functional relationships of human frrritin H and T, c,hains deduced from cDNA clones. .I. Niol. C’hrm. 260. 117!5.5--11761.
Cast’agnoli. I,.. Scarpa. M.. Kokkinidis. 91.. Bannrr. D. W.. Tsrrnoglou, I). & (&areni. (G. (1989). (ienetit and structural analysis of the ColEl Rap (Ram) protein. EURO J. 8. til-4319. C’ollawan,.J. F.. Cowan. L. H.. (Irow. H.. Svhwaht~. (‘. Ki
Fish. u’. M’. (1987).Isolation and F)artial amino acid sequenceof three subunit spet+s of porcine spletql ferritin: evidence of’ multiple H subunits. ilrch. Riochem. Hiophys. 259. 1W---l 13.
(‘ostanzo. F.. Colomho, M.. Staemplis. 8.. Satltoro. (I.. Naront*. )I., Frank. It.. Delius. H. & C’ortrsr. K. (19%). Structure of geneand pseudogenes of human apoferritin H. Nucl. =Icids flea. 14, 721 -736. Dickey, L. B.. Sreedharan.S., Theil. E. C’., Uidsbury, ,J. R.. Wang. Y. H. & Kaufman. K. E. (1!)87). Differences in the regulation of messenger RNA for housekeeping and specialized-cell ferritin. -1. Riol. (‘hem. 262, 7901-7907.
Loop
Mutations
in thP Fe&tin
J)yson. H. .J.. Rance, M., Houghten, R,. A.. Lerner, R. & W’right P. E. (1988). Folding of immunogenic prpt~ide fragments of protein in water solution. 1. Sequence requirements for the formation of a reverse t,urn. .J. iVo1. Biol. 201; 161-200. Frt,row. J. S.. Cardillo, T. & Sherman, F. (1989). Deletion and replacement of omega loops in yeast iso-l-cytochrome c. Proteins, 6, 372-381. Ford, G. C.. Harrison, P. M., Rice. D. W., Smith, J. A. M., Treffry. A.. White. J. L. & Yariv. J. (1983). Ferritin: design and formation of an iron-storage molecule. Phil. Trans. Roy. Sot. Lond. ser. B, 304, 551-565. Gerl, M. Jz ,Jaenicke, R. (1988). Self-assembly of apoferritin from horse spleen after reversible chemical modification with 2,3-dimethylmaleic anhydride. Biochemi&y, 27, 40894096. Haase-J’ettingel, C. A. & King, J. (1988). Formation of aggregates from a thermolabile in uivo folding intermediate irr 1’22 tailspike maturation. A model for inclusion body foi,mation. J. Biol. Chem. 263, 497i-49x3. Heustrrprrutr. M, & Crichton, R. R. (1981). Amino acid sequence of horse spleen apoferritin. FEBS Letkrs. 129; 32% 329. Kunkrl. T. :I\.. Roberts, .J. D. & Zakour, R,. A. (1987). Rapid and rficient~ site-specific mutagenesis without 153. Jrhenotypics selection. hfethods Enqymol. 367.-382. J,armmli. li. K. (1970). (Xeavage of structural proteins during the assembly of the head of bacteriophage T4. Snture (London), 227. 680-685. Lawson. I>. M.. Artymiuk. P. tJ.. Yewdall. S. J.. Smith, .I. >I, A.. J,ivingstone. tJ. C., Treffrey. A., J,uzzago. .I., J,cvi. S . Arosio, I’., Cesareni, G., Thomas. (:. I).. Shaw. WT. V. & Harrison. P. M. (1991). Solving the structure of human H ferritin by genetically cnginrering intramolecular crystal contacts. Xaturr (London). 349, 541-544. Lesk. .\. M. & Rose, G. 1). (1981). Folding unit,s in ylobular prot,eins. Proc. Kut. Acad. Ski., l,‘.S.A. 78. 43044308. Leszczynski. .J. F. & Rose. G. D. (1986). Loops in globular proteins: a novel category of secondary structure. ,Sci~nce, 234. 849-855. Levi. S.. (‘esarrni. G., Arosio, P.; Lorenzetti. R., Sollazzo. M.. Alhrrtini. A. & Cortese. R. (1987). (‘haracterization of human ferritin H chain synthesized in E. foli. Gene, 51. 269-274. Lewis, J’. X.. Momany, F. A. & Sheraga, H. A. (1971). Folding of polypeptide chains in proteins: a proposed mechanism for folding. Proc. Nat. Acad. Sci., U.S.A. 68, 2293~-2297. Luzzago. A. & (‘esareni, G. (1989). Isolation of point mutations that affect the folding of t,he H-chain of ferritin in E.&i. EMRO 7. 8. 569-576. Edited
Protein
543
Matouschek. A., Kellis, J. T., Serrano. I,.. Kycroft. M. & Fersht, A. R. (1990). Transient folding intermediates characterized by protein engineering. Baturr (London). 346, 44G445. McQueen. D. S., Ray. A., Walden W. E.. Ray, 13. K.. Brown, P. H. & Thatch, R. E. (1988). Nucleotide sequence of cDNA encoding rabbit ferritin I, chain. lVucl. Acids. Res. 16, 7741. ,Messing, ,J., Gronenborn, B., Mueller-Hill, H. $ Hofschneider. P. H. (1977). Filamentous coliphage Ml3 as a cloning vehicle: insertion of a Hind111 fragment of the Zac regulatory region in Ml3 replicative form in wltro. Pror. ,Vat. Acrid. Sri , I’.#.,-!. 74, 364tL-3646. .Miya,zaki, Y-.. Setoguchi, M., Higuchi. M. Y.. Yoshida. S:., Akizuki. S. & Yamamoto. S. (1988). Xucleotide sequence of cDI$A encoding the heavy subunit of mouse macrophage ferritin. Xucl. .4crds I&s. 16. 10373. Moskait)is. J. E.. Pastori. R. L. & Schoenberg. I). R. (1990). Sequence of Xenoprrs la~/,i.c ferritin mRNA. Sucl. Acids Res. 18. 2184. Murray, M. T.. White, K. B Munro. H. X. (1987). (‘onservation of ferritin heavy subunit gene strut ture: implications for the regulation of ferritin gene rxJiression. I’roc. Xat. Acad. A”c~i.. f’.S.A. 84. 5438-7442. Remaut), E.. Tsao. H. & Fiers. W. (1!)83). Improved plasmid vectors with a thermoinducihle expression and temperature-regulated runaway replication. &n,r. 22, 103-l 13. Rose. (>. D.. \Vint,ers, R. H. & W’etlaufer. D. B. (1976). A t~establr tnodel for protein folding. FEKIS Letters. 63. IO- 16. Rose. (:. D.. (iierash. L. M. h Smith. .J. ;\. (1985). Turns iii peJ)tidr and proteins. A&WI. Protrjirr (‘hem. 37. 1 109. Yarnbrook. ,J.. Fritsch. E. F. & Maniatis. T. (1989). Lliolecular C’loning: A Labor&q 12lrrnucxl. (old Spring Harbor Laboratory Press. ( ‘old Spring Harbor. SY. Santoro. C.. Marone. M.. Ferrone. M.. (Iostanzo. F., (‘olombo. M., Minganti. C., (ortrsr. K. & Silengo. L. (1986). Cloning of the gene coding for human I, apoferritin. Nucl. Acids Rrs. 14. 2863-2X76. Starzvk, R. M.. Webster. T. A. 8r Schimmel. T. (1987). J?videncr for dispensable sequence inserted into a nucleotide fold. Science, 237. 1614 161% Stevens. I’. W:., Dogson, ,J. B. & Engel. .I. 1). (1987). Structure and expression of the c,hickcn ferritin H-subunit gene. Mol. C’~ll. Hiol. 7. 1751-1758. Zimmerman. S. S. & Sheraga, H. A. (1977). J,ocal interactions in bends of proteins. I’roc. Sort. .-1cad. Sci., l’.A’.A. 74. 4126-4129.
by A. R. Fersht
Loop Mutations Can Cause a Substantial Conformational Change in the Carboxy Terminus of the Ferritin Protein Roberto Jappelli’,
Alessandra Luzzago2 7, Paola Tataseo’, Ida Pernice’ and Gianni Cesareni’ $ 1Dipartimento di Biologia Universitd di Roma Tor Vergata Via Carnevale 00173 Roma, Italy ‘EMBL, Postfach 1022 09 6900 Heidelberg, Germany
(Received 10 January
1992; accepted 20 May 1992)
Although some protein folding theories sustain that the peptides (loops) that connect elements of more compact secondary structure may be important in the folding process, most of the data accumulated until now seemsto contradict this notion. To approach this problem we have isolated and characterized a number of mutants in which the amino acid sequence of the peptide that connects helix D and helix E in the H-chain of human ferritin has been randomized. Our results indicate that, though no single loop residue is absolutely required for ferritin to attain the native conformation, most of the mutants that we have obtained by random regional mutagenesis, affect its folding/assembly process. This conclusion was reached utilizing a sensitive test that associatesthe color formed by a colony synthesizing a hybrid ferritin-b-galactosidase protein to the ability of the ferritin domain to fold and assemble as the native protein. The characterization of the folding/assembly properties of our collection of mutants and the comparison of the mutant loop sequences, have allowed us to draw the following conclusions. Mutants that have positively charged residues at position 159, 160 or 161 fail to assembleinto the native protein shell and form an insoluble aggregate. Interestingly some loop amino acid sequences cause the E-helix to reverse direction and to expose its COOH group, normally hidden inside the protein cavity, to the solvent. The propensity of a given ferritin mutant to fold into this “non-native” conformation can be attenuated by the introduction of Gly at position 159 and 164, as in the natural ferritin. Keywords: protein folding; loops; turns; ferritin; protein fusion
1. Introduction
ture (Lewis et aE., 1971; Zimmerman & Sheraga, 1977; Dyson et al., 1988). Alternatively they are Loops are peptide segments with non-repetitive considered to be essentially passive structures derivvalues of 4 and $ angles, that have a structural role ing from formative interactions in other parts of the in connecting elements of protein secondary strucprotein (Rose et aE.. 1976; Lesk & Rose, 1981). The ture, i.e. helices and sheets. They are located on the observation that, in structurally related proteins, surface of proteins and often play important funcloop residues are generally less conserved than the tional roles contributing to the formation of cataresidues that form ordered secondary structure lytic and binding sites (Rose et al., 1985). elements is consistent with the second hypothesis Some theories sustain that loops play an active role in protein folding, forming early in this process (Leszczynski & Rose, 1986). Site-directed mutagenesis and deletion analysis of a number of and pulling together elements of secondary strucdifferent loops also supports this hypothesis (Starzyk et aE., 1987; Castagnoli et al., 1989; Fetrow t Presentaddress: I.R.B.M., Via Pontina km 30.600, et al., 1989; L. Castagnoli et al., unpublished results). 00040 Pomezia, Roma, Italy. Furthermore, loop structures form late in the $ Author to whom all correspondence should be addressed. process of folding of the protein barnase 532 0022-2836/92/180532-12
$08.00/O
81:) 1992 Academic
P’rma Limited
Loop Xzitations
in the Fw-ritin
Protkn
/
533
4-FOLD AXIS
Figure 1. The xt~ructure of the D-E loop of the H-chain of human ferritin. Thin lines indicat,e the residue side-chains and the tracing of t,he ferritin fold in the region between the end of helix D and the start of the helix E. Only some of t,he side-chains have been drawn. The thick line is the a-carbon atom backbone of the protein. Broken lines represent hydrogen bonds. The .&fold axis of the assembled apoferritin molecule is represented by a t’hick arrow.
(Matouschek et nl.. 1990). To gain further insight into the role of loops and turns in determining the stable
structure
of a protein.
we have
performed
a
detailed genetic analysis of a loop connecting t,wo hrlices in the protein ferritin. The three-dimensional struct.ure of the H-chain of human ferritin was determined at high resolution (Lawson
it al..
1991) and is shown
to have a hack-
bone rssent,ially identical to that of horse spleen ferritin (Ford et tsl., 1983). The monomer is formed by a bundle
of four helices
connected
by a six amino
acid residue loop (D-E loop) to a fifth shorter helix (helix E) that forms an acme angle with the axis of the bundle. An apoferritin molecule is composed of 24 subunits. arranged in a 432 symmetry, that form a hollow shell perforated by six hydrophobic and eight hydrophilic channels. The inspection of the atomic model indicates that the residues that form
the
1)-I
terminus extending outside, is favoured in some mutant proeeins with amino acid substitutions in the E-helix. Flip conformation is determined by the formation of the hydrophobic channels that run parallel to the 4-fold symmetry axes and is desbabilized by subst,ituting the side-chains that form the channels with hydrophilic side-cha,ins. Here.
we ask whether
the amino
acid sequence
of
t,hr I)-);: loop is important in determining a c~wrc~ct folding of the ferrit in molewlc and t hr orient at ion of the E-helix.
rnutapenc% \Vith c,ligorlIcc.leotitir (:(‘T(:(‘T(‘(‘T’r(‘T(:(:TT”(:(:(’(:(:;\;\T.~l’
(c,) P~r.rifin
2. Materials (a)
Bacterial
strnina
DSd
methods
Ferritin expression plasmids used in this work are derivatives of p2HFT (Levi et al.. 1987). Two bacterial strains were used t)hroughout this work. The)- both contain pcI857. a plasmid compatible with p2HFT, which carries the temperature sensitive allele of the i repressor cI857 and confers resistance to the antibiotic kanamyrin (Rrmaut et al.. 1983). This plasmid is used to control the d pL promoter in the ferritin H-chain expression vector p2HFT (Levi pt al.. 1987). 71172 is a supE strain and was derived from ill18 (Alac-pro/F’[larIqlacZdF5115 proAB+]supE: Messing ef al.. 1977) by transformation with pcI8.57. When necessary. plasmids carrying amber mut,ations were transformed “into GC382. a derivative of GC76 (Alar-pro/F’[ZacTq lacZAM15 proAB’]lnup”) cnontaining the plasmid ~~1857. Microbiological techniques. recombinant DNA and D?jA sequencing were according to standard protocols (Sambrook ef al.. 1989).
(b) M&ant
and
:( :(:‘I’
pur~fimtio~c
ferritin exprrsaion. 0.1 ml of an owrnight wit wr of 71,‘72 (or (XW2) containing plasmid ~JPHFT (or afly of its mutant, derivatives) was inoculated into a IO0 ml flask containing IO ml of TJ-broth (100 pg ampic*illin,rnl and 50 /fg kanarnyin:‘ml) and incubated. with shaking, at 30°C until the c*ult,urc> reached an d,oo of’ approxirnatrl~ 0.7. The flask was then rapidly shifted to -I-“‘(’ by atItling 20 ml of prewarmed Lbroth and incubated. with shaking. at t)hat t’emprraturr for 3 h. The bacteria \vt’rtA c~ollrcted Ir~y rentrifupation. resuspended into I ml of 20 rnhl-Tris (pH 7.4) and tysed by sonication. (‘ell debris (IJellrt I ) were then pelletted by centrifuging for 10 min at 1 ‘(’ ill an Epprndorf ~entrlfuge. Heat labile proteins wrrt denatured by heating in the Eppendorf tube for 10 min at 70°C’. The precipitated heat labile proteins (1Jellet r’) wert* srparat)ed by centrifugat ion in an Eppendorf c.rntrifuyt~ for 15 min. In the case of wild-type f+rrit,in. at this st,agtL approximately iO”,, of’ thr tot,al prot,tsill i,r t hc2 s,rptxr natant was ti,rritin. Further puritication (up to 93”,,) was ac~hic~vtyl 1~) adding an equal \olumr of a solution oontainilry 200,, (I-:\-) I’IX: (fJolvc~th~lt~~~r-~l~(~ol) 6000 and 2.5 >I-Sac’\. leaving for I h ‘in ic*r and prt+Jitating thcl asstmblrd apoferritin in an Eppendorf cntsntrifuge (10 I,ritl). ‘rho pellet (IJellrl 3) \~as washtyl rapidly with 20 nIlI-Tris (IJH 7.4). This washing step van be omitted if c*olonies
formrd
t)v thr
corresponding mutants have varying intensities of blue color that can be easily identified after 20 h incubation at :37”‘(’ in the presence of BCTG (50 pg/ml) and of II’TC: (26 pg/ml). A pale blue color could be identified in the mutant,s that are classified as + only after a furt’her incubation of 36 h at 4”(1. while the colonies of - rnut,ants remained white even after prolonged incubation. The numbers in the 1’1, I?2 and 1’3 columns represent the percentage of frrritin that is rrc.overed in the 3 fractions as described in Materials and Methods. The mutants laheled with an asterisk were caonstrucatrtl for other purposes by site-directed mutagenesis
by the procedure described in &Iaterials and Methods. Three fractions corresponding to the insoluble (PI), thermo-sensitive (P2) and high molecular mass thermoresistant proteins (P3) were analyzed by denaturing PAGE and the percentage of ferritin in each fraction was scored by visual inspection. These results are presented in Figure 4. According to the results of this experiment, the mutated proteins could be grouped int,o three broad classes. To the first class belong mutant ferritins whose properties are indistinguishable from those of wild-type and that can be quantitatively recovered as a soluble high molecular mass thermoresistant protein. Proteins belonging to the second class were unable to fold correctly and could only be recovered as an insoluble aggregate. The third class of mutants were distinguished from the previous two because their products were present. in varying amounts, in each of the three pellets. Inspection of the sequence of the mutated loops identifies a striking correlation between mutants encoding a protein with folding defects and the presence of positively charged side-chains at positions 159, 160 and 161. Negatively charged residues
in these same positions do not, seem to cause equivalent, effects (see mut)ant’ f’fi4). (f) Mutations that favor the Flop co@mnation As already shown for some mutations that alter t,he amino acid sequence of the E-helix. mutations in the loop can result in a second, less dratnatk, f’olding defect, bv causing the E-helix to protrude outside the ferriiin shell. All the mutants that could be purified as soluble high molecular mass proteins were tested by incubating their protein product in the presence of trypsin. Under the c:onditJions used. wild-type ferritin (regardless of the presence of the a-peptlde tail) is totally resistant to proteolysis. By contrast, the carboxy-terminal trypsin fragment of the mutants isolated by random mutagenesis became sensitive to proteolysis in various degrees, and a new band consistent with an amino-terminal ferritin fragment terminating at Lys1’73 could be identified by denaturing PAGE (see for inst,ance fliX in Fig. 2(b)). The results of digestion experiments. carried out on the mutant,s that could be purified in a soluble
in thP i7writin
Loop Mutations
Mutant
539
Trypsin sensitivity
159
Colony color
Sol.
164
LeuAla LeuAla LeuAla ArgLysMet . AspAsnThrAspLeuSer. LeuAla ArgLysMet . GlySerLeuArgGluVal. LeuAla AsgLysMet . GlyArgAspSerArgLeu. LeuAla ArgLysMet . GlyValAsnTyrPheIle. LeuAla ArgLysMet . ArgThrSerGlyValIle. LeuAla ArgLysMet . GluSerValTrpAsnPro. LeuAla ArgLysMet ArglLysMet ArgLysMet
Protein
. GlyAlaProGluSerGly. . GlyGlyProGluSerGly. . GlyGlyProGlySerGly.
0 0 0
f 27Cwt) f39” f83* f64 f41 f69 f71
20 60 60 60 80 100
f70 f68
100 100 100 100 100 20 70 10 40
+++ ++ ++ ++ +i+++
Figure 5. Trypsin sensitivity of’ “ soluble” mutants. The numbersin the Trypsin sensitivity csolumnrepresentthe percentageof frrritin moleculeswhoserarboxy-terminal peptide is digestedby trypsin order the conditionsdescribedin Materials and Methods. Numbersin the Sol. column representthe percentageof ferritin that, in thr sprc~iticmutant, is recsovrrrdin the thrrmo-resistant high molecularmasssolubleform (P3). Seealsothe legendto Fig. 4.
form. is reportled in Figure 5, where “100” indicates complete loss of the full-length ferritin peptide, while “0” indicates absence of any proteolvsis. Three intermediate classescould also be identified by comparing the intensity of the bands corresponding to the full-length and the 173 amino acid residue-long proteolytic peptide. The characterization of the loop mutants that we have isolated led us to the following tentative conclusions. (I) The blue color of the colonies is associa,ted with mutant’s whose protein product is either recoverrd in the insoluble fraction or, though assembled in a ferritin-like shell, exposes the c*arboxy terminus to the solvent (Flop conformation). The only mutants that partially violate this rule arcAf43 and f38 that form colonies that are only pale blue (Fig. 4), in spite of synthesizing a protein product t)hat, is mostly insoluble, and f64 which forms intense blue colonies, though it directs the synthesis of a caompletety soluble and predominantl!
REVERSION
“Flip” ferritin product. These apparent incongruities could be explained by ad hoc hypotheses; f43 and f38 might form a less strucatured and more compa,ct aggregate of a different natlure to that formed bv. for instance, f14, and this could explain t,he diRerent levels of fl-galactosidase activity. The phenotype of f64, on the other hand. is consistent with a slower assembly process that might favor the int,rraction between the a-peptidr and the chromosomally encloded o-peptide of /?-galact,osidase. (2) Substit,uting residues 1.59. 160 and I61 with a positively charged side-chain causesferritin to form an aggregat,r. Again f43 seemsto violate this rule as it is the &lv mutant that forms a highly insoluble prot,ein desiite the absence of posit,ively charged residues. (3) Given the relatively small number of mutant prot)rins that) can be purified in the soluble form (Fig. 5) , the sensitivity of the carboxy terminus to proteotytic degradation ran not be correlated with substitution of specific residues in
OF MUTANT
Mutant
Blue
159 164 AngLysMet . GlyAlaProGluSerGly.
LeuAla f27(w.t.)
ArgLysMet
LeuAla
F45 % ferritin Pl P2 P3
-
0
+++
50
AngLysMet . TrpAlaAlaProSerLeu. LeuAla f63
+
0
ArglLysMet . GlyArgLysProSerGly. LeuAla f62
+
60
LeuAla f61
-
0
. TrpAugLysProSerLeu.
ArgEysMet . GlyAlaAlaProSerGly. Figure
6. Reversionof
mutant
fX5.
f45
SW
the legendsto Figs 4
and
Trypsin sensitivity
0 100
0
0
ND
0 100
50
50
0
ND
0 100
20
40
5. pill.
not’ detjerminrd
f80 mm
f64
f45 Pm
f27
I5x-z
Pl
Ferritin
a- pept ide
Ferritin
f27
f29
In
f68 -;
Y 000120001200012$ T012000120001200~
Carboxy
-peptidase
Ferritin
a-peptide
%rri+ln e Ferrltln
Gly 159-am
(b) Figure 7. l’henotvpic reversion of mutant f45. (a) Gel electrophoresis of the 3 protein fracations PI, 1’2 and 1’3 text) prepared from f&j and from the 3 mutants that have been c&onstrurted in an atkmpt to revert the phenatyl)r The sequence of the mutant loops is indicated in the top line w-ith the amino avid residues rrpresenbed in the c-ode. (h) Denat,uring gel electrophoresis of mut.ant ferritins f27, f29. f6l and f63 incubated in t,he presence of amounts of trypsin (see Materials and Methods). To identif?r the bands corresponding to t,rypsin and to soybean inhibitor. the last slot was loaded with a control reaction inrubat,ed with the maximum amount of t’rypsin and t.rypsin inhibitor in the absence of ferritin. f29 is a Flop mutation alread? described (Luzzago & Cksareni. c+orresponding to LrrrlSS-Arg. Above each mutant name is the loop amino acid residue sequence in the 1 Irtjt.rhr
(SW also of f45. 1 letkr varying trypsin soybeau 1989) IW~C~.
Loop Mutations
in
the loop. However, in the next section we test the hypothesis that Gly159 and Gly164. by increasing the flexibilitv of the loop, might favor the formation of the nat’ivr conformation with the carboxy t,erminus inside the cavity. (g)
Stepwisr
reversion
of mutand
f4.5
In order to t.est these hypotheses we set up an experiment aimed at reverting the mutant phenotype of f4.5. (loop sequence: Trp159Arg160Lys161Pro162 Serl63Leu164). f45 directs the synthesis of a ferritin protein t,hat can only be recovered either in t.he insoluble or in the thermo-sensitive fraction, at a ratio of approximately 1 : 1 (see Fig. 7(a)). Ry using synthet.ic oligonucleotides we constructed three new loop muta,tions and we analyzed the phenotypes of the corresponding proteins (Fig. 6). In agreement with hypothesis 2, above. f45 could be soluhilized hy removing the two positively charged residues at position 160 and 161. In fact, the protein synt)hrsized by mutant f63 with Ala. both at position 160 and 161, could now bta completeIS recovered in the soluble fraction, proving that’ hypothesis i+ is correct. By contrast, restoring positions 159 and 164 to Gly (mutant f62) did not have any substantial elect on the relative proportions of ferrit’in recovered in the three different fractions (Fig. 7(a)). The phenotype of f63, however, is not completely wild-type since the carboxy terminus of
the Fwritin
Protein
541
the assembled mutant ferritin is, to a certain degree, sensitive to proteolysis (Fig. 7(b)). The hypothesis that we have formulated in (3), above, predicts that complete reversion of the tnutant phenotype should be accomplished by reintroducing the two Gly residues flanking the loop. This was achieved by constructing mutant f6l. As predicted by hypotheses I and 3. this new mutant forms white colonies (Fig. 6), which can not be distinguished from wild-type, and the carboxy terminus of its protein product has acquired an increased resistance to proteolysis when compared to its parent f63, which lacks thr (‘11~ residues (Fig. i(b)).
4. Discussion Protein folding pathways can be envisaged as a series of exchanging conformers t,hat. are only transiently populated during the process that leads t,o the formation of thermodynamically stable conformations. Our work has demonstrated that the end point of ferritin folding and assembly can be any of three easily distinguishable stable structures (Fig. 8). One of these is the “native” Flip conformation whose structure has been det.ermined at, high resolution (Lawson et al.; 1991). As already demonstrated in the case of other proteins (Haase-Pettingel & King, 1988), “folding errors” can lead to the formation of an insoluble,
I
Figure 8. Shemat,ir and simplified model of the ferritin assembly pathway. Details of the transition indic*ated by I have been described by Gerl & Jaenicke (1988). The diagram is used to illustrate that the proportion of the 3 stable ferritin structures is determined by the relative abundance of 2 populations of folding intermediates whose hypothetical structures arc indicated by I and II. The Flip and the Flop conformations are depicted as an equatorial slip perpendicular t,o I of the 6 4-fold axes. Ko information is available on t,hr structure of the insoluble aggregat,r S.
st,rucQrall;y ill-defined. aggregate. TXfferently from most prot,ems, however! ferritin can assemble into a shell whose monomer conformation differs from the native one in the carboxy-terminal E-helix (Flop). The proportion of subunits that’ are either in the Flip or in the Flop conformation, in a given ferritin mutant, is det,ermined by the relative abundance of two hypothetical populations of assemhly-competent folding monomers. One of these populations is compatible with t,he formation of the hydrophobic interactions between the E-helix and the 4-helix bundle in a single ferritin subunit’ and among IChelices along the 4fold axis. while the other is not Whether the tjwo populations are in thermodynamic equilibrium depends on the relat’ivr kinetics of the two processes of folding and assembly. Once the shell is formed, however. no exchange is likely to occur between the t’wo populations of conformers. The proportion of molecules that ends up in one of the three stable structures can be altered by changing the ferritin primary sequence. Previously wt’ have shown that., hy introducing charged residues in the hydrophobic interface between the E-helices along the 4-fold axis, it is possible to favor the Flop conformation (Luzzago & Cesareni, 1989). Here we have asked whether the arnino acid sequenceof the short loop connecting helix D and E cva,ninfluence t,he path taken hy ferritin during t.he folding process. The ability to construct ferritin molecules with all thr loop residues substit’ut,ed that’ are mostly Flip (seefor instance f64 in Fig. 5) indicat’es t’hat none of them plays a crucial role in determining the native Flip (sonformation. Our results, however. suggest that most loop sequences interfere with the “normal” folding/assembly process as indicated by t#hr high percentage of mutant)s forming blurt colonies. This is rather surprising when compared with similar experiments on different proteins (Starzyk et nl., 1987; (lastagnoli et al., 1989: Fetrow et rrcl., 1989: Castagnoli et al.. unpublished results). Furthermore the residues that we altered do not seemto play any crucial role in stabilizing the native structure. We believe that in the ferritin molecule even minor changes that alter the kinetic constants of the t,ransitions illustrated in Figure 8 (or similar) are amplified hy the assembly process t,ha,t freezes the rxchange between different. pools of folding/ assembly intermediates before thermodynamic equilibrium is reached. In other words it is unlikely that the drastically different’ conformations of Flip and Flop ferritins reflect substant’ially different values of the AG of the different mutants in the two conformations. We suspect that, t,he Flip conformation is the most stable, even in those mutants that eventually end up in Flop. This statement is difficult to prove since, until now, we have not succeededin obtaining two homogeneouspreparations of one mutant in the two different conformations. However. qualitative considerat,ions about t’he number of favorable
csontactslost (between the N-helices along t,he &fold axis and hrtwern E-helices and t,he 4-helix-bundle) in going from Flip to Flop, and the experimental characterization of mutant proteins that, lacking the carboxyterminal E-helix, show a“thermo-instability” compa,rable to Flop mutants. are consistent with the previous st’atement (unpublished rrsults). Alt8ernativrly. some of the loop residues pls~ :t crucial role (not, easily identifiable in the native conformation) in stabilizing a “productive:” folding int’ermediate. These considerations underline the importance of the kinetic parameters, that govern t,he transitions along the folding pathway, in determining the conformation of a protein. The characterization of our collection of mutated proteins has allowed us to begin to understand how the sequence of the six amino acid residue loop infiuences t,he partitioning between the t,hree different, conformations of ferritin. The observation that the two Cly residues flanking the loop favour Flip is not surprising and can be ascribed to a necessity for loop flexibility in order to easily form t.he bend in the native protein. Lessexpected is t.hr negative effect of positively charged residues at positions 159. 160 and 161. Preliminary experiments point to an unfavorahlr interaction of positive charges in these positions with Lys157 at the carboxy terminus of helix 1) (set Fig. 1). as suggested by the abilit,y t,o suppress the ‘-aggregation phenotype” of some mutant proteins by scconci-sit,?mutants that map in the sequencers that, ent~odt~rrsidues 156 and 1.57. IVe thank I’. Harrison for information and advice in thr initial stage of this project and P. Arosio for critit,all> reading this manuscript. This work was supportt4 1)~ 1’. F. Tngegnrria Genetica. t’. F. Kiot~ecnologit~ (1 ~)iostrurnelltazioneand by the E.C. J3RIDGE ~~ograrn. .\.I, was supported
by an tC.(‘. training
fellowship
References Boyd. I).. Vrvoli. C’.. Belches. I). &I., ,lain, 8. Ii. k Drgsdale. *J. IV. (1985). Structural and functional relationships of human frrritin H and T, c,hains deduced from cDNA clones. .I. Niol. C’hrm. 260. 117!5.5--11761.
Cast’agnoli. I,.. Scarpa. M.. Kokkinidis. 91.. Bannrr. D. W.. Tsrrnoglou, I). & (&areni. (G. (1989). (ienetit and structural analysis of the ColEl Rap (Ram) protein. EURO J. 8. til-4319. C’ollawan,.J. F.. Cowan. L. H.. (Irow. H.. Svhwaht~. (‘. Ki
Fish. u’. M’. (1987).Isolation and F)artial amino acid sequenceof three subunit spet+s of porcine spletql ferritin: evidence of’ multiple H subunits. ilrch. Riochem. Hiophys. 259. 1W---l 13.
(‘ostanzo. F.. Colomho, M.. Staemplis. 8.. Satltoro. (I.. Naront*. )I., Frank. It.. Delius. H. & C’ortrsr. K. (19%). Structure of geneand pseudogenes of human apoferritin H. Nucl. =Icids flea. 14, 721 -736. Dickey, L. B.. Sreedharan.S., Theil. E. C’., Uidsbury, ,J. R.. Wang. Y. H. & Kaufman. K. E. (1!)87). Differences in the regulation of messenger RNA for housekeeping and specialized-cell ferritin. -1. Riol. (‘hem. 262, 7901-7907.
Loop
Mutations
in thP Fe&tin
J)yson. H. .J.. Rance, M., Houghten, R,. A.. Lerner, R. & W’right P. E. (1988). Folding of immunogenic prpt~ide fragments of protein in water solution. 1. Sequence requirements for the formation of a reverse t,urn. .J. iVo1. Biol. 201; 161-200. Frt,row. J. S.. Cardillo, T. & Sherman, F. (1989). Deletion and replacement of omega loops in yeast iso-l-cytochrome c. Proteins, 6, 372-381. Ford, G. C.. Harrison, P. M., Rice. D. W., Smith, J. A. M., Treffry. A.. White. J. L. & Yariv. J. (1983). Ferritin: design and formation of an iron-storage molecule. Phil. Trans. Roy. Sot. Lond. ser. B, 304, 551-565. Gerl, M. Jz ,Jaenicke, R. (1988). Self-assembly of apoferritin from horse spleen after reversible chemical modification with 2,3-dimethylmaleic anhydride. Biochemi&y, 27, 40894096. Haase-J’ettingel, C. A. & King, J. (1988). Formation of aggregates from a thermolabile in uivo folding intermediate irr 1’22 tailspike maturation. A model for inclusion body foi,mation. J. Biol. Chem. 263, 497i-49x3. Heustrrprrutr. M, & Crichton, R. R. (1981). Amino acid sequence of horse spleen apoferritin. FEBS Letkrs. 129; 32% 329. Kunkrl. T. :I\.. Roberts, .J. D. & Zakour, R,. A. (1987). Rapid and rficient~ site-specific mutagenesis without 153. Jrhenotypics selection. hfethods Enqymol. 367.-382. J,armmli. li. K. (1970). (Xeavage of structural proteins during the assembly of the head of bacteriophage T4. Snture (London), 227. 680-685. Lawson. I>. M.. Artymiuk. P. tJ.. Yewdall. S. J.. Smith, .I. >I, A.. J,ivingstone. tJ. C., Treffrey. A., J,uzzago. .I., J,cvi. S . Arosio, I’., Cesareni, G., Thomas. (:. I).. Shaw. WT. V. & Harrison. P. M. (1991). Solving the structure of human H ferritin by genetically cnginrering intramolecular crystal contacts. Xaturr (London). 349, 541-544. Lesk. .\. M. & Rose, G. 1). (1981). Folding unit,s in ylobular prot,eins. Proc. Kut. Acad. Ski., l,‘.S.A. 78. 43044308. Leszczynski. .J. F. & Rose. G. D. (1986). Loops in globular proteins: a novel category of secondary structure. ,Sci~nce, 234. 849-855. Levi. S.. (‘esarrni. G., Arosio, P.; Lorenzetti. R., Sollazzo. M.. Alhrrtini. A. & Cortese. R. (1987). (‘haracterization of human ferritin H chain synthesized in E. foli. Gene, 51. 269-274. Lewis, J’. X.. Momany, F. A. & Sheraga, H. A. (1971). Folding of polypeptide chains in proteins: a proposed mechanism for folding. Proc. Nat. Acad. Sci., U.S.A. 68, 2293~-2297. Luzzago. A. & (‘esareni, G. (1989). Isolation of point mutations that affect the folding of t,he H-chain of ferritin in E.&i. EMRO 7. 8. 569-576. Edited
Protein
543
Matouschek. A., Kellis, J. T., Serrano. I,.. Kycroft. M. & Fersht, A. R. (1990). Transient folding intermediates characterized by protein engineering. Baturr (London). 346, 44G445. McQueen. D. S., Ray. A., Walden W. E.. Ray, 13. K.. Brown, P. H. & Thatch, R. E. (1988). Nucleotide sequence of cDNA encoding rabbit ferritin I, chain. lVucl. Acids. Res. 16, 7741. ,Messing, ,J., Gronenborn, B., Mueller-Hill, H. $ Hofschneider. P. H. (1977). Filamentous coliphage Ml3 as a cloning vehicle: insertion of a Hind111 fragment of the Zac regulatory region in Ml3 replicative form in wltro. Pror. ,Vat. Acrid. Sri , I’.#.,-!. 74, 364tL-3646. .Miya,zaki, Y-.. Setoguchi, M., Higuchi. M. Y.. Yoshida. S:., Akizuki. S. & Yamamoto. S. (1988). Xucleotide sequence of cDI$A encoding the heavy subunit of mouse macrophage ferritin. Xucl. .4crds I&s. 16. 10373. Moskait)is. J. E.. Pastori. R. L. & Schoenberg. I). R. (1990). Sequence of Xenoprrs la~/,i.c ferritin mRNA. Sucl. Acids Res. 18. 2184. Murray, M. T.. White, K. B Munro. H. X. (1987). (‘onservation of ferritin heavy subunit gene strut ture: implications for the regulation of ferritin gene rxJiression. I’roc. Xat. Acad. A”c~i.. f’.S.A. 84. 5438-7442. Remaut), E.. Tsao. H. & Fiers. W. (1!)83). Improved plasmid vectors with a thermoinducihle expression and temperature-regulated runaway replication. &n,r. 22, 103-l 13. Rose. (>. D.. \Vint,ers, R. H. & W’etlaufer. D. B. (1976). A t~establr tnodel for protein folding. FEKIS Letters. 63. IO- 16. Rose. (:. D.. (iierash. L. M. h Smith. .J. ;\. (1985). Turns iii peJ)tidr and proteins. A&WI. Protrjirr (‘hem. 37. 1 109. Yarnbrook. ,J.. Fritsch. E. F. & Maniatis. T. (1989). Lliolecular C’loning: A Labor&q 12lrrnucxl. (old Spring Harbor Laboratory Press. ( ‘old Spring Harbor. SY. Santoro. C.. Marone. M.. Ferrone. M.. (Iostanzo. F., (‘olombo. M., Minganti. C., (ortrsr. K. & Silengo. L. (1986). Cloning of the gene coding for human I, apoferritin. Nucl. Acids Rrs. 14. 2863-2X76. Starzvk, R. M.. Webster. T. A. 8r Schimmel. T. (1987). J?videncr for dispensable sequence inserted into a nucleotide fold. Science, 237. 1614 161% Stevens. I’. W:., Dogson, ,J. B. & Engel. .I. 1). (1987). Structure and expression of the c,hickcn ferritin H-subunit gene. Mol. C’~ll. Hiol. 7. 1751-1758. Zimmerman. S. S. & Sheraga, H. A. (1977). J,ocal interactions in bends of proteins. I’roc. Sort. .-1cad. Sci., l’.A’.A. 74. 4126-4129.
by A. R. Fersht
