Biochemistry 2000, 39, 575-583
`
`575
`
`Aggregation Events Occur Prior to Stable Intermediate Formation during Refolding
`of Interleukin 1(cid:226)†
`
`John M. Finke, Melinda Roy, Bruno H. Zimm, and Patricia A. Jennings*
`Department of Chemistry and Biochemistry, UniVersity of California, San Diego, La Jolla, California 92093-0359
`ReceiVed July 1, 1999; ReVised Manuscript ReceiVed October 26, 1999
`
`ABSTRACT: A point mutation, lysine 97 f isoleucine (K97I), in a surface loop in the (cid:226)-sheet protein
`interleukin 1(cid:226) (IL-1(cid:226)), exhibits increased levels of inclusion body (IB) formation relative to the wild-
`type protein (WT) when expressed in Escherichia coli. Despite the common observation that less stable
`proteins are often found in IBs, K97I is more stable than WT. We examined the folding pathway of the
`mutant and wild-type proteins at pH 6.5 and 25 (cid:176) C with manual-mixing and stopped-flow optical
`spectroscopy to determine whether changes in the properties of transiently populated species in vitro
`correlate with the observation of increased aggregation in vivo. The refolding reactions of the WT and
`K97I proteins are both described by three exponential processes. Two exponential processes characterize
`fast events (0.1-1.0 s) in folding while the third exponential process correlates with a slow (70 s) single
`pathway to and from the native state. The K97I replacement affects the earlier steps in the refolding
`pathway. Aggregation, absent in the WT refolding reaction, occurs in K97I above a critical protein
`concentration of 18 (cid:237)M. This observation is consistent with an initial nucleation step mediating protein
`aggregation. Stopped-flow kinetic studies of the K97I aggregation process demonstrate that K97I aggregates
`most rapidly during the earliest refolding times, when unfolded protein conformers remain highly populated
`and the concentration of folding intermediates is low. Folding and aggregation studies together support
`a model in which the formation of stable folding intermediates afford protection against further K97I
`aggregation.
`
`Understanding the factors responsible for the aggregation
`of expressed proteins is a problem of importance not only
`in biotechnology but also in the health-related industries.
`Deposition of insoluble protein aggregates results in the
`formation of inclusion bodies during the bacterial expression
`of recombinant proteins (3-8) as well as a variety of
`pathological conditions in mammals, including Alzheimer’s
`and other amyloid-related disease states (3, 9-12). A
`common characteristic of protein deposition diseases, as well
`as inclusion body formation during bacterial expression, is
`the sensitivity of the extent of aggregation to point mutations
`in the protein of interest (3, 4, 11, 13). The common belief
`is that single amino acid replacements creating off-pathway
`association events are mediated either by intermediates or
`by native states, as a single mutation is thought to be unlikely
`to be able to alter significantly the unfolded biopolymer (14).
`However, it is known that mutations can and do affect the
`stability and transient structure of the unfolded state (15).
`Thus, the effects of point mutations on aggregation formed
`during protein folding may help identify the species and
`factors involved in protein deposition.
`To address fully protein aggregation phenomena, detailed
`structural and biochemical information on the protein being
`
`† This work supported by Hellman Faculty Fellowship (P.A.J.), Sloan
`Fellowship (P.A.J.), NIH Grant GM54038 (P.A.J.), and Public Health
`Service Grant CA09523 (J.M.F. and M.R.).
`* To whom correspondence should be addressed: University of
`California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0359.
`Phone (858) 534-6417; Fax (858) 534-6714; Email pajennin@ucsd.edu.
`
`studied must be available in addition to a consistent method
`of modulating aggregation. Consequently, recombinant hu-
`man interleukin 1(cid:226) (IL-1(cid:226))1 is an excellent model system
`for characterizing the effects of point mutations on the
`competition between on-pathway folding events and off-
`pathway aggregation. The wild-type protein (WT) is found
`predominantly (>90%) in the soluble fraction of whole-cell
`lysates when expressed at 37 (cid:176) C in Escherichia coli, and a
`number of point mutations have been identified that show
`dramatic effects on the levels of inclusion bodies formed
`(>90% insoluble) (4, 16, 17). Of particular interest is the
`mutation in the surface loop 86-99 that replaces the
`positively charged lysine side chain at position 97 with the
`hydrophobic side chain of isoleucine (18; see Figure 1 and
`Materials and Methods). This mutation K97I results in a
`native state that is thermodynamically more stable than the
`wild-type protein yet is found predominantly in inclusion
`bodies when expressed at 37 (cid:176) C in E. coli. No alteration in
`the kinetics of the slow step in folding, nor the ribosomal
`synthesis rates, nor protein-chaperone interactions have been
`found with this mutation (17). FTIR analysis indicates that
`
`1 Abbreviations: FTIR, Fourier transform infrared spectroscopy;
`IPTG, isopropyl (cid:226)-D-thiogalactopyranoside; PMSF, phenylmethane-
`sulfonyl fluoride; EDTA, ethylenediaminetetraacetic acid; DTT, dithio-
`threitol; (cid:226)me, 2-mercaptoethanol; Tris-HCl,
`tris(hydroxymethyl)-
`aminomethane hydrochloride; MES, 2-(4-morpholino)ethanesulfonic
`acid; GdnHCl, guanidine hydrochloride; Gdn-SCN, guanidine thiocy-
`anate; FPLC, fast protein liquid chromatography; SDS-PAGE, sodium
`dodecyl sulfate-polyacrylamide gel electrophoresis; ESI-MS, electro-
`spray ionization mass spectrometry; IL-1(cid:226), interleukin-1(cid:226).
`
`10.1021/bi991518m CCC: $19.00 © 2000 American Chemical Society
`Published on Web 12/28/1999
`
`Amgen Exhibit 2048
`Apotex Inc. et al. v. Amgen Inc. et al., IPR2016-01542
`Page 1
`
`

`

`576 Biochemistry, Vol. 39, No. 3, 2000
`
`Finke et al.
`
`a flow rate of 3 mL/min. Purity was judged to be greater
`than 95% as assessed by SDS-PAGE. Protein concentrations
`were calculated by using an experimentally determined (cid:15)280
`) 11.26 mM-1 cm-1 (21). Mass spectral analysis and DNA
`sequencing identified the mutant IL-1(cid:226) as a K97I variant,
`and not K97V as it had been previously reported (4). Purified
`protein was dialyzed extensively into 10 mM MES buffer
`containing 2 mM EDTA, 90 mM NaCl, and 1 mM (cid:226)me at
`pH 6.5 for all experiments discussed.
`Equilibrium Fluorescence. Equilibrium unfolding titrations
`were measured by use of intrinsic tryptophan fluorescence
`emission intensity (22, 23). Protein samples were diluted to
`varying final denaturant concentrations and equilibrated
`overnight. Protein concentrations ranged from 6 to 18 (cid:237)M
`in separate experiments. Fluorescence spectra were acquired
`on a Fluoromax-2 spectrofluorometer (SPEX, Edison, NJ).
`Fluorescence emission was measured as total intensity over
`the 300-450 nm emission range after excitation at 293 nm.
`Manual-Mixing Fluorescence. Unfolding experiments were
`initiated by addition of native IL-1(cid:226) stock into a cuvette to
`final GdnHCl concentrations ranging from 2 to 5 M and final
`IL-1(cid:226) concentrations ranging from 6 to 24 (cid:237)M. Refolding
`experiments were initiated by dilution of unfolded IL-1(cid:226) at
`2.2 M GdnHCl to final GdnHCl concentrations ranging from
`0.2 to 1.4 M and final IL-1(cid:226) concentrations ranging from 6
`to 36 (cid:237)M. The kinetics of folding reactions with relaxation
`times greater than 10 s were measured by the time-dependent
`change in fluorescence emission at 343 nm (slit 1 mm) while
`excitation was at 293 nm (slit 1 mm) with a Fluoromax-2
`spectrofluorometer equipped with a Neslab RTE-111 tem-
`perature controller.
`Stopped-Flow Fluorescence. Folding reactions with re-
`laxation times less than 10 s were monitored with an Applied
`Photophysics SX.17MV (Applied Photophysics, London)
`stopped-flow unit with a path length of 0.1 cm. Refolding
`experiments were initiated by a 1:10 dilution of unfolded
`IL-1(cid:226) at 2.2 M GdnHCl, except where indicated, into varying
`final IL-1(cid:226) concentrations and final GdnHCl concentrations.
`Unfolding experiments used 1:1 ratio mixing sizes such that
`1 part native IL-1(cid:226) dilutes into 1 part GdnHCl buffer.
`Excitation was at 293 nm and emission was collected through
`a >320 nm cutoff filter. Each kinetic trace is the average of
`10-20 data acquisitions.
`Interrupted Refolding Experiments. The reverse double-
`jump experiment (24)
`
`step 1
`unfolded 98
`(ti)
`
`step 2
`folded 98
`
`unfolded
`
`was performed to measure the time-dependent production
`of native wild-type and K97I protein. Folding (step 1) was
`initiated by diluting 4 parts unfolded protein in 3.0 M
`GdnHCl with 9 parts buffer to reach a final buffer concentra-
`tion of 0.8 M GdnHCl. After various times (ti), refolding
`was interrupted by transferring the solution into a final
`GdnHCl concentration of 4.5 M (step 2). Partially folded
`intermediates present
`immediately prior to step 2 are
`extremely unstable (0.8 M GdnHCl refolding conditions),
`unfold quickly during step 2, and are therefore not measured.
`However, under these conditions, the population of native
`IL-1(cid:226) unfolds at the expected rate ((cid:244)step2 (cid:24) 400 s). Conse-
`quently, native IL-1(cid:226) at each time point ti can be reliably
`
`FIGURE 1: MOLMOL representation (44) of the solution structure
`of human interleukin 1(cid:226) (45) down the (cid:226)-barrel with secondary
`structure elements as defined by Kabsch and Sander (46). Residues
`demonstrating early amide-proton protection are shown as dark
`strands. Residues demonstrating late amide-proton protection are
`shown as light strands. The residue 86-99 loop microdomain is
`indicated and the extended side chain of lysine 97 is shown.
`
`K97I inclusion bodies in vivo and aggregates in vitro contain
`a significant and similar degree of secondary structure, yet
`both are different from the native state (19).
`We now report the results of manual-mixing and stopped-
`flow optical experiments on the folding reaction of WT and
`K97I at pH 6.5 and 25 (cid:176) C. Using kinetic analyses of both
`folding and aggregation, we have determined that K97I self-
`associates when unfolded conformers predominate and that
`formation of stable folding intermediates protects against
`further association. We have proposed structural details of
`the initial aggregation stage and how aggregation is subse-
`quently prevented during the folding process, a prerequisite
`to any further studies of the final insoluble aggregates.
`
`MATERIALS AND METHODS
`Expression and Purification of IL-1(cid:226) and K97I. The gene
`encoding either wild-type IL-1(cid:226) (WT) or the K97I sequence
`variant were subcloned into a pET 24-d vector (Novagen)
`and transformed into E. coli BL21(DE3) cells. Cells were
`grown in LB to an OD600 of 0.8 and protein expression was
`induced with IPTG at a final concentration of 1 mM.
`Rifampicin was added 45 min after induction to a final
`concentration of 0.1 mg/mL. Three hours after induction,
`cells were harvested by centrifugation at 10000g for 30 min.
`Purification of IL-1(cid:226) was based on the procedure of
`Meyers et al. (20) but with a number of modifications. Cells
`were resuspended in 10 mM KPO4, 0.2 mM EDTA, 5 mM
`DTT, and 1 mM PMSF at pH 8.0 and then lysed by
`sonication at 4 (cid:176) C, followed by centrifugation at 4300g for
`30 min. Soluble IL-1(cid:226) in the supernatant was made 80%
`saturated in ammonium sulfate and then precipitated by
`centrifugation as above. The pellet was then dissolved in
`buffer A (25 mM NH4OAc, 2 mM EDTA, and 1 mM (cid:226)me
`at pH 4.5) and dialyzed overnight in buffer A at 4 (cid:176) C.
`The dialysate was centrifuged as above, filtered, and
`applied to a Resource S cation-exchange column (Pharmacia)
`equilibrated in buffer A. IL-1(cid:226) was eluted in a 40 column
`volume linear gradient of 25-240 mM NH4OAc, pH 4.5, at
`
`Page 2
`
`

`

`Aggregation of Interleukin 1(cid:226)
`
`Biochemistry, Vol. 39, No. 3, 2000 577
`
`quantitated by the amplitude of the resulting unfolding kinetic
`exponential (cid:244)step2. All amplitudes were normalized to the
`amplitude recovered from step 2 on 100% native protein.
`No amplitude was observed when unfolded IL-1(cid:226) (3.0 M
`GdnHCl) was diluted into the final conditions (4.5 M
`GdnHCl), confirming that measured amplitudes could result
`only from native IL-1(cid:226) formed from folding after step 1 and
`not from extraneous processes. Spectroscopic conditions were
`as described under Manual-Mixing Fluorescence.
`Stopped-Flow Light Scattering. Protein aggregation during
`refolding was monitored with an Applied Photophysics
`SX.17MV (Applied Photophysics, London) stopped-flow unit
`with a path length of 0.1 cm. Aggregation was initiated by
`a 1:10 dilution of IL-1(cid:226) from 2.2 M GdnHCl (except where
`noted) to the indicated final IL-1(cid:226) concentrations of protein
`and final GdnHCl concentrations. Aggregation was measured
`by light scattering at a wavelength of 500 nm and collected
`at a 90(cid:176) angle through a >320 nm cutoff filter. Each kinetic
`trace is the average of 10-20 data acquisitions.
`Fraction Aggregated. Refolding experiments were initiated
`by a 1:10 dilution of IL-1(cid:226) from 2.2 M GdnHCl, except
`where noted, to the indicated final IL-1(cid:226) concentrations of
`protein and 0.4 M GdnHCl. Samples were centrifuged at
`7000g and the completeness of the aggregate separation was
`assessed by the absence of turbidity at 500 nm. The dissolved
`protein concentration was measured by the absorbance at
`280 nm and calculated by use of the extinction coefficient
`(cid:15)280 ) 11.26 mM-1 cm-1 and the Beer-Lambert relationship
`(21).
`Data Analysis. Equilibrium data were fit as described
`previously (4). Manual mixing kinetics, stopped-flow kinet-
`ics, and interrupted refolding data were fit globally (25) to
`eq 1 using the Marquardt algorithm (26) and in-house
`software:
`
`i
`
`A(t) ) (cid:229)
`
`Ai exp(-t/(cid:244)i) + A(¥
`
`)
`
`(1)
`
`The number of kinetic processes i, relaxation time (cid:244)i, signal
`amplitude Ai of each exponential kinetic process i, and the
`final signal value A(¥
`) at equilibrium were determined using
`the fit quality represented in the chi-squared (ł
`2) values (27),
`r
`the random dispersion of residuals, and the logical consis-
`tency of the generated fitting parameters. The relaxation time
`(cid:244) equals the inverse of the observed rate constant (i.e., 1/kobs)
`and does not directly measure the microscopic rate constant
`for a kinetic process (28).
`
`RESULTS
`
`Equilibrium Fluorescence. The equilibrium transition for
`K97I IL-1(cid:226) is shifted to higher denaturant concentrations,
`consistent with a 0.9 kcal/mol increase in thermodynamic
`stability of the native protein with mutation, as observed
`previously (16, 18, 22).
`Fluorescence Measured Kinetics. A plot of the change in
`fluorescence intensity above 320 nm as a function of time
`for a stopped-flow refolding jump of WT IL-1(cid:226) from 2.2 to
`0.2 M GdnHCl is given in Figure 2A. The first 20 s of the
`reaction are shown for clarity. The initial
`increase in
`fluorescence intensity is characterized by two exponential
`processes and is followed by a slow process in which the
`decay in fluorescence intensity approaches the expected
`
`FIGURE 2: Real-time kinetic fluorescence measurements during the
`refolding of WT IL-1(cid:226). (A) Refolding from 2.2 to 0.2 M GdnHCl
`by stopped-flow mixing. Individual data points ((cid:226)) are shown along
`with the 3-exponential fit (s). (Inset) Residuals of the kinetics fit
`to three exponential processes (cid:244)1, (cid:244)2, and (cid:244)3. (B) Refolding from
`2.2 to 0.2 M GdnHCl by manual mixing. Under these conditions,
`the displayed (cid:244)1 process is characterized by a decrease in fluores-
`cence with time. (C) Refolding from 2.2 to 1.0 M GdnHCl by
`manual mixing. Under these conditions, the displayed (cid:244)1 process
`is characterized by an increase in fluorescence with time. Given
`the change in sign for the amplitude of (cid:244)1 between panels B and C,
`interpretation of amplitude changes over GdnHCl concentrations
`for the three kinetic processes (cid:244)1, (cid:244)2, and (cid:244)3 will be ambiguous.
`
`equilibrium value. This nonmonotonic change in fluorescence
`intensity as a function of time measures the population of a
`highly fluorescent intermediate formed prior to the native
`protein and is consistent with the finding of a populated
`intermediate with pulse-labeling techniques (1, 2). The
`calculated fit of the data to a three-exponential process is
`shown for comparison (Figure 2A). The plot of the residual
`errors for the fit of the data to a three-exponential equation
`is given in the inset to Figure 2A. As demonstrated in Figure
`2 panels B (0.2 M GdnHCl) and C (1.0 M GdnHCl), the
`slowest process has either a negative (Figure 2B) or positive
`(Figure 2C) fluorescence amplitude, depending upon the final
`denaturant concentration in the refolding jumps. This ob-
`servation is consistent with the changes in equilibrium
`fluorescence measurements of the native state observed at
`0.2 and 1.0 M GdnHCl, respectively. Global analysis of both
`WT and K97I refolding data consistently fit best to three
`exponential processes and are designated as (cid:244)3 (fast), (cid:244)2
`(medium), and (cid:244)1 (slow). Analysis of both WT and K97I
`unfolding data under all conditions consistently fit best to
`one exponential process and is designated (cid:244)1(slow). The two
`faster processes observed in the refolding reactions, (cid:244)3 and
`(cid:244)2, were not observed in unfolding by either manual-mixing
`or stopped-flow techniques (data not shown).
`
`Page 3
`
`

`

`578 Biochemistry, Vol. 39, No. 3, 2000
`
`Finke et al.
`
`FIGURE 3: Manual mixing kinetic fluorescence measurement of
`folding and unfolding kinetic parameters for WT A1/(cid:244)1 refolding
`(O), WT A1/(cid:244)1 unfolding (b), K97I A1/(cid:244)1 refolding (]), and K97I
`A1/(cid:244)1 unfolding ([). (A) Amplitude A1 of (cid:244)1 folding/unfolding
`process with respect to final GdnHCl concentration. (B) Relaxation
`time of (cid:244)1 folding/unfolding process with respect to final GdnHCl
`concentration. The values of (cid:244)1 are essentially identical for WT
`and K97I, as observed previously (17).
`
`Fluorescence-detected folding kinetics for both WT and
`K97I were performed over a series of denaturant concentra-
`tions and 18 (cid:237)M protein. The observed amplitude A1 and
`slow relaxation time (cid:244)1, determined in manual-mixing
`experiments as a function of final denaturant concentration,
`are given in Figure 3. The refolding relaxation time (cid:244)1 varies
`from 30 s under strongly refolding conditions (0.2 M
`GdnHCl) to 1900 s in the transition region (1.0 M GdnHCl),
`where solution conditions support the formation of a highly
`fluorescent nativelike species (18). Unfolding relaxation
`times vary from 160 s under strongly unfolding conditions
`(5 M GdnHCl) to 3100 s in the transition region (2 M
`GdnHCl). Neither the observed relaxation times nor the
`fluorescence amplitudes for the slow (cid:244)1 process differ
`appreciably between the WT and K97I mutant proteins,
`consistent with previously published observations (16).
`In stopped-flow fluorescence experiments, an initial
`increase in fluorescence intensity observed in refolding
`experiments at 18 (cid:237)M protein is described by two exponential
`process, (cid:244)3 and (cid:244)2. Interpretation of the GdnHCl dependence
`of the amplitudes of (cid:244)3 and (cid:244)2, A3 and A2, is unreliable
`because the amplitude A1 changes from negative to positive
`at increasing GdnHCl concentrations (Figure 2A) and affects
`the fitted values of A3, A2, and A1. Therefore, a qualitative
`analysis of fluorescence amplitudes follows. The observed
`amplitudes A3 and A2 for both the WT and K97I protein are
`shown as a function of final denaturant concentration in
`Figure 4A. The major difference in amplitudes of the fast
`processes between WT and K97I is in A2. K97I A2 is much
`
`FIGURE 4: Stopped-flow kinetic fluorescence measurements for
`refolding kinetic parameters WT A3/(cid:244)3 (0), WT A2/(cid:244)2 ("), K97I
`A3/(cid:244)3 (4), and K97I A2/(cid:244)2 (3). Error bars represent 95% confidence
`intervals. WT and K97I A3/(cid:244)3 are not shown above 1.0 M GdnHCl
`because the amplitude A3 can no longer be resolved. (A) Amplitude
`(in PMT volts) of kinetic folding processes WT A3, WT A2, K97I
`A3, and K97I A2. (B) Relaxation time of WT (cid:244)3, WT (cid:244)2, K97I (cid:244)3,
`and K97I (cid:244)2 folding processes with respect
`to final GdnHCl
`concentration. The values of K97I (cid:244)2 do not follow simple
`denaturant concentration dependence.
`
`lower than WT A2 at 0.2 M GdnHCl but increases to the
`value of the WT A2 at 0.8 M GdnHCl and behaves similarly
`to WT at higher GdnHCl concentrations. Although a
`quantitative interpretation of the amplitudes is precluded (see
`above), it is nonetheless clear that the major amplitude
`differences between K97I and WT occur in A2.
`In Figure 4B, the WT relaxation times (cid:244)3 and (cid:244)2 at 18 (cid:237)M
`protein display the characteristic denaturant concentration
`dependence expected for a protein folding reaction (29). K97I
`(cid:244)3 is slightly accelerated with respect to WT (cid:244)3 and increases
`logarithmically from 50 ms at 0.2 M GdnHCl to 780 ms at
`1.0 M GdnHCl. In contrast, K97I (cid:244)2 displays kinetic behavior
`that does not follow simple denaturant concentration depen-
`dence. K97I (cid:244)2 displays a relaxation time of approximately
`1 s, which is independent of GdnHCl concentrations below
`0.6 M, as opposed to WT (cid:244)2, which decreases to 260 ms at
`0.2 M GdnHCl. Repeated experiments (4(cid:2)) with different
`protein preparations demonstrate that differences between
`WT and K97I are reproducible and statistically significant,
`as shown by 95% confidence intervals in Figure 4. Above
`0.6 M GdnHCl, K97I (cid:244)2 is denaturant-dependent and similar
`in rate to WT (cid:244)2. For both WT and K97I above 1.0 M
`GdnHCl, the (cid:244)3 process is no longer observable and (cid:244)2
`reaches a maximum near 4 s.
`Interrupted Refolding Experiments. At pH 5.0, the fast
`processes in the refolding reaction of the WT protein are
`not a result of parallel paths to the native protein (2). To
`
`Page 4
`
`

`

`Aggregation of Interleukin 1(cid:226)
`
`Biochemistry, Vol. 39, No. 3, 2000 579
`
`test whether native WT and K97I also accumulate from a
`single folding route at pH 6.5, we performed interrupted
`refolding experiments (24). The time-dependent recoveries
`of the native-state amplitude during step 2 of the interrupted
`folding experiment (see Materials and Methods) for both WT
`and K97I are each accurately fit to a single-exponential
`process. The relaxation time for this process is 420 s for the
`WT and 370 s for K97I (data not shown), consistent with
`the slowest refolding process (cid:244)1, as seen in Figure 3B. If a
`fast refolding process ((cid:244)3 or (cid:244)2) also led to native protein
`production, a significant amount of native-state amplitude
`would be observed during step 2 at the earliest time point (ti
`) 10 s), which is not the case here.
`Variation of Initial Denaturant Concentrations. To address
`whether residual structure in the unfolded state under high
`denaturant concentration is responsible for the observed
`differences in the (cid:244)2 process for the WT and K97I proteins,
`a series of experiments were performed at 18 (cid:237)M protein in
`which the initial GdnHCl concentrations in the refolding
`reactions were varied. The amplitudes and relaxation times
`for the refolding processes for both WT and K97I are
`independent of the initial unfolding conditions.
`Aggregation. The final protein concentration in refolding
`was varied to determine whether protein association could
`account for the observed differences in the A2 amplitude and
`the (cid:244)2 relaxation time at 0.4 M final GdnHCl for K97I as
`compared to WT (Figure 4). We observed protein association
`during K97I refolding by two detection methods:
`(1)
`fluorescence at 293 nm excitation, 320-400 nm emission,
`and (2) light scattering at 500 nm.
`When the refolding relaxation times (cid:244)3 and (cid:244)2 of WT and
`K97I are observed by stopped-flow fluorescence with 0.4
`M final GdnHCl, only the (cid:244)2 process of K97I is dependent
`on protein concentration (Figure 5). The (cid:244)2 rate decreases
`2-fold and the normalized amplitude A2 (PMT volts/(cid:237)M)
`decreases 40% as the protein concentration is increased from
`3 to 42 (cid:237)M.
`Protein association during refolding in 0.4 M final GdnHCl
`was followed more directly by stopped-flow light scattering
`at 500 nm. A plot of the signal change at 500 nm as a
`function of refolding time for WT and K97I is given in
`Figure 6. As IL-1(cid:226) contains no chromophores that either
`absorb or emit light at this wavelength, any observed signal
`change can be attributed to particulate formation as a result
`of protein aggregation. Over the range of protein concentra-
`tions measured, 3-48 (cid:237)M, WT did not show any appreciable
`increase in light scattering. At 18 (cid:237)M K97I, there is no
`measurable change in the optical signal with time. However,
`at higher protein concentrations, K97I displays measurable
`light scattering occurring within the first seconds of acquisi-
`tion.
`Light scattering behavior during refolding was also studied
`while first varying the initial GdnHCl concentration (Figure
`7A) and then varying the final GdnHCl concentration (Figure
`7B) with a constant protein concentration of 0.55 mg/mL
`K97I. Variations in the initial GdnHCl did not affect the
`light scattering behavior of K97I at either 0.4 or 0.6 M final
`GdnHCl (Figure 7A). This is consistent with the indepen-
`dence of the folding process to initial unfolding conditions
`as monitored by fluorescence (see above). However, increas-
`ing the final GdnHCl concentration in the refolding process
`
`FIGURE 5: Effect of increasing protein concentration from 3 to 48
`(cid:237)M on measured folding parameters WT A3/(cid:244)3 (0), WT A2/(cid:244)2 ("),
`K97I A3/(cid:244)3 (4), and K97I A2/(cid:244)2 (3). (A) Amplitudes A3 and A2 of
`respective kinetic processes (cid:244)3 and (cid:244)2 measured during the refolding
`of WT and K97I at 3-48 (cid:237)M protein. (B) Refolding relaxation
`times (cid:244)3 and (cid:244)2 measured during the refolding of WT and K97I at
`3-48 (cid:237)M protein. K97I amplitude A2 decreases 2-fold and K97I
`relaxation time (cid:244)2 increases 2-fold as concentration is increased,
`indicating an association process during K97I refolding not present
`in WT refolding.
`
`FIGURE 6: Stopped-flow light scattering behavior during refolding
`of WT and K97I at increasing protein concentrations from 3 to 48
`(cid:237)M. (Left panel) Time dependence of light scattering in WT
`refolding versus protein concentration. (Right panel) Time depen-
`dence of light scattering in K97I refolding versus protein concentra-
`tion. Both panels A and B are represented on the same scale. At a
`final GdnHCl concentration of 0.4 M GdnHCl and protein
`concentrations >18 (cid:237)M, K97I exhibits light scattering behavior
`consistent with aggregation.
`
`demonstrated a reduction in the light scattering amplitude,
`which is due to a drop in aggregation (Figure 7B).
`
`Page 5
`
`

`

`580 Biochemistry, Vol. 39, No. 3, 2000
`
`Finke et al.
`
`FIGURE 9: Analysis of stopped-flow light scattering data during
`refolding of K97I to 0.4 M GdnHCl and 42 (cid:237)M K97I: Light
`scattering intensity in volts over 0-2 s after mixing and time
`derivative of light scattering in ¢volts/¢time over 0-2 s after
`mixing where ¢time interval ) (0.005 s. The decrease of the time
`derivative, taken together with the observed folding rates, supports
`a model in which aggregation takes place before stable IL-1(cid:226)
`structure formation.
`
`K97I concentration. The percent of K97I aggregation at any
`final protein concentration in Figure 8 has been shown to
`be independent of the initial concentration of unfolded protein
`prior to dilution and refolding (data not shown).
`
`DISCUSSION
`
`Protein aggregation is a problem in the recovery of
`bioactive recombinant proteins and may also lead to patho-
`logical conditions in vivo and in vitro. Although many
`proteins aggregate under native conditions, it may be that
`pathological or recombinant protein aggregation reactions
`are mediated through association of a nonnative species (7,
`11, 12, 19, 30-38). If nonnative states are responsible for
`aggregation, it is essential that they be characterized. In the
`current study, we performed kinetic experiments under
`conditions where folding and aggregation compete to gain a
`better understanding of these processes.
`Aggregation Occurs Prior to Stable Intermediate Forma-
`tion. At elevated protein concentrations the mutation of K97
`to isoleucine affects early events in the folding reaction of
`IL-1(cid:226) (Figures 4 and 5), particularly the (cid:244)2 folding process
`and the extent of self-association. With regard to increasing
`protein concentration, the correlation between the lowered
`amplitude and rate of the (cid:244)2 refolding process (Figure 5) and
`the increasing amplitude of light scattering (Figure 6)
`indicates that folding and aggregation are competing pro-
`cesses in K97I.
`Figure 9 shows the intensity of light scattering and its time
`derivative plotted as a function of time. The time derivative
`of the observed light scattering signal is directly proportional
`to the concentration and the square of the molecular weight
`of the species involved in aggregation (39). The aggregation
`rate should be maximal under conditions where the aggregat-
`ing species is most highly populated, as the aggregation
`reaction is at least second-order in the concentration depen-
`dence. The time derivative of the light scattering intensity
`is dependent on the aggregation rate, and consequently the
`concentration of the associating species. Therefore, if ag-
`
`FIGURE 7: Effect of initial and final GdnHCl concentration during
`refolding of K97I on stopped-flow light scattering behavior. (A)
`Light scattering at initial GdnHCl concentrations 2.2, 3.0, and 4.0
`M and final concentrations 0.4 and 0.6 M. (B) Light scattering at
`an initial GdnHCl concentration of 2.2 M and final concentrations
`0.4, 0.6, 0.8, and 1.0 M. Initial GdnHCl concentration does not
`affect the extent of aggregation, while increasing the final GdnHCl
`decreases aggregation.
`
`FIGURE 8: Fraction of K97I aggregated in post-refolding equilib-
`rium relative to soluble K97I as a function of the total concentration
`of K97I. The fraction of aggregated K97I was determined quan-
`titatively through UV absorbance at 280 nm of the dissolved protein
`(O) and qualitatively through solution turbidity at 500 nm (b). The
`relatively low fraction of K97I aggregate (<50%) during refolding
`experiments supports an aggregation model mediated by an early,
`transiently populated species.
`
`Aggregation was quantified by measuring the fraction of
`K97I found in the dissolved and precipitated fractions after
`the completion of folding at 0.4 M final GdnHCl. The
`percentage of protein found in the precipitated fraction was
`assayed by the observed sample turbidity at 500 nm and by
`quantitating the amount of protein remaining in the soluble
`portion (Figure 8). There is good agreement between the two
`sets of data, which indicates a cooperative transition between
`the soluble and insoluble percentage as a function of total
`
`Page 6
`
`

`

`Aggregation of Interleukin 1(cid:226)
`
`Biochemistry, Vol. 39, No. 3, 2000 581
`
`gregation occurs when the unfolded populations predominate,
`the time derivative is expected to start from its maximum
`value and subsequently fall as the species depopulates. If an
`intermediate (I) aggregates, the time derivative is expected
`to rise and fall in concert with the transient population of
`this species. If the native state (N) aggregates, the time
`derivative should initially be zero and increase with a rate
`reflecting the population of the native protein.
`By use of the measured folding rates (Figure 4A) and the
`light scattering data (Figure 6) at 42 (cid:237)M K97I and 0.4 M
`GdnHCl, the most probable aggregating species can be
`identified. The light scattering time derivative falls from an
`initial maximal value to nearly zero at 2 s after the initiation
`of folding (Figure 9). This same result is found at all protein
`and GdnHCl concentrations investigated. Consequently, the
`species whose predicted aggregation explains these observa-
`tions is the unfolded ensemble U. The aggregation of U is
`also supported by the observation that refolding at 0.4 M
`GdnHCl and 42 (cid:237)M K97I produces a relatively low fraction
`(<50%) of aggregated protein (Figure 8), since