From Immunogenic Peptides to Intrinsically Disordered Proteins

Abstract

It is hard to evaluate the role of individual mentors in the genesis of important ideas. In the case of our realization that proteins do not have to be stably folded to be functional, the influence of Richard Lerner and our collaborative work in the 1980s on the conformations of immunogenic peptides provided a base level of thinking about the nature of polypeptides in water solutions that led us to formulate and develop our ideas on the importance of intrinsic disorder in proteins. This review describes how the insights gained into the behavior of peptides led directly to the realization that proteins were not only capable of being functional while disordered, but also that disorder provided a distinct functional advantage in many important cellular processes.

Anti-Peptide Antibodies

One of Richard Lerner’s areas of interest in the early 1980s was the phenomenon of anti-peptide antibodies.^[1] Immunogenic peptides that elicited antibodies that would recognize the protein from which they were derived would be a notable advance in the preparation of synthetic vaccines, and an immunological map of an important vaccine target, influenza hemagglutinin (HA), had been prepared.^[2] In an effort to determine the structural basis for the ability of peptide-elicited antibodies to react with the whole protein, Richard recruited a number of structural biologists to Scripps, including Ian Wilson, who had just solved the crystal structure of HA in the lab of Don Wiley.^[3] However, the static structures obtained by crystallography, for example, for the major antigenic determinant of HA,^[4] did not necessarily capture the structural basis for the anti-peptide antibody phenomenon, which seemed instead to be related to the local mobility of the polypeptide chain, as measured by crystallographic temperature factors.^[5] Richard wanted to explore the situation in solution, which we were able to do using solution NMR spectroscopy. Richard provided unparalleled support for the fledgling NMR unit at Scripps, including a state-of-the-art 500MHz NMR spectrometer for our sole use, an unheard-of advantage at the time, since most spectrometers were associated with entire departments or whole universities and were very hard to access.

Our first task was to examine a synthetic peptide whose sequence was derived from the major antigenic determinant of HA. The dominant opinion in the field in the 1980s was that small peptides were completely unstructured in water solution, although they might form structure in organic solvents. Our NMR analysis of the 9-residue HA peptide (sequence Tyr-Pro-Tyr-Asp-Val-Pro-Asp-Tyr-Ala) revealed a folded structure in water solution, although it was clear that this conformation was only a small part of the overall conformational ensemble of the peptide.^[6] That is, the peptide, while conformationally heterogeneous, showed an NMR-detectable preference for a recognizable element of secondary structure, in this case a β-turn in the sequence Tyr-Pro-Tyr-Asp (YPYD). This was a tantalizing observation that challenged the prevailing dogma of the time and led us to perform an extensive NMR survey of several series of peptides to determine the sequence requirements for stabilization of turn-like structures in water. ^[7–9] Our collaboration with Richard Lerner led to the realization that highly immunogenic peptides frequently populate folded structures in aqueous solution, providing insights into the molecular basis for induction of protein-reactive antipeptide antibodies.^[10–13]

Peptide Models of Protein Folding

Our laboratory became the hub of a major effort to define the conformational preferences of peptide fragments of proteins, including myohemerythrin,^[14–15] plastocyanin,^[16] and myoglobin.^[17–20] The relationship between the presence of conformational preferences for elements of secondary structure and the folding pathways of proteins led us into a major effort over many years to delineate the folding pathway of apomyoglobin (reviewed in ^[21]). Our hypothesis that nascent elements of secondary structure present in parts of the protein sequence would be an important impetus for the initiation of protein folding^[22] initially received some support: fragments of helical proteins contained varying conformational preferences for local helical dihedral angles,^{[14, 17–19]} while fragments of β-sheet-rich proteins showed no such preferences.^[16] However, extensive later work on apomyoglobin showed that the presence of helical propensity in a particular part of the protein was likely not the operative trigger for initiation of folding.^[23] Instead, it appeared that local clusters of residues with long side chains that bury a large surface area upon folding were responsible for the initiation and propagation of folding in apomyoglobin.^[24–25] Our peptide and protein folding projects arose directly from our initial collaboration with Richard Lerner on conformational propensities of immunogenic peptides.

Functional Proteins without Defined Three-Dimensional Structure

The 1990s saw the increasing availability of gene sequences for proteins that were of physiological interest, but which had never been prepared using traditional biochemical methods. In particular, genetic methods (gene deletions or mutations, for example) could be used to map the most important parts of the protein sequence for function. The protein sequences inferred from the gene sequences could be prepared by cloning and expression in bacteria or other media, and their conformational preferences could be assessed, as we had done previously for peptides and unfolded or partly folded proteins. The results were surprising. Many of the proteins we and others examined at this time proved to be largely, if not entirely, unstructured in aqueous solution.^[26–28] That is, important functional elements of the proteins, identified by genetic methods, were not folded into globular structures. Interestingly, in several cases, these sequences became folded upon binding to their physiological partners.^{[26, 28–32]} These observations have given rise to an extensive new field, focused on the characterization of intrinsically disordered proteins, made possible in part by the NMR methods developed during our earlier studies characterizing transient structure in peptides and partly folded proteins.^[33–36] Although many proteins, such as enzymes, scaffolding proteins and membrane proteins, are required to be correctly folded to achieve their function, a large number of proteins are not folded in the absence of their partners. Some proteins do not fold even in the presence of their partners, forming complexes that are termed fuzzy.^[37]

Where are Disordered Proteins Found?

The availability of gene sequences facilitated characterization of several disordered proteins.^{[26–30, 38–41]} However, as more extensive gene sequence data became available, it became possible to correlate the sequence signatures of intrinsically disordered proteins, and to generate computer programs to predict regions of the sequence that are likely to be unable to fold on their own. In 2021, an evaluation of 43 methods of disorder prediction was published.^[42] Analysis of the types of sequences that showed the most propensity for intrinsic disorder revealed fascinating correlations with certain metabolic pathways (such as transcription, translation, and cell cycle factors^[43–44]), with certain disease states (such as cancer^[45] and prion disease^[46]), and with viral proteins.^[47]

Many proteins contain both disordered and structured regions. An example is shown in Figure 1. The transcriptional coactivator CREB-binding protein (CBP) contains 2440 amino acids. Much of the protein consists of intrinsically disordered regions that are rich in polar amino acids or proline, frequently in repeat sequences. Interspersed within the protein are structured domains, many small, and some containing metal cofactors. The largest structured domain is a histone acetyl transferase (HAT). One of the challenges in studying proteins like CBP is that, while structured domains can be characterized by X-ray crystallography, or by NMR if they are sufficiently small, and the disordered regions can be characterized by NMR, CD and SAXS in solution, the characterization of the full length protein and the interactions that may occur between structured and disordered domains cannot be accomplished by X-ray crystallography or NMR alone. The recent advances in cryo-electron microscopy, giving high-resolution structures of large macromolecules, are also not applicable, since the disordered regions are too heterogeneous to be resolved. NMR approaches using inteins or sortase to perform segmental isotope labeling are highly promising for characterization of full-length proteins containing both structured and disordered regions. ^[48–50]

Figure 1.

The amino acid sequence of human cyclic-AMP-response element binding protein (CREB)-binding protein (CBP). Amino acids are color-coded according to the type: green: hydrophilic (Gly, Ala, Asn, Gln, Ser, Thr) plus Pro; yellow: hydrophobic (Val, Leu, Ile, Met, Phe, Tyr); red: acidic (Asp, Glu); blue: basic (Lys, Arg, His); purple: low-frequency (Cys, Trp). Structured domains are indicated by colored boxes below the sequence and mapped to the schematic representation of the domain structure of CBP shown on the left.

Advantages of Disorder

Intrinsically disordered proteins are widespread in eukaryotes and viruses, and to a lesser extent also in bacteria. Their prevalence argues that they confer unique advantages on the organisms that employ them. Some of these advantages are summarized in this section.

(a). Enhancement of Specificity in Complex Formation

Because a disordered sequence is not itself structured, it is easy for it to bind to a partner under the right circumstances. An example is provided by the interaction of the nuclear coactivator binding domain (NCBD) of CBP^[51] (Figure 1). Although it contains some measurable helical structure, this domain is not cooperatively folded, as illustrated by the CD spectrum of its unfolding in urea (Figure 2, green). Its partner protein, activator for thyroid and retinoid receptors (ACTR) is even less structured (Figure 2, black). The complex between these two domains is well-folded, with a cooperative unfolding transition in urea (Figure 2, red).

Figure 2.

(left) Circular dichroism (CD) spectra of ACTR (black), NCBD (green) and the ACTR/NCBD complex (red). (right) Dependence of the CD signal at 222 nm on the concentration of added urea.

The three-dimensional structure of the complex^[51] (Figure 3) shows ACTR wrapped around the NCBD, with both components showing additional structure in the complex compared to the free proteins, which we term mutual synergistic folding. This process results in a highly complementary interaction that would not be possible if the component proteins were individually folded. The surface area buried, approximately 1500 Ų, is much larger than expected for such small proteins: to attain a comparable interaction surface using only fully-folded proteins would require them to be much larger.^[52]

Figure 3.

Structure of the ACTR-NCBD complex.^[51] A. Ribbon representation of the lowest-energy structure in the ensemble of NMR solution structures. Green: NCBD; orange: ACTR. B. Space-filling representation of the NCBD portion of the complex (green) with the ACTR potion shown as a backbone trace (orange). Select ACTR side chains are included to illustrate the extent of the contact surface.

(b). Biological Switches

Particularly in eukaryotic systems, metabolic control is frequently mediated through biological switches that function through post-translational modifications (PTM). Regions of the protein that are intrinsically disordered are frequently the sites for post-translational modifications such as phosphorylation, acetylation, and ubiquitination.^[53] In some cases, the PTM may result in changes in the structural state of the IDP, either increasing the level of structure in the free protein, for example, upon phosphorylation of the 4E-BP2 protein,^[54] or by promotion of disorder, for example, by oxidative stress-related loss of zinc from the redox-regulated chaperone Hsp33.^[55–56] The inherent flexibility of the IDP and the heterogeneity of its conformational ensemble allows the free protein to take up vastly different structures in complex with different partners. An example of this is provided by the transcription factor HIF-1α which, at normal cellular oxygen levels, is hydroxylated in two locations, one by prolyl hydroxylases, leading to degradation of HIF-1α, and one by an enzyme known as factor inhibiting HIF (FIH) on an asparagine residue in the C-terminal activation domain (CTAD). Hydroxylation at Asn803 lowers the affinity of HIF-1α for the TAZ1 domain of CBP: under hypoxic conditions, Asn803 is not hydroxylated, leading to interaction with TAZ1 and promotion of transcription of hypoxia-responsive genes. The segment of HIF-1α CTAD containing Asn803 forms an extended structure in complex with FIH^[57] (Figure 4A), but the same segment forms a helix in the HIF-1α CTAD-TAZ1 complex (Figure 4B).

Figure 4.

Conformation of HIF-1α CTAD in two complexes. A. complex of residues 794–804 of HIF-1α CTAD (green backbone and yellow side chains and CO) with FIH (gray ribbon). B. complex of residues 794–804 of HIF-1α CTAD (green backbone and yellow side chains) with the TAZ1 domain of CBP (gray ribbon). The side chain of Asn803 is shown in black (C), red (O) and blue (N).

(c). Unidirectional Regulation of Signaling

The cell must respond rapidly to conditions such as hypoxia or oxidative stress. This is seen for many of the signaling pathways where IDPs participate: under normal conditions, transcription factors that code for stress responses are synthesized, but are either rapidly degraded (for example, HIF-1α) or sequestered in inactive form until needed (for example, NF-κB). Equally important for cellular homeostasis is the ability for signals to be turned off, rapidly and completely, after the cell has returned to a normal state. For example, following hypoxia, once oxygen levels are returned to normal, the hypoxia response genes must be turned off. The hypoxic response can result in high cellular concentrations of the HIF-1α transcription factor. Binding of HIF-1α to partners such as the TAZ1 domain of CBP/p300 is tight (K_d 10 nM) and spontaneous dissociation of the HIF-1α is very slow. The hypoxic response is terminated by another IDP, CITED2, that functions as a negative feedback regulator which rapidly and efficiently displaces HIF-1α from TAZ1 to downregulate transcription of stress response genes (Figure 5). ^[58–59]

Figure 5.

Schematic representation of the transcription of the CITED2 gene as a response to hypoxia and the negative feedback of the CITED2 protein on the activation of HIF-1α.

CITED2 binds to the TAZ1:HIF-1α complex to form a transient ternary complex and displaces HIF-1α by a facilitated dissociation mechanism.^[59–60] Competition is unidirectional; CITED2 efficiently displaces HIF-1α from the TAZ1: HIF-1α complex but HIF-1α is ineffective in displacing CITED2 from the TAZ1:CITED2 complex. This unidirectional switch depends entirely on the flexibility and multivalency of the HIF-1α and CITED2 ligands, both of which are IDPs, and on allosteric conformational changes in the TAZ1 protein.^[60–61] Although free TAZ1 is cooperatively folded, with added stability provided by three zinc ions coordinated by histidine and cysteine side chains, it shows distinct structural differences in its complexes with different ligands. Figure 6 illustrates the differences in the structure of TAZ1 in the two complexes: the α1 helix of TAZ1 is straight in the HIF-1α complex and bent in the CITED2 complex, and the α4 helix of TAZ1 is longer in the CITED2 complex.

Figure 6.

NMR structures of the complex between TAZ1 of CBP (blue) with (left panel) the transactivation domain of HIF-1α and (right panel) CITED2. HIF-1α and CITED2 are shown colored green at the N-terminus, orange in the middle of the sequence and red at the C-terminus. Arrows point out the differences in the TAZ1 structure between the two complexes.

These structural differences in TAZ1, combined with the ability of multivalent disordered ligands to dissociate partially from their partners, allows the CITED2 to effectively strip the HIF-1α from the TAZ2 without the necessity for prior complete dissociation of HIF-1α. Biophysical evidence in the form of NMR and fluorescence competition experiments^{[59, 61]} clearly show that replacement of HIF-1α by CITED2 takes place with the formation of a transient ternary complex, and that the process is unidirectional: CITED2 replaces HIF-1α, but HIF-1α cannot displace CITED2, despite the fact that the affinities of the binary TAZ1-HIF-1α and TAZ1-CITED2 complexes are the same (K_d = 10 nM). A schematic representation of the mechanism of replacement of HIF-1α by CITED2 is shown in Figure 7. Beyond the CITED2/HIF-1α switch, it is likely that facilitated dissociation processes will be found to play a broad role in regulation of biological processes. Indeed, other recent examples include molecular stripping in the NF-κB/IκB/DNA genetic regulatory network^[62], dissociation of transcription factors from DNA binding sites^[63], and dissociation of high affinity regulatory complexes formed by polyelectrolytes.^[64]

Figure 7.

Schematic diagram showing the mechanism inferred for the replacement of HIF-1α (orange) from its TAZ1 (gray) complex by CITED2 (blue).^[59] The α_A region of the bound HIF-1α is relatively flexible^[65] and is readily replaced by the α_A region of CITED2, forming a ternary complex (TAZ1-HIF-1α-CITED2), which enables competition between the LPQL motif of HIF-1α and the LPEL motif of CITED2 for the same binding site. Binding of CITED α_A and the LPEL motif causes TAZ1 to assume the CITED2-bound structure, weakening the affinity of the remaining α_B and α_C portions of HIF-1α for TAZ1. The stronger thermodynamic coupling (Δg^C) between CITED2 α_A and LPEL, compared to the weaker coupling (Δg^H) between the components of HIF-1α contributes to the overall directionality in preferential formation of the CITED2 complex in the presence of HIF-1α.

Conclusions

The insights into conformational propensities of immunogenic peptides derived from our early collaborations with Richard Lerner played a major part in our understanding of intrinsically disordered proteins and their importance in living systems. We are immensely grateful to Richard for his extensive support and interest in our work, particularly in his generous sponsorship of the biomolecular NMR facility. Truly, we owe Richard a huge debt of gratitude.