Amino acid composition of intrinsically disordered proteins was collected from DisProt database (release version 2022_12) and was analyzed using in-house built Python scripts and Microsoft Excel. The DisProt database amino acid composition was determined for the whole dataset (10,544 records) and a filtered dataset (4158 records), from which duplicates based on the sequence and region ID were removed (Supplementary Table 1).

The cis-Pro neighboring sequence preference with proof by NMR measurements was studied for several IDPs [12, 17, 19, 20, 22, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36] (Supplementary Table 2). As the dataset has increased since our previous work, we have updated the results [17]. The current dataset contains 15 IDPs with 167 central i proline environments, and the proline i $\pm$ 4 environments were investigated. Residues were divided into 7 groups based on the traditional classification of the amino acids: (1) Gly; (2) aliphatic side chain (Ala, Val, Leu, Ile, Met); (3) polar side chain (Ser, Thr, Cys, Asn, Gln); (4) positively charged side chain (Arg, Lys, His); (5) negatively charged side chain (Asp, Glu); (6) aromatic side chain (Phe, Tyr, Trp); and (7) Pro. Two-sided binomial tests at a significance level of 0.05 and 0.1 were conducted for the statistical analysis.

The designed peptides (Table 1) were synthesized using solid-phase peptide synthesis, applying the Fmoc/tBu strategy using a CEM microwave-assisted fully automated peptide synthesizer. The syntheses were carried out at a 0.25 mmol synthesis scale using a TentaGel S Ram resin (Rapp Polymere GmbH, Tübingen, Germany) with a loading of 0.23 mmol/g amino function. The crude peptides were detached from the solid support using TFA (90%) in the presence of water (5%), 1, 4-dithiothreitol (DTT, 2.5%), and TIS (2.5%). The crude peptides were purified using a C18 RP-HPLC on a PerfectSil 100 ODS3 5 µm column (250 $\times$ 10 mm). All compounds were $>$95% pure by HPLC analysis. The cleaved peptides were analyzed by RP-HPLC/MS (A: 0.1% TFA; B: 80% AcN/water) on a PerfectSil 100 ODS-3 5 µm column (250 $\times$ 4.6 mm), with a flow of 1 mL/min. The mass accuracy of the products was determined by ESI-MS using a Waters SQ Detector2 (Waters Corporation, Milford, MA, USA). The mass spectra were recorded in positive ion mode in the 200.0–3000.0 m/z range (Supplementary Figs. 1,2).

Typical NMR samples contained 1 mM peptide in 10% D${}_2$O and 0.05 mM 3-(trimethylsilyl)-1-propanesulfonic acid sodium salt (DSS), and the pH was adjusted to 3.0. All NMR spectra were recorded on a Bruker Avance III 700 spectrometer (Bruker GmbH, Ettlingen, Germany) (700.05 MHz for ${}^1$H; 70.94 MHz for ${}^{15}$N; 176.03 MHz for ${}^{13}$C) using a Prodigy TCI H&F-C/N-D, 5 mm z-gradient probe head. The measurements were conducted at 298 K. ${}^1$H chemical shifts were referenced to the internal DSS standard, whereas ${}^{15}$N and ${}^{13}$C chemical shifts were referenced indirectly via the gyromagnetic ratios.

Resonance assignment and sequential connectivities were determined from classical 2D homonuclear ${}^1$H,${}^1$H-TOCSY (mixing time: 80 ms) and ${}^1$H,${}^1$H-NOESY (mixing time: 250 ms) measurements. The typical spectral resolution was 2048 $\times$ 512, and the measurements were acquired with 8 and 16 transients, respectively. ${}^1$H,${}^{15}$N SOFAST-HMQC, and ${}^1$H,${}^{13}$C-HSQC measurements were performed on peptides with a naturally abundant isotope content. The ${}^1$H,${}^{15}$N SOFAST-HMQC spectra were acquired with 2048 $\times$ 128 resolution, and the number of scans varied between 160–480 to detect the lowly populated minor forms. ${}^1$H,${}^{13}$C-HSQC were acquired using a 2048 $\times$ 256 resolution with 32 transients. All spectra were processed using TopSpin 3.6.0. (Bruker GmbH, Ettlingen, Germany) Peak assignment was completed using Sparky (University of California, San Francisco, CA, USA) [37].

Proline, with 7.41% occurrence, is the 5th most frequent amino acid in IDPs according to DisProt (release version 2022_12) (Fig. 2, Supplementary Table 1). In order to confirm the amino acid preference, the proline preceding and succeeding neighboring residue type was collected for the Pro residues and Pro-Pro motifs (Fig. 2, Supplementary Tables 3,4). Statistical analysis of these data shows that the distribution of amino acid type in the Pro $\pm$ 1 positions differs significantly from the DisProt amino acid distribution (Table 2). This signifies that the number of aliphatic and Pro residues significantly increases in these positions, whereas charged and aromatic residues are significantly less frequent. The number of Gly and polar residues are significantly different in the Pro preceding position, but no significant differences in occurrence can be found for i + 1 position. The composition of X-Pro-Pro and Pro-Pro-X motifs differ as well: in both cases, the number of aliphatic residues and Pro is increased, and interestingly the number of negatively charged residues shows significant deviation from the DisProt database (Supplementary Table 5).

Fig. 2.

Amino acid type occurrences in DisProt database and the Pro and Pro-Pro neighboring residues (denoted as X-P, P-X, X-PP, and PP-X, respectively). Gly (gray), aliphatic (black), polar (green), positively charged (blue), negatively charged (red), aromatic (yellow), and Pro (dark blue).

It is important to note that prolines are often situated in proline-rich regions with repetitive Pro containing motifs that often form polyproline II-type helices. While the number of polyproline (containing consecutive proline residues) sequences longer than 20 residues in the UniProt database (https://www.uniprot.org/) is more than 6000, these motifs are underrepresented in the DisProt database, as here the longest polyproline sequence is only 13 residues long (DisProt ID: DP02591r001).

In order to determine the sequence dependence of the cis-Pro amount (calculated as [cis]/([cis] + [trans])), we updated our previously published dataset of IDPs and expanded our previous dataset to the Pro neighboring i $\pm$ 4 range [17]. The updated dataset now contains 15 IDPs (Supplementary Table 2) with 167 central proline residues. The cis-Pro content depends on the peptide sequence length. Therefore peptides (less than 20 residues) are not included in the dataset [34]. Note that polyproline sequences with several consecutive prolines are also excluded due to ambiguous peak assignment.

The i $\pm$ 4 Pro neighboring sequences contain a total of 1312 amino acids (not considering Pro in the i position), and the amino acid type occurrence slightly differs from DisProt (Fig. 3, Supplementary Table 6); thus, we use the amino acid composition of our overall dataset as a reference.

Fig. 3.

Comparison of amino acid type occurrences. (A) In the current database and the corresponding two subgroups according to the cis-Pro ratio. (B) Amino acid type occurrences in each position in the i $\pm$ 4 regions in the current dataset. Gly (gray), aliphatic (black), polar (green), positively charged (blue), negatively charged (red), aromatic (yellow), and Pro (dark blue).

The Pro neighboring sequences were divided into two groups according to the cis-Pro isomer content (Fig. 3A). For cis-Pro content 10%, the amino acid type occurrence is similar to the overall dataset. However, significant differences are found for the sequences with $>$10% cis-Pro content: the positively charged residues are significantly less. In contrast, prolines occur significantly more than expected (Fig. 3A, Supplementary Table 7).

In the individual positions for the complete dataset (Fig. 3B), some residue types deviate significantly in the occurrence. The largest differences can be found for Pro and polar residues, where more than a 12% deviation between the highest and the least populated positions can be observed. In addition, two-sided binomial tests at a significance level of 0.05 were conducted to investigate which amino acid types (Fig. 3B) alter in the individual positions. We found that the distant positions (i-3, i-4) do not deviate significantly from the reference. Negatively charged residues are significantly more frequent in the i-2 position. There are significantly more polar residues in i-1 and aliphatic amino acids in i + 1 position. Prolines occur significantly more in i + 4 position and less in i $\pm$ 1 positions. Aromatic amino acids are more frequent in i + 2, and less in i + 3 positions.

In sequences with more than 10% cis-Pro content-considering a 0.05 significance level—the positively charged residues are significantly less frequent in i-3 and i-1 positions (Fig. 4). In the i-1 position, aromatic and polar amino acids occur significantly more often, whereas the number of Pro is reduced. Aromatic residues are common as well in the i + 1 position. Note, Pro occurrence is significantly higher in i + 1, i + 2, and even in i + 4 positions, and the number of Gly is also increased at i $\pm$ 4.

Fig. 4.

Acid-type occurrences in more than 10% cis-Pro containing sequences in i $\mathrm{\pm }$ 4 range. Amino acid type occurrences in each position in the i $\pm$ 4 regions in the current dataset. Gly (gray), aliphatic (black), polar (green), positively charged (blue), negatively charged (red), aromatic (yellow), and Pro (dark blue).

Compared to our previous work, the general rules hold [17]: The number of aromatic amino acids adjacent to Pro (i $\pm$ 1) is increased. In Pro-Pro motifs, the cis-Pro amount is higher for cis-Pro-trans-Pro than for trans-Pro-cis-Pro. Positive charges at i-3 and i-1 indicate a decreased cis-Pro content.

The increased number of Asp and Glu in i-2, i-1, and i + 3 positions does not hold at 0.05 significance level, only at p = 0.1.

To validate our findings, model peptides with designed mutations were synthesized (Table 1). All peptides were 12–15 residues long and were enriched in Lys and Arg residues to test the effect of the positive charge. Since aromatic residues in the i $\pm$ 1 position of the Pro has the strongest effect, Peptide I.A contains two Phe in the Pro vicinity. To test the effect of a positive charge, instead of the aromatic sidechain on the cis-Pro content, a Phe5Arg mutation (Peptide I.B) was designed. Peptide II.A contains two prolines: in the case of Pro6, two Phe residues are located in the Pro preceding region: in the i-1 and the more distant i-4 position. In this case, the i + 1 residue is a negatively charged Asp. In the Pro10 neighboring region, there are no aromatic rings, the i-4–i-1 region contains residues that do not significant affect the cis-Pro ratio, and the i +1–i +4 positions are enriched in positively charged residues. In peptides II.B-II.D Pro-Ala mutations were introduced. Peptide III contains two prolines: the i $\pm$ 2 regions of Pro6 are similar to Pro6 in Peptide II.A (Gly-Phe-Pro-Asp/Glu-Lys). The main differences are in positions i-4 and i-3, where Lys is located. Pro14 is close to the C-terminus, and the preceding i-4–i-1 region contains positive charges in positions that should not influence the cis-Pro content.

NMR measurements were performed to test these assumptions. Peak assignment was performed using 2D homo- and heteronuclear measurements on peptide samples with natural isotope abundance. Since the concentration of the minor form is low, approximately 50–200 µM for 1 mM protein, and ${}^{13}$C isotope abundance is ~1%, measurement times are lengthy. Pro isomerization results in a peak multiplication for the Pro neighboring residues (Fig. 5, Supplementary Fig. 3). For quantitative cis-Pro determination, the Pro H$\delta$2-C$\delta$ and H$\delta$3-C$\delta$ cross peaks were integrated on the ${}^1$H,${}^{13}$C-HSQC spectrum, since these peaks could be unambiguously assigned as signal overlap rarely occurs in this region (Table 3).

Fig. 5.

Cis-Pro detection on 1D ${}^{\mathrm{1}}$H and 2D ${}^{\mathrm{1}}$H,${}^{\mathrm{13}}$C-HSQC spectra of Peptide I.A. (A) 1D ${}^1$H spectrum of Peptide I.A. Proline H$\alpha$ (4.2–4.6 ppm) and sidechain H$\beta$ and H$\gamma$ (~1.8–2.2 ppm) peak assignment is not possible due to signal overlaps. Pro H$\delta$ region is highlighted using a black rectangle. (B) Peak multiplication in the Pro H$\delta$ region of Peptide I.A: the population of cis-conformer is ~26%. (C) Pro H$\delta$-C$\delta$ region on the ${}^1$H,${}^{13}$C-HSQC spectrum of Peptide I.A (blue) and I.B (red). The Pro preceding Phe5 in Peptide I.A was mutated to an Arg residue, causing a significant decrease in the cis-Pro content.

In Peptide I.A Pro6 has phenylalanine in i $\pm$ 1 position; therefore, the cis-Pro content is increased by 26%. If the disadvantageous Phe to Arg mutation occurs at i-1 position (Peptide I.B), the cis-Pro ratio decreases to 9% (Fig. 5C).

Peptide II.A contains two prolines where two sets of minor peaks of different signal intensities are detected. Here, the major Pro6 and Pro10 H$\delta$2-C$\delta$ and H$\delta$3-C$\delta$ cross peaks overlap on the ${}^1$H,${}^{13}$C-HSQC spectrum. Therefore, Pro-Ala mutants were synthesized for unambiguous peak assignment. The Pro6Ala and Pro10Ala mutations cause the absence of the corresponding cis-Pro isomer, indicating that the high-intensity cis-Pro peak (20%) belongs to Pro6, while the cis-P10 content is significantly lower (9%). This agrees with our previous observations: the phenylalanine in i-1 position indicates an increased minor content, whereas sequential enrichment in positively charged residues hinders the cis-Pro formation.

Peptide III. contains two proline residues. Pro6 succeeds an aromatic phenylalanine which is advantageous for a high cis/trans ratio (Supplementary Fig. 3B). However, the several positively charged lysines (i-4, i-3, i + 2, i + 4) reduce this effect, resulting in a 16% cis-Pro amount. The Pro14 neighboring sequence lacks aromatic residues, and the unfavorable positive charge sidechain is located in the neutral i-2 position producing a 10% minor isomer.