Heterologous expression of naturally evolved putative de novo proteins with chaperones.

Over the past decade, evidence has accumulated that new protein-coding genes can emerge de novo from previously non-coding DNA. Most studies have focused on large scale computational predictions of de novo protein-coding genes across a wide range of organisms. In contrast, experimental data concerning the folding and function of de novo proteins are scarce. This might be due to difficulties in handling de novo proteins in vitro, as most are short and predicted to be disordered. Here, we propose a guideline for the effective expression of eukaryotic de novo proteins in Escherichia coli. We used 11 sequences from Drosophila melanogaster and 10 from Homo sapiens, that are predicted de novo proteins from former studies, for heterologous expression. The candidate de novo proteins have varying secondary structure and disorder content. Using multiple combinations of purification tags, E. coli expression strains, and chaperone systems, we were able to increase the number of solubly expressed putative de novo proteins from 30% to 62%. Our findings indicate that the best combination for expressing putative de novo proteins in E. coli is a GST-tag with T7 Express cells and co-expressed chaperones. We found that, overall, proteins with higher predicted disorder were easier to express. STATEMENT: Today, we know that proteins do not only evolve by duplication and divergence of existing proteins but also arise from previously non-coding DNA. These proteins are called de novo proteins. Their properties are still poorly understood and their experimental analysis faces major obstacles. Here, we aim to present a starting point for soluble expression of de novo proteins with the help of chaperones and thereby enable further characterization.

INTRODUCTION

De novo genes originate from intergenic or non‐coding DNA regions , , , , , , in contrast to genes that emerge by duplication , or rearrangement from existing gene fragments. Therefore, recent, true de novo genes have no precursor by definition and have not been subjected to selection for particular structures or functions for long, if at all. Due to their recent emergence, de novo genes tend to be shorter, evolve more rapidly, and have lower expression than established genes. , Short‐length and accelerated evolution make it difficult to reliably detect (or reject) homologs of orphan genes and thereby identify true de novo genes. By combining homology and synteny based approaches for de novo gene identification, the origin of de novo genes can be detected more accurately.

Several de novo protein‐coding genes have been identified and confirmed across a wide range of eukaryotes. , , , , , , , , , , These de novo genes were mainly analyzed with comparative genomics and transcriptomics. A recent study by Grandchamp et al. showed that proto‐genes, an intermediate step in de novo gene emergence, contain regulatory sequences similar to established genes. Depending on the genomic position of the recently emerged proto‐gene, introns may already be present in the proto‐gene, making it harder to distinguish from established genes. However, without experimental evidence on structure and function, our evolutionary understanding of how de novo proteins emerge, remains incomplete.

Difficulties in handling de novo proteins, together with the novelty of the research area, might be the reason for the lack of experimental studies on de novo proteins. So far only two de novo proteins were expressed and characterized experimentally, Goddard (Gdrd) and Bsc4. In both cases, the expressed de novo protein was difficult to analyze due to unstable or incorrect folding (Bsc4) or unusual behavior in SDS‐PAGE (Gdrd). Compared to well‐studied proteins with expression and purification data available, de novo proteins tend to behave differently when using standard protocols.

Several studies, foremost some from the laboratory of Dan Tawfik, , , , inspired us to apply co‐expression with chaperones to achieve soluble expression of de novo proteins. Since de novo proteins evolve rapidly by becoming coding from scratch, they probably lack a stable structural configuration and contain high amounts of disorder. , Those properties determine the levels of soluble and insoluble fractions of a protein during in vitro experiments and could explain the obstacles faced during their expression. , On the other hand, it is not yet clear if de novo proteins undergo a similar hindrance in their native organism or only in the expression hosts. While Tawfik and colleagues used chaperones to explore the sequence space of enzymes and enable soluble expression of mutants, , , we hypothesized that de novo protein expression might also profit from chaperones. With their “emergence from dark genomic matter” in the DNA and predicted lack of stability and high disorder, de novo proteins are prospective targets for chaperones because their solubility can be increased. , Increased solublity can be relevant for protein purification and any follow‐up experiments.

The chaperonin GroEL and its co‐chaperone GroES are found throughout the bacterial domain, while their homologs, HSP60 and HSP10, respectively, are found in eukaryotes. GroEL/GroES play a pivotal role in the translocation, dis‐aggregation, function, and folding of newly synthesized peptides after translation. , , ,

The other chaperone system used here is DnaK, DnaJ, and GrpE (homologous to HSP70 and HSP40 in eukaryotes). For simplicity we will refer to the chaperone system GroEL/ GroES as only GroEL and to DnaK, DnaJ, and GrpE as DnaK only. While the GroEL system targets misfolded and unfolded proteins, DnaK can refold an already aggregated protein to its native state using ATP (see Figure 1a). , , , The two different chaperone systems can be exploited for challenging heterologous expression of proteins which are foreign to the host and thus prevent misfolding and aggregation which is often associated with heterologous expression. , , , , ,

FIGURE 1

(a) Mechanism of chaperone assisted protein folding after Thomas et al. The nascent protein is bound by the DnaK/J complex and release is triggered by GrpE under ATP hydrolysis. After release, the protein is either correctly folded, degraded (proteolysis), or remains unfolded. The unfolded protein can either aggregate or bind to the GroEL/ES complex. GroEL/ES either releases the folded protein by ATP hydrolysis or the protein is degraded. (b) Overview of the workflow on de novo protein expression: We first selected candidate proteins from Drosophila melanogaster (11, including Atlas) and 10 from Homo sapiens from a pool of putative de novo genes for expression. The 21 sequences were codon optimized for E. coli and ordered from Twist. For expression, different tags (GST and His), different E. coli expression cells (star, T7), and different chaperones (GroEL and DnaK systems) were tested. The success of protein expression was verified by Western blot (WB) and mass spectrometry (MS)

For this study, we used 21 putative de novo proteins, 11 from Drosophila melanogaster (termed here as DM1‐10 and Atlas) and 10 from Homo sapiens (termed here as HS1‐10) as shown in Figure 1b). The sequences of all used putative de novo proteins can be found on Zenodo (https://doi.org/10.5281/zenodo.6512224), for genomic location and official gene names see Table S1. These de novo proteins have been recently published by Heames et al. and Dowling et al., respectively. Additionally, we tested our method on a recently published and better characterized putative de novo protein from D. melanogaster, called Atlas. Atlas appears to function as a DNA binding protein that facilitates the packaging of chromatin in developing D. melanogaster sperm. Since experimental work with de novo proteins is still underrepresented (compared with computational studies) and challenging, we want to propose a guideline for successful expression of putative de novo proteins in E. coli. We combined different chaperone systems (GroEL and DnaK) with different combinations of E. coli strains (BL21 Star™ [DE3] and T7 Express) in order to express putative de novo proteins solubly. To verify successful expression of target proteins, Western blots were performed and samples sent for tryptic digest followed by mass spectrometry. We identified the best combination for expression of putative de novo proteins in E. coli. After first expressions with His‐tag alone resulted in soluble expression for only 1/21 proteins, we increased the total number of solubly expressed putative de novo proteins to 13/21 with GST‐tag and chaperones. The different chaperone systems increased or enabled soluble expression in four cases, while DnaK only helped in two, GroEL in all of those four.

RESULTS

Structural content of the putative de novo proteins

Disorder predictions

We performed disorder predictions with IUPred2a , on all candidate de novo proteins. For this we calculated the percentage of residues predicted to be disordered (Figure 2), as opposed to the overall average disorder score (Figure S1). This allows direct comparison to secondary structure predictions (Figure 3). Our first objective here was to choose candidate de novo proteins with different levels of intrinsic disorder to observe any difference in their ability to express. If any trend in predicted disorder and soluble expression or susceptibility to chaperones was observed, this could help choosing promising candidates for characterization in future experiments. The predicted disorder ranged from around 3%–100% as shown in Figure 2. DM5 was predicted to have least disorder content, while DM6, DM3, HS10, and DM8 appear to be entirely disordered. The putative de novo protein Atlas has predicted disorder of 60%.

FIGURE 2

Percentage of disorder as calculated with IUPred2a. All candidate de novo proteins used for expression experiments ordered by their disorder level from the left to right. Unicolor bars belong to the unsuccessfully expressed proteins, striped bars to the successfully expressed ones

FIGURE 3

Percentage of random coils, α‐helices, and β‐sheets predicted by Porter 5.0 for each de novo protein candidate. Left to right following increasing disorder level based on Figure 2

Secondary structure predictions

Predictions of secondary structure elements were performed using Porter 5.0 , for all candidate proteins and are shown in Figure 3. The predicted random coils should be equivalent to the disordered regions predicted by IUPred2a. While the results indicate a high amount of random coils for most candidates, they do not completely follow the trend of the disorder predictions by IUPred2a (compare Figure 2). DM3, for example, is predicted to be ∼100% disordered by IUPred2a, while, on the other hand, it is predicted to have over 20% β‐sheet and ∼70% random coils by Porter 5.0.

Our goal was to choose a cohort of de novo proteins that consist of a diverse range in composition of structural elements. We assumed that a protein containing more secondary structure elements should be better accessible for soluble expression with chaperones. Notably, DM1 (∼70% α‐helical), DM2 (∼70% α‐helical), DM4 (∼55% α‐helical, ∼10% β‐sheets), DM5 (∼60% α‐helical, ∼10% β‐sheets), and DM10 (∼70% α‐helical, ∼10% β‐sheets) are predicted to have secondary structure contents of 50% or more, with α‐helices to be more frequent than β‐sheets and less than 50% random coils. HS1 (∼50% random coils, ∼45% α‐helical, ∼5% β‐sheets), HS3 (∼70% random coils, ∼25% α‐helical, ∼5% β‐sheets), HS4 (∼65% random coils, ∼30% α‐helical, ∼5% β‐sheets), HS5 (∼55% random coils, ∼5% α‐helical, ∼40% β‐sheets), HS6 (∼60% random coils, ∼5% α‐helical, ∼35% β‐sheets), HS7 (∼70% random coils, ∼5% α‐helical, ∼25% β‐sheets), DM3 (∼70% random coils, ∼30% sheets), DM7 (∼65% random coils, ∼10% α‐helical, ∼25% β‐sheets), and DM9 (∼60% random coils, ∼10% α‐helical, ∼30% β‐sheets), on the other hand, are predicted to be mostly random coils (disordered) with otherwise higher amounts of β‐sheets predicted. DM6 (∼90% random coils), DM8 (∼100% random coils), HS2 (∼85% random coils), HS9 (∼95% random coils), and HS10 (∼100% random coils) are predicted to contain more or less only random coils.

Expression of putative de novo proteins

Candidate proteins of Drosophila melanogaster

Our initial approach was similar to the successful expression of characterized putative de novo protein Gdrd. Therefore, we aimed to express our 11 putative de novo protein candidates with an N‐terminal 6xHis‐tag in E. coli BL21 Star™ (DE3) cells, and verify expression via SDS‐PAGE and mass spectrometry. However, for our candidates, the expression level was either very low or not detectable, as can be seen in Figure S3. We switched to different E. coli cells (T7 Express), but expression remained unsuccessful. Shifting from an N‐terminal 6xHis‐Tag to a C‐terminal 6xHis‐tag showed similar negative results. Considering the size and levels of disorder, we switched to a larger tag for increased solubility and stability, choosing an N‐terminal GST‐tag. In this way, we were able to observe a higher success rate in soluble expression of our target proteins. But not all proteins could be expressed at satisfying levels, especially solubility needed to be increased for some (Figure S3).

Inspired by successful work carried out by Tawfik et al., , , we hypothesized that chaperones could improve thermodynamic stability of these evolutionarily young proteins thus enabling their soluble expression. We repeated our experiments with the addition of the two chaperone systems (i) GroEL and (ii) DnaK. We were able to increase the number of solubly expressed de novo candidate proteins of D. melanogaster using the combination of either GroEL or DnaK and N‐terminal GST‐tag (see Figure 4). However, for the candidate proteins DM6, DM9, and DM10 no soluble expression was achievable, despite the use of different tags, strains, or chaperones. Only in the case of Atlas, the combination of N‐terminal 6xHis‐tag and GroEL worked best. We tested all combinations in BL21 Star™ (DE3) and T7 Express E. coli cells. Six candidate proteins were expressed in T7, two were expressed in BL21 Star™ (DE3) cells. Three proteins were not expressable in either strain. In summary, with the combination of chaperones and switching to N‐terminal GST‐tag, we were able to express 8/11 of the D. melanogaster putative de novo protein candidates (see Table 1).

FIGURE 4

Western blots with anti‐His antibody. Boxes indicate the height of the target protein band: (34 kDa): highest solubilty without chaperones, then GroEL, then DnaK; highly soluble. (36 kDa): only insoluble, even with chaperones. (33 kDa): DnaK highest solubilty, then base, then GroEL; very soluble. (34 kDa): DnaK highest solubilty, then GroEl, then base; very insoluble. (39 kDa): GroEL only one with soluble fraction, runs a bit high. (36 kDa): Dnak highest solubilty, then base, then GroEL very soluble. (37 kDa): all similar, different expression levels, first base, then GroEL, then DnaK; more insoluble. Atlas (20 kDa): GroEL highest solubilty, nothing in base and DnaK

TABLE 1

Expression conditions and results of D. melanogaster de novo proteins. Base = no chaperones, GroEL = GroEL/ES, DnaK = DnaK/J/GrpE. Plus signs mean visible expression, two plus signs strong expression, 0 means no visible expression. Arrows indicate the change in expression with added chaperones in comparison with base

Protein	Cell/tag	Base	GroEL	DnaK	GroEL effect	DnaK effect	Disorder (%)
DM1	T7/GST	+ +	+ +	+	−	↓	15
DM2	Star/GST	+ +	+ +	0	−	↓	13
DM3	T7/GST	+ +	+	+ +	↓	−	97
DM4	T7/GST	+ +	+	+	↓	↓	21
DM5	T7/GST	0	+	0	↑	−	3
DM6	−/−	0	0	0	−	−	97
DM7	T7/GST	+ +	+	+ +	↓	−	35
DM8	T7/GST	+	+	+	−	−	100
DM9	−/−	0	0	0	−	−	49
DM10	−/−	0	0	0	−	−	15
Atlas	Star/6xHis	0	+ +	0	↑	−	56

Comparison of different chaperone conditions for D. melanogaster proteins

Western blots were used for comparison of the soluble expression levels with and without chaperones, in order to test our hypothesis that chaperones would increase soluble expression of the target proteins. The optimal conditions identified by SDS‐PAGEs were repeated under three settings: (i) without chaperones (base), (ii) with GroEL, and (iii) with DnaK. Surprisingly, we did not observe increased solubility for most putative de novo proteins when adding chaperones (see Figure 4 and Table 1).

In contrast, we observed soluble expression for most proteins without chaperones, for example, DM1, DM2, DM3, DM4, and DM7. In combination with GroEL, the intensity of the bands in the soluble fraction and therefore amount of soluble protein, even decreased for DM3, DM4, and DM7. For DM2 and DM5, the amount of soluble protein increased when co‐expressed with GroEL. When DnaK was co‐expressed, protein solubility either appeared to decrease (DM1, DM2, and DM4) or was similar to the base (DM3 and DM7). DM8 showed similar soluble expression for all three conditions with most of the protein being insoluble. In the case of Atlas and DM5, soluble protein expression was increased or enabled with the addition of the GroEL chaperone system while DnaK and base expression resulted in no or very little soluble protein. While we cannot confirm that co‐expression with DnaK in fact decreases the amount of soluble protein (DM1, DM2, and DM4), we do not see increased soluble expression for any of the candidate proteins in the presence of DnaK as we do for GroEL (DM5 and Atlas).

Candidate proteins of Homo sapiens

The 10 putative human de novo proteins were expressed following the same protocol as the D. melanogaster proteins by combining the different E. coli expression cells, tags, and chaperone systems (Figure S4). We detected a similar trend here as for the D. melanogaster proteins (N‐terminal GST‐tag in E. coli T7 express cells; see Table 2). One protein (HS8), however, was only weakly expressed with an N‐terminal 6xHis‐tag but using also E. coli T7 express cells. Without the addition of chaperones only HS7, HS8, and HS10 were successfully expressed and soluble. After co‐expression with chaperones, as described for D. melanogaster proteins, two more H. sapiens proteins could be expressed. Unfortunately, H. sapiens protein candidates HS1, HS3, HS4, HS5, and HS6 showed no expression at all, even with chaperones. In total, we were able to express 5 out of 10 putative de novo proteins following our protocol (see Table 2).

TABLE 2

Expression conditions and results of H. sapiens de novo proteins. Base = no chaperones, GroEL = GroEL/ES, DnaK = DnaK/J/GrpE. Plus signs mean visible expression, two plus signs strong expression, 0 means no visible expression. Arrows indicate the change in expression with added chaperones in comparison with base

Protein	Cell/tag	Base	GroEL	DnaK	GroEL effect	DnaK effect	Disorder (%)
HS1	−/−	0	0	0	−	−	59
HS2	T7/GST	0	+	+	↑	↑	84
HS3	−/−	0	0	0	−	−	60
HS4	−/−	0	0	0	−	−	54
HS5	−/−	0	0	0	−	−	51
HS6	−/−	0	0	0	−	−	51
HS7	T7/GST	+ +	+ +	0	−	↓	82
HS8	T7/6xHis	+	+	+	−	−	60
HS9	T7/GST	0	+	+	↑	↑	56
HS10	T7/GST	+	+	+	−	−	99

Comparison of different chaperone conditions for H. sapiens proteins

Western blots were used for comparison of the three different chaperone expressions (i) base, (ii) GroEL, and (iii) DnaK, as described above. Two out of the five successful candidates (HS2 and HS9) showed very weak or no soluble expression without chaperones, but solubility could be increased with both chaperone systems. HS8 and HS10 showed low soluble expression overall, but no change in solubility was visible when co‐expressing with either chaperone system. The candidate de novo protein HS7 already showed strong soluble expression at base (Figure 5). However, the addition of GroEL seemed to increase soluble expression further, while DnaK co‐expression led to low or no protein being detected. Overall, the trend observed for the D. melanogaster proteinś was consistent with the trend observed for the H. sapiens proteins. GroEL increased soluble expression for most putative de novo proteins while DnaK lacked substantial influence on protein solubility.

FIGURE 5

Western blots with anti‐His antibody. Boxes indicate the height of the target protein band: (44 kDa): upper bands (lower are degraded protein or double bands) most in DnaK, then GroEL, then base; very insoluble. (50 kDa): GroEL best, then base, nothing in DnaK. Possible protein degradation; very soluble. (16 kDa): upper bands most in DnaK, then GroEL, then base; very insoluble. (42 kDa): upper bands (lower are degraded protein or double bands) most in DnaK, then GroEL, then base; very insoluble. (43 kDa): upper bands (lower are degraded protein or double bands) most in GroEL, then DnaK, then base; very insoluble

DISCUSSION

De novo proteins have first been detected more than a decade ago and the mechanism of their emergence has been studied intensely ever since. , Still, there are concerns (i) regarding the reliability of their computational identification , , and (ii) if and how they code for functional proteins. To shed light on these concerns, de novo proteins need to be studied experimentally and theoretically. The handling of de novo proteins by heterologous expression and purification is often difficult because solubility is low and purification yields little amounts and potentially unstable proteins. Moreover, identifying the function of these young genes, is another challenging task. In this study, we present a guideline for expressing de novo proteins in E. coli.

Expression cells

E. coli is the most widely used model organism for recombinant expression. However, foreign proteins can be toxic to E. coli by interfering with the physiology or leading to protein aggregation. This may result in low expression yields, growth defects, or even cell death. , , Optimized expression hosts and plasmids , , or chaperones can be used to overcome the expression issues caused by proteins which are a metabolic burden for the host. Here, we used two different types of the E. coli strains (DE3): BL21 Star™ and T7 Express. Both strains resulted in effective protein expression and a relatively high yield of the de novo proteins, with T7 Express being the best option. The de novo proteins studied here are possibly a toxic, metabolic burden to the E. coli cells, suggesting T7 cells are the better choice of expression cell. BL21 Star™ (DE3) contains a T7‐RNA‐polymerase under control of lacUV5 promoter together with higher mRNA stability. This leads to stable mRNA transcripts and higher amount of target protein. However, BL21 Star™ (DE3) cells have increased basal expression of heterologous genes and cannot express toxic genes. In contrast, the T7 Express cells have a reduced basal expression of target proteins than BL21 Star™ (DE3) cells. Therefore, toxic proteins can be expressed better in T7 cells compared with BL21 Star™ (New England Biolabs).

Comparing different protein tags

Based on our study, an N‐terminal GST‐tag was the more appropriate choice than a 6xHis‐tag. Some de novo protein candidates are quite small (8–12 kDa), so a larger tag like GST might already stabilize in a chaperone‐like manner. , However, Atlas and HS8, that is, 2/21, were only expressed with an N‐terminal 6xHis‐tag. With a mass of only 1 kDa, 6xHis‐tag is the better choice for further structural characterization using circular dichroism (CD), multi‐angle light scattering (MALS) or nuclear magnetic resonance (NMR), since a small tag has less influence on protein folding. In contrast, the larger GST‐tag needs to be cleaved for most follow‐up experiments. When removing the tag, the de novo protein might behave differently and could degrade or aggregate.

Influence of chaperones on protein expression and solubility

Our Western blot results indicate that GroEL slightly outperforms DnaK in terms of increased protein solubility. In some cases, both chaperone systems increase or enable soluble expression (HS2 and HS9, 2/21) but for most proteins GroEL leads to more soluble protein than DnaK (DM1, DM2, DM5, Atlas, and HS7, 5/21) (Figure 6). DnaK requires easily accessible hydrophobic fragments that can be predicted from the protein sequence, while GroEL demands no defined binding motifs. However, in the case of our proteins, we found no connection between predicted DnaK binding sites and influence of DnaK on protein expression level (Figure S2). While GroEL could increase solubility for single de novo proteins in 5/21 cases, this system was not successful in Heames et al., in which we used a library of 1800 putative de novo proteins (4–8 kDa) in a cell‐free expression system.

FIGURE 6

Fractions of soluble and insoluble expression for candidate de novo proteins with expression of GroEL or DnaK and without (base). Values are intensities of Western blot bands

In some cases (DM1, DM2, HS7, and Atlas), DnaK decreased the solubility of proteins that expressed soluble without chaperones. One could assume that DnaK and target protein expression compete for cellular ressources and that for already soluble proteins this has no positive trade‐off. Another reason could be DnaK's role as central organizer in the chaperone network of E. coli. Binding to DnaK could trigger degradation of the “toxic” target protein but also overexpression of DnaK additionally to the endogenous version could bring the cellular metabolism out of balance. ,

We cannot verify that changes with co‐expression of chaperones is solely due to effects of chaperones on putative de novo proteins or on overall amount of protein expression. Our main interest here is to optimize expression for follow‐up experiments and not to draw general conclusions on chaperone interaction with de novo proteins.

Drawing conclusions from heterologous expression experiments toward in vivo interactions of proteins and chaperone systems are fragmentary and can only serve as hypotheses in need of further verification using in vivo experiments.

Comparing putative de novo proteins from D. melanogaster to H. sapiens

In total, we were able to successfully express 13 out of 21 putative de novo proteins in E. coli cells (eight in D. melanogaster and five in H. sapiens). For both, D. melanogaster and H. sapiens candidate putative de novo proteins, the combination of GST‐tag and E. coli T7 Express cells were the best performing (10 out of 13). We performed test expressions and compared the levels of soluble expression for different chaperone combinations shown in Figures 4 and 5. Expression results from putative de novo protein candidates DM5, Atlas, HS2, and HS9 were in line with our original hypothesis that chaperones enhance solubility of de novo proteins in heterologous expression systems. However, the choice of appropriate tag and expression cells in the first step was equally, if not more, important. When using the N‐terminal His‐tag that proved successful for putative de novo protein Gdrd, only two (Atlas and HS7) of our candidate proteins were expressed. When switching to the N‐terminal GST‐tag another seven D. melanogaster and four more H. sapiens protein candidates were expressed. Unfortunately, we were not able to express 8/21 of the candidate proteins in E. coli at all (HS1, HS3 – HS6, DM6, DM9, and DM10), despite trying different expression strains, tags and chaperone systems.

Disorder and secondary structure predictions

When examining the predicted structural properties of the human de novo protein candidates, we observe a slight trend toward better expression of the more disordered proteins. This trend can be observed for the IUPred2a disorder predictions (Figure 2) but becomes more apparent for the overall secondary structure predictions (Figure 3). The unsuccessful expression candidates HS1, HS3, and HS4 showed a higher predicted α‐helical content of approximately 40% while HS5 and HS6 had a higher predicted β‐sheet content of around 30%–40% compared with the other human candidate proteins HS2, HS7, HS8, HS9, and HS10) which are predicted to contain over 70% random coils (or 60% disorder). The described differences in predicted secondary structure content and disorder level might be the reason why these putative de novo candidates could not be expressed in E. coli cells even with the help of chaperones.

For D. melanogaster protein candidates, this trend was not observed. Here, several of the proteins with lower disorder predicted (DM1, DM4, and DM7) were expressed solubly without addition of chaperones. Yet, DM6 (∼90% disorder predicted) was not expressed successfully. However, the two proteins with 100% random coils predicted by Porter 5.0 and highest disorder predictions by IUPred2a (DM8 and HS10) did not show any change in solubility when chaperones were co‐expressed. Considering that such highly disordered proteins do not need chaperones, this observation was expected.

Deviations of the level of predicted disorder and predicted secondary structures, especially random coils, for each protein can be explained by the differences in IUPred2a and Porter 5.0. IUPred2a provides energy estimations for each amino acid residue resulting in quasi‐probabilities of disorder. On the other hand, Porter 5.0 is based on a neural network relying on sequence alignments and co‐evolutionary information. These fundamentally different approaches can lead to inconsistent results in some cases (e.g., HS9 and DM3) while not invalidating one another.

CONCLUSION AND OUTLOOK

Exemplifying the general trend for soluble de novo protein expression is only the first step toward enabling further in vitro experiments for functional and structural characterization. Further advancement will lead to efficient and stable purification, followed by functional assays such as peptide phage display to identify binding partners. , This technique has proven to be useful for high‐throughput screening of intrinsically disordered regions for short linear motifs, especially for human proteins. Soluble expression and purification will be crucial for structural characterization via CD, NMR, and Cryo‐EM. Due to their small size and high disorder content, only NMR and potentially Cryo‐EM will be capable of solving the structure of de novo proteins experimentally. Even in light of the recent dawn of computational structure prediction, , experimental structural and functional determination remains necessary, especially for de novo proteins. While contemporary prediction methods can certainly provide a first estimate on structure, the intrinsic nature of de novo proteins, with their short length, high disorder content and lack of homology, will demand some scepticism while analyzing such predictions. , Surprisingly, the AlphaFold prediction of Goddard (GEO12017p1, AlphaFold Protein Structure Database) is in line with its partial experimental characterization and its central helix is predicted with high confidence. Despite the lack of homology, which is a core demand for the MSA generation of AlphaFold2 and a hallmark of de novo proteins, one could assume that de novo proteins are of such small size that AlphaFold can solve their local folding. This will have to be validated in future studies. This study of 21 putative de novo proteins from H. sapiens and D. melanogaster, including previously in vivo characterized putative de novo protein Atlas, showed that chaperones may help expressing de novo proteins in E. coli cells. However, not all putative de novo proteins needed chaperones for soluble expression and sometimes even expressed better without. Fusion of the target de novo proteins to a GST‐tag and using T7 Express cells as hosts proved to be the most successful combination. Our work may serve as a guide to facilitating future analyses of putative de novo proteins or other difficult (short and/or disordered) target proteins in E. coli.

MATERIALS AND METHODS

Online data availability

All SDS‐PAGEs, MS results, Western blots, and scripts are deposited in Zenodo database (https://doi.org/10.5281/zenodo.6512224).

Computational methods

Candidate selection

We selected a total of 21 putative de novo protein candidates. Ten are uncharacterized putative de novo proteins from Homo sapiens and are referred to here as HS1‐10. Ten proteins originate from Drosophila melanogaster and are referred to as DM1‐10. One is the functionally characterized putative de novo protein Atlas from D. melanogaster. The 21 candidates contain different levels of disorder and secondary structure elements (α‐helix, β‐sheet, and mixture of both) and different sequence lengths (see Figure 2). We selected only candidate sequences without exon/intron structure and without long single amino acid repeats. All putative de novo proteins have confirmed expression in their native organism.

Predictions

We performed disorder predictions with IUPred2a , using default options long disorder for entire proteins. We calculated the average disorder score of the whole sequence and percentage of residues predicted to be disordered. The percentage of disorder was calculated by taking the amount of disordered residues (disorder score > 0.5) and dividing it by the sequence length of the protein. We also predicted average disorder and percentage of disordered residues with a disorder threshold of 0.8 (Figure S1). A python script was used to automate predictions and disorder proportion for all candidates. We performed α‐helix and β‐sheet predictions to verify the amount of disordered residues predicted by IUPred2a. Secondary structure predictions were performed with Porter 5.0 (SS3). , The predicted secondary structure elements for each residue were counted with a Javascript and divided by the total number of residues to obtain a percentage score for each structural element. DnaK binding sites were predicted using the ChaperISM suite (v1) in quantitative mode with default settings.

Experimental methods

Cloning of putative de novo candidates

Putative de novo candidates were synthesized as strings DNA from Twist Bioscience, San Francisco, codon optimzed for E. coli and without restriction sites used for cloning (BamHI, HindIII, NcoI, XhoI) inside the sequence. The wild‐type DNA for Atlas was provided by Geoff Findlay. To introduce restriction sites at the ends, we used different primers (a fasta file containing the DNA sequences and primer used can be found online on Zenodo (https://doi.org/10.5281/zenodo.6512224) For cloning into pHAT2 vector (N‐terminal 6xHis) we used restriction enzymes combination of BamHI/XhoI + HindIII, for pETM‐30 (N‐terminal 6xHis‐GST‐TEV), we used NcoI+HindIII. Both vectors were from the EMBL vector database, Heidelberg, introduced stop‐codon was TAA for all constructs. We digested the PCR product with both restriction enzymes respectively (FastDigest, Thermo Scientific) for 3 h at 37°C. Digest of the vector (1 h, 37°C) was purified from agarose gel (Zymo Research). We ligated both with an insert:vector ratio of 1:4 using Ligase (Thermo Scientific; 1 h, 22°C). The ligation mix was purified (Zymo Research) and 2 μl of the purified reaction mix was used to transform into 50 μl of chemically competent E. coli TOP10 cells. Cells were incubated for 30 min on ice, followed by a 90 sec heat‐shock at 42°C. 500 μl of LB‐Media (5 g yeast extract, 6 g tryptone, 5 g NaCl) was added for recovery and incubated for 1 h at 37°C. After incubation, the resuspended cell pellets were plated on LB‐agar containing 50 μg/ml ampicillin (AMP, Carl Roth, pHAT2, and EMBL vector database) or Kanamycin (KAN, Carl Roth, pETM‐30, and EMBL vector database) and incubated at 37°C over night.

Successful transformation was verified by colony PCR and sequencing at Microsynth, Seqlab, Germany. The plasmid DNA bearing the chaperone combinations GroEL/ES (pGro7) or DnaK/J/GrpE (pKJE) from Takara Biotech chaperone kit , were first transformed into E. coli Top10 cells and then into expression strains (BL21 Star™ (DE3) and T7 Express). Chaperone plasmid bearing cells were made chemically competent (Inoue method) , and used for transformation with the plasmid containing the target protein sequence. Final expression cells contained two plasmids: chaperone plasmid and target protein plasmid. The chaperone plasmids are chloramphenicol (CAM) resistant, so the double plasmid cells are either AMP + CAM (pHAT2, N‐terminal 6xHistag) or KAN + CAM (pETM‐30, N‐terminal GST‐tag) resistant.

Test‐Expression of candidate de novo proteins

To identify in which strain and plasmid proteins were expressed we performed test expressions. 10 ml of LB + AMP + CAM or LB + KAN + CAM were inoculated from a glycerol stock of all three expression cells bearing both plasmids (target protein and chaperone) and grown until turbid (6–8 h, 37°C). We split the solutions into 3 × 3 ml and incubated for 30 min at different temperatures (37°C, 28°C, and 20°C) before adding IPTG (Carl Roth) for a final concentration of 0.5 mM and shaking over night. When using the cells with chaperone plasmids we made the following adjustment: L‐arabinose (final concentration 3 mM, Carl Roth) was added from the beginning for immediate induction of chaperone expression. Therefore, after inducing the de novo protein expression with IPTG the chaperones were already present in order to help folding the de novo proteins.

A total of 500 μL of each cell culture were centrifuged (15,000 rpm, 2 min). Pellets were resuspended and lyzed in 50 μl of a mix of Bugbuster and Lysonase (both Merck AG) through vortexing for 10 min. After centrifugation the supernatant was mixed with the same volume of SDS‐loading buffer (standard). The pellet was resuspended in 5x diluted Bugbuster, centrifuged, and resuspended in 50 μl SDS‐loading buffer. 15 μl of each fraction was loaded on an SDS‐PAGE, either 10% Bis‐Tris or 12.5% TGS, run on 200 V for 50 min and dyed using ReadyBlue™ staining.

For the final Western blots, the determined optimal combination of strain, expression vector, and chaperone plasmid were used. 20 ml cultures of 2YT + AMP + CAM or 2YT + KAN + CAM were inoculated with 1 ml of the overnight culture. L‐arabinose (final concentration 3 mM) was added to the samples, but not to the control without chaperones and grown at 37°C, 180 rpm for 4–6 hr until turbid. The cultures were incubated at 28°C, 180 rpm for 30 min before induction with IPTG (final concentration 0.5 mM) and incubated overnight under these conditions. Final samples were harvested and handled as prior performed test expressions.

Western blot

The SDS‐PAGEs were run as described above but without ReadyBlue™staining. The gel was equilibrated in transfer buffer (20% Methanol) for a few seconds. A polyvinylidene fluoride (PVDF) membrane with a pore size of 0.22 μm was activated by methanol (2 min) and equilibrated in transfer buffer. The semi‐dry transfer was performed at 25 V for 30 min using the BioRad standard protocol. The membrane was blocked at room temperature for 1 hr using 5% bovine serum albumin BSA in phosphate‐buffered saline with tween (PBS‐T) then washed in PBS‐T and incubated for 1 h with anti‐His antibody (MA1‐21315‐HRP) diluted 1:500. For chemiluminescence, 0.5 ml luminol was mixed with 0.5 ml peroxide and distributed evenly on the membrane. Intensities of the different bands were measured in ImageJ after default background subtraction. The different fractions (soluble/insoluble of base, GroEL and DnaK) are calculated as fractions of the intensities of all relevant protein bands to also compare the amount of expression between the different conditions.

Mass spectrometry

Tryptic digest followed by mass spectrometry for peptide detection of the candidate proteins was performed by the Core Unit Proteomics group of Prof. Dr. Simone König, UKM Muenster.

AUTHOR CONTRIBUTIONS

Lars A. Eicholt: Data curation (lead); investigation (lead); validation (lead); visualization (lead); writing – original draft (lead). Margaux Aubel: Data curation (equal); investigation (equal); validation (equal); visualization (equal); writing – original draft (equal). Katrin Berk: Data curation (supporting); validation (supporting); writing – original draft (supporting). Erich Bornberg‐Bauer: Conceptualization (supporting); funding acquisition (lead); project administration (supporting); resources (lead); supervision (supporting); writing – original draft (supporting). Andreas Lange: Conceptualization (lead); investigation (supporting); project administration (lead); supervision (lead); validation (supporting); writing – original draft (supporting).

CONFLICT OF INTERESTS

The authors declare no competing interests.

Figure S1 Percentage of average disorder as calculated with IUPred2a. All candidate de novo proteins used for expression experiments ordered by their average disorder level from left to right.

Click here for additional data file.

Figure S2 Number of DnaK binding Heptamers as predicted with ChaperISM suite (v1) per candidate de novo protein. DM3 was expressed solubly with DnaK, but no binding sites were predicted.

Click here for additional data file.

Figure S3 SDS‐PAGEs with all D. melanogaser de novo protein candidates co‐expressed with GroEL system and without. The red boxes indicate GroEL chaperone, green boxes the target protein.

Click here for additional data file.

Figure S4 SDS‐PAGEs with solubly expressed de novo protein candidates of H. sapiens co‐expressed with GroEL system and without. The red boxes indicate GroEL chaperone, green boxes the target protein.

Click here for additional data file.

Table S1 Names of the putative de novo proteins, annotated gene names (only for D. melanogaster) and genomic locations on genome assembly BDGP6.32 (D. melanogaster) and GRCh38.p13 (H. sapiens).

Click here for additional data file.