The Effect of Divalent Cations on the Thermostability of Type II Polyketide Synthase Acyl Carrier Proteins

Abstract

The successful engineering of biosynthetic pathways hinges on understanding the factors that influence acyl carrier protein (ACP) stability and function. The stability and structure of ACPs can be influenced by the presence of divalent cations, but how this relates to primary sequence remains poorly understood. As part of a course-based undergraduate research experience, we investigated the thermostability of type II polyketide synthase (PKS) ACPs. We observed an approximate 40 °C range in the thermostability amongst the 14 ACPs studied, as well as an increase in stability (5 – 26 °C) of the ACPs in the presence of divalent cations. Distribution of charges in the helix II-loop-helix III region was found to impact the enthalpy of denaturation. Taken together, our results reveal clues as to how the sequence of type II PKS ACPs relates to their structural stability, information that can be used to study how ACP sequence relates to function.

Introduction

Microorganisms produce pharmacologically-relevant natural products of enormous structural complexity.¹ In particular, type II polyketide natural products are a prolific source of antibiotics (e.g., tetracycline) and anticancer agents (e.g., the anthracycline skeleton of doxorubicin).¹ These molecules are assembled within microorganisms by type II polyketide synthases (PKSs), large enzyme assemblies encoded by biosynthetic gene clusters. The engineering of existing type II PKSs represents a powerful route to gain access to chemical diversity relevant to human health, but the success of this approach hinges on understanding the molecular features that guide protein-protein interactions within these synthases.²

Acyl carrier proteins (ACPs) serve as central hubs in PKSs, carrying reactive intermediates to partner proteins within the synthase, while protecting their molecular cargo from undesirable reactions. ACPs are highly conserved, small (approximately 9 kDa), and acidic proteins with a characteristic four-helix bundle structure comprised of three major helices (H_I, H_II, and H_IV), a minor helix (H_III), a hydrophobic core, and a flexible loop connecting H_I and H_II (Figure 1).³ They are expressed in an inactive form, known as the apo form. A conserved serine residue residing at the N-terminus of H_II is post-translationally modified with a 4′-phosphopantetheine (Ppant) moiety, converting the protein to its holo form (Figure S1).⁴ The Ppant moiety acts as a swinging arm through which the terminal thiol tethers intermediates that are shuttled as thioesters. These substrates remain reactive enough to support the transacylation reactions and decarboxylative Claisen-like condensations that build the polyketide chain. To protect the acyl intermediates from nonselective reactivity, an ACP can sequester its molecular cargo within its hydrophobic cavity.⁵ Upon interactions with a partner enzyme, the acyl chain can flip from the hydrophobic pocket of the ACP to that of the enzymatic counterpart in a solvent-protected environment. This overall mechanism, which appears to be universal among type II synthases, is known as the chain-flipping process.^5–7

Figure 1. Comparison of primary sequences and structures of ACPs studied.

A. Multiple sequence alignment of helix II and III of ACPs highlight the variability in charge distribution in this region (acidic amino acids highlighted in red and basic amino acids in cyan). Ec-ACP cation binding sites are marked with green stars. B. Overlay of ACP solved structures and homology models reveal similar global folds amongst the ACPs studied, with the most variability observed in the region between helix III and IV (see Figure S3 and Table S2 for details).

The flexible nature of the ACP structure is thought to play an important role in recruiting and interacting with enzymatic partners.⁸ Interestingly, ACPs have been referred to as intrinsically unstructured by virtue of the disorder conferred by the electrostatic repulsive forces that result from their extraordinary acidic and polar character.⁸ ACP conformations seem to be influenced by environmental factors, such as the presence or absence of divalent cations, enzymatic partners, or acylation of the Ppant arm.^9,10 A conserved, negatively-charged patch of residues along H_II of the ACP has been identified as a “recognition helix” for binding to partner enzymes through electrostatic complementarity.¹¹ More recently, the role of hydrophobic interactions in ACP-protein recognition has also been highlighted.¹² Taken together, these observations suggest that multiple factors converge to confer ACPs with such elegantly-tuned conformational dynamics.

Despite their flexible nature, ACPs from across different species and biosynthetic pathways display strikingly similar global folds (e.g., fatty acid synthase (FAS) ACPs Helicobacter pylori, Bacillus subtilis, rat, Thermatoga maritima, and PKS ACPs act, fren, otc).¹³ Interestingly, the sequence identity and charge distribution between these ACP homologs vary widely.¹³ In addition, despite maintaining a similar global fold, some FAS ACPs and PKS ACPs display deviating helix and loop orientations with varying helix lengths.¹³

Perhaps as a consequence of these sequence and structural deviations, ACPs exhibit a wide range of stability and conformations across species and conditions (such as pH, temperature, or divalent cation concentration).^10,14–18 For example, the ACP in Helicobacter pylori (Hp-ACP) shows a distinct pH-dependent conformational change, such that it is partially unfolded at neutral pH, but tightly folded at pH 6.¹⁸ The thermostability of Hp-ACP also changes significantly depending on whether it is in its apo or holo state.¹⁶ In the presence of divalent cations, the ACP from Enterococcus faecalis (Ef-ACP), a thermophilic bacterium found in the gastrointestinal tracts of humans, has a much higher thermostability than that of the FAS E. coli ACP (Ec-ACP; 79 °C compared to 67 °C).¹⁷ Vibrio harveyi ACP (Vh-ACP) is unfolded at pH 7 despite sharing 86% sequence identity with Ec-ACP, but Vh-ACP folding can be induced by binding to FAS enzymes or divalent cations, or attachment of acyl chain groups.¹⁹

The effect of divalent cation binding has been well-studied in Ec-ACP. Divalent cations have previously been shown to stabilize native ACP conformation by binding to two sites of acidic residues located in the HII and surrounding loop regions: site A (E30, D35, D38), and site B (E47, D51, E53, D56; see Figure 1A).²⁰ Differential scanning calorimetry (DSC) has been used to examine the effects of both monovalent and divalent cation binding on the thermodynamics of unfolding.¹⁴ Results showed that although both monovalent and divalent salts increased thermostability, divalent cation binding was coupled with a large increase in the change in enthalpy of unfolding (ΔH_cal,), suggesting specific binding to the native state. Other studies showed that mutations to these cation binding residues in Vh-ACP appeared to affect interactions with ACP-dependent enzymes.²¹

The explosive rate of biosynthetic gene cluster sequencing motivated our efforts to connect ACP primary sequence to properties such as stability and cation binding. Since ACPs serve as a linchpin in PKSs, such predictive power could be leveraged to build functional hybrid PKSs and therefore gain access to new chemical diversity. Towards this goal, we cloned, and expressed 13 type II PKS ACPs from Streptomyces (Figure 1) that were inferred to be evolutionarily-diverse based on previous cluster-wide evolutionary studies and pairwise gene coevolution information.²² Since type II PKSs share common ancestry with E. coli FASs,²² we studied the E. coli FAS ACP, Ec-ACP, in parallel. Inspired by previous studies that suggest Ec-ACP exhibits higher thermostability in the presence of Ca^2+, and intrigued by the notion that cation-induced conformational changes could play an important role in directing polyketide biosynthesis, we also investigated how the presence of divalent cations influences type II PKS ACP thermostability. By studying 14 distinct ACPs, we were able to reveal clues as to how the sequence of type II ACPs relates to their structural stability, information which can be further built upon in the future to predict features of ACP function based on primarily sequence.

We integrated this research challenge into a course-based undergraduate research experience (CURE)^23–25, thereby leveraging a “many hands make light work” approach to studying multiple ACPs in parallel, while concomitantly fulfilling the pedagogical goals of an upper-level biochemistry course. In doing so, a team of undergraduate students were exposed to original research challenges at the chemistry-biology interface, thereby expanding the impact of our work into training future scientists and scientifically literate citizens.

Materials and Methods

General.

All reagents, unless otherwise stated, were purchased from Sigma Aldrich. Nickel-NTA agarose resin and isopropyl β-D-1-thiogalactopyranoside (IPTG) were purchased from Gold Biotechnology. Sodium dodecyl sulfate (SDS), acrylamide, and tetramethylethylenediamine (TEMED) were purchased from Bio-Rad and Luria-Broth powder was purchased from IBI Scientific. During the protein expression process, cells and cell lysate were pelleted using an Avanti J-E centrifuge (Beckman Coulter). Proteins were concentrated using an Allegra X-14R centrifuge with three types of 3-kDa molecular weight cutoff (MWCO) centrifugal units: Amicon Ultra (Millipore, 15 mL capacity), Vivaspin (Sartorius Stedim Biotech, 2 mL capacity), and Vivaspin (Sartorius Stedim Biotech, 0.5 mL capacity). The Ec-ACP plasmid pTL14²⁶ (N- and C-terminal His₆ tags; kanamycin resistant) was provided by the Khosla Lab at Stanford University. The Ec-ACP plasmid pKJ5535 (N-terminal His₆ tag only; kanamycin resistant) was constructed as described in Supporting Information. The ACP_act plasmid (pMC002067; carbenicillin resistant) was provided by the Chang Lab at University of California Berkeley.

Molecular cloning of ACP Expression Constructs.

ACP DNA sequences flanked with NdeI/EcoRI or NdeI/BamHI restriction sites were ordered as gBlock DNA fragments from Integrated DNA Technologies (IDT) for insertion into a pET28a vector for expression of the encoded ACPs with an N-terminal His₆-tag. The gBlock DNA was digested with the appropriate restriction enzymes, gel-purified, and cloned into the corresponding sites of pre-digested pET28a vector using the Roche Rapid DNA Ligation kit. Mutagenic PCR was used to obtain Ec-ACP mutant plasmids. See Supporting Information for details.

Expression of ACPs.

ACPs were expressed and purified as described previously.²⁷ In brief, ACP expression plasmids were transformed into chemically competent BAP1 cells for expression.²⁸ Seed cultures (10 mL LB, 50 μg kanamycin/mL or 100 μg/mL carbenicillin) were grown at 37 °C with shaking and then added to 1 L LB production cultures (50 μg kanamycin/mL or 100 μg/mL carbenicillin). Production cultures were grown at 37 °C with shaking until OD₆₀₀ was between 0.4 – 0.6. They were then induced with 250 μL of 1 M IPTG and incubated at 18 °C for 18 – 21 h with shaking. Cells were harvested by centrifugation (4500 × g, 4 °C, 15 min), resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 2 mM dithiothreitol (DTT), 10% glycerol), and sonicated while on ice using an XL-200 Microson sonicator. Supernatant was clarified by centrifugation (17000 × g, 45 min), and the supernatant was incubated overnight at 4 °C with 3 mL nickel-NTA agarose slurry equilibrated in lysis buffer. After collecting the flow through, resin was washed with 20 – 25 mL of Wash 1 buffer (50 mM Tris-HCl, pH 7.5, 2 mM DTT, 10 mM imidazole), 20 – 25 mL of Wash 2 buffer (50 mM Tris-HCl, pH 7.5, 2 mM DTT, 30 mM imidazole), and protein was eluted from the resin with 10 – 15 mL of elution buffer (50 mM Tris-HCl, pH 7.5, 2 mM DTT, 300 mM imidazole). ACPs were concentrated using 3 kDa MWCO centricons (4000 × g, 4 °C) to 0.8 – 1.0 mM, flash-frozen in liquid nitrogen, and stored in 10% glycerol at −80 °C.

Conversion to holo-ACP.

Coenzyme A (2.5 mM) was added to a solution of ACP (0.8 – 1.0 mM) with Sfp R4–4²⁹ (1.5 μM) in 50 mM sodium phosphate, 10 mM MgCl₂, pH 7.6, for a total volume of 1 mL. The solution was incubated for 12 hrs at room temperature. The reaction mixture was desalted into 50 mM phosphate buffer, pH 7.6 using a HiPrep 26/10 column (GE). Percent conversion to holo-ACP was determined using liquid chromatography mass spectrometry (LCMS).

LCMS Analysis of ACPs.

ACP (20 μL of a 0.1 mg/mL solution in 50 mM sodium phosphate buffer, pH 7.6) was analyzed by LCMS (Agilent G6125BW) equipped with a Waters XBridge Protein BEH C4 Column (300A, 3.5 μm, 2.1 mm × 50 mm) heated to 45 °C for analysis by ESI MS in the positive mode. The following gradient was used (Solvent A = water + 0.1% formic acid; Solvent B = acetonitrile + 0.1% formic acid): 0–1 min 95% A, 3.1 min 5% A, 4.52 min 5% A, 4.92–9 min 95% A. Data were deconvoluted using ESIprot³⁰ and the observed MW was compared to the calculated MW for apo and holo ACP.

Sedimentation Velocity Experiments.

All experiments were performed using a Beckman model Optima XL-A Analytical Ultracentrifuge (AUC) equipped with an An-60 Ti rotor. Sedimentation velocity runs used two-channel Epon, charcoal-filled centerpieces with 1.2 cm path length containing 400 μL sample and 410 μL buffer as a reference. Sedimentation boundaries were measured at a speed of 50,000 rpm for the studied ACPs. All measurements were performed at 20 °C using a step size of 0.003 cm, a time delay of 0 s, and a total of 70 –100 scans. Samples were monitored at 280 nm with a required starting absorbance between 0.3 and 1.0.

Temperature-corrected partial specific volumes (as weight-averages), densities, and viscosities were calculated using Sednterp (beta version).³¹ Heterogeneity of the mixtures was determined by using the dc/dt method³² implemented in DCDT+ (v.2.4.0), and a model-independent, continuous c(s) distribution using the 1 discrete component model in Sedfit (v.15.3).³³ The 1 discrete component model takes into consideration small molar mass components that do not sediment even at 50,000 rpm.³³ Confidence intervals for temperature-corrected sedimentation coefficient, s(20,w), diffusion coefficient, D(20,w), and molecular mass (kDa) were computed using a bootstrapping method, confidence probability level 90% (±1.65 sigma) in DCDT+ (v.2.4.0).³² Data, fits, and residual distributions were plotted using the Gussi interface (v. 1.3.2)³⁴ implemented in Sedfit (v.15.3).

Circular Dichroism Measurements of ACPs.

Circular dichroism (CD) measurements were performed on ACPs (10–50 μM, in 10 mM Tris-HCl, pH 7.6) using a cell with a 1 mm pathlength (Hellma Analytics). To study the effects of divalent cations, CaCl₂ (10 mM, final concentration) was added immediately prior to transferring sample into cuvette. Wavelength scans were collected at the 260 – 190 nm range on an Aviv model 410 spectropolarimeter. Changes in signal at 222 nm were followed as a function of temperature using the following parameters: ~ 10 – 100 °C (actual range depended on stability of ACP), 2 °C steps, 2-minute equilibration, heating rate of 2 °C min⁻¹, 30 second signal averaging time, and a 1 nm bandwidth. Data were analyzed via CDpal³⁵ using a two-state unfolding model N⇌U with the standard assumption that ΔC_p = 0. We evaluated the robustness of our assumption that ΔC_p = 0 by: i) comparing fits when ΔC_p was set as zero versus when it was allowed to vary as a fitting parameter; and ii) using the approximation that ΔC_p = 12 (cal/mol/deg/res) × 111 residues = 1.33 kcal/K mol.³⁵ We found little variability in parameters obtained using these different approaches and therefore employed a value of ΔC_p = 0 for data analysis. Estimated errors for each fitted parameter were calculated using a Jackknifing method implemented in CDpal (see Supporting Information for details). The difference in the value for enthalpy of unfolding in the presence and absence of CaCl₂ was calculated for each ACP (ΔΔH=ΔH_+CaCl2-ΔH_{no salt}).

Calculation of Isoelectric Points and Aliphatic Index.

Analysis of isoelectric point and aliphatic index was conducted using the ExPASy ProtParam tool³⁶ with the corresponding ACP sequences.

Generation of Homology Models.

The primary sequences of the type II polyketide ACPs were used to generate homology models via SWISS-MODEL.³⁷ The models were viewed and imaged using Pymol 2.0. RMSD values were obtained by aligning the generated model with the template. QMEAN scores above −4 indicated that the model was of overall good quality. See Supporting Information for details.

Implementation as a Course-Based Undergraduate Research Experience (CURE).

Initial experiments were conducted in the context of a course-based undergraduate research experience (CURE)^23–25 by 16 junior undergraduate students enrolled in the 2016 Laboratory in Biochemical Research at Haverford College. Through “literature club”-style lectures, students gained an understanding of the basic role of ACPs in type II polyketide biosynthesis. As a class, we identified a key unanswered question amenable to a “many hands make light work” approach: How does ACP sequence relate to ACP structure, stability, and function? We then identified 13 ACPs (representing various stages of evolution based on whole-cluster phylogenetic analyses²²) to clone, express, purify, and characterize as a team. Throughout the 14-week course, students designed and executed experiments, encountered unexpected findings, shared results in “group meeting”-style lectures, and networked with invited guest speakers with expertise relevant to the project. Two students (V. Courouble and M. Rivas) replicated the findings from the course and conducted additional experiments as part of their senior thesis research projects.

Results

Sequence and structure homology of ACPs.

A multiple sequence alignment (MSA) of the 13 PKS ACPs and Ec-ACP was generated in Clustal Omega³⁸ and visualized using Jalview³⁹ (Figure 1, Figure S2). At the N-terminus of H_II, all ACPs contain a serine, functioning as the point of attachment for the Ppant arm, and a highly conserved flanking region. Additional conserved motifs include: i) a threonine at the N-terminus of H_IV; ii) a hydrophobic patch in H_IV; iii) a negative charge patch connecting the unstructured region between H_I and H_II; and iv) a negative charge patch in H III. The abundance of acidic residues within the ACPs results in calculated net charges at physiological pH ranging from −6.3 to −16.2 (See Table S1). The charge distribution of acidic and basic residues within H_II and H_III varied between the studied ACPs. Though Ec-ACP has 14 acidic and zero basic residues between positions 30 and 60 (containing H_II and H_III), the type II PKS ACPs have a range of 5–9 acidic residues and 1–3 basic residues (Figure 1). None of the type II PKS ACPs retain all the cation binding residues present in Ec-ACP: D35 and D56 are well-conserved, whereas D38 and D51 are not conserved for any of the type II ACPs studied. The histidine at position 75 of Ec-ACP, which was previously identified as a key residue for stabilizing ACP conformation, is not conserved in the studied type II ACPs (Figure S2). All ACPs studied were predicted to exist in the canonical helical bundle structure based on homology models (Figure 1, Figure S3, Table S2), although several models suggest the possibility of extended loop regions (e.g. ACP_tcm, ACP_whiE).

Expression and Purification of holo-ACPs.

All ACPs were expressed as soluble proteins and purified by Ni-affinity purification (see SDS PAGE gel in Figure S4). It has been previously established that the loaded state of ACPs (apo vs holo vs acyl) can affect their thermostability.¹⁶ Whereas the BAP1 cell line used for protein expression harbors a phosphopantetheinyl transferase capable of converting apo ACP to holo ACP, not all ACPs are entirely converted to the holo state.²⁸ Therefore, we evaluated the ratio of apo:holo ACP in our purified protein samples via LCMS. ACPs that were not fully holo after expression in BAP1 cells were treated with the promiscuous phosphopantetheinyl transferase (PPTase) Sfp R4–4²⁹ in vitro and re-analyzed via LCMS (Table S3, S4). Following the in vitro Sfp reaction, four remained fully or mostly in the apo form, three were a 50:50 ratio of holo:apo, and the remainder were found to be exclusively in the holo state (Figure S5). We limit our discussion of ACP thermostability in the main text to those ACPs that were exclusively, or mostly, in the holo state.

Oligomeric state of ACPs.

A critical assumption in inferring thermodynamic parameters from thermal denaturation data is that the ACPs exhibit two-state denaturation and exist primarily as monomers in solution. To test the robustness of this assumption, we analyzed select ACPs (those for which we could obtain protein in an entirely, or a majority, holo state) using sedimentation velocity experiments with an analytical ultracentrifuge (SV-AUC). We observed that the sedimentation profiles of ACPs fit to a model-free continuous c(s) distribution and a major peak at s(20,w) = 1.4 S for most of the ACPs (Figure S6). Ec-ACP expressed from pTL14 (12.3 kDa) sediments slightly differently from the type II PKS ACPs, with its major peak being more broad (9.97 – 12.8 kDa) and centered at a higher s-value s(20,w) = 1.7. This observation suggests that Ec-ACP sediments at an apparently “larger” molecular radius, which emulates how the protein behaves in SDS-PAGE (Figure S4).⁴⁰ The SV-AUC results confirm that the ACPs studied exist primarily in a monomeric state (Figure S6), allowing us to use a two-state assumption for ACP unfolding. We observed that Ec-ACP forms a tetramer at high (>200 μM) concentrations, but whether this self-association is physiologically relevant remains to be investigated.

Thermostability of Ec-ACP.

CD was used to examine the effects of divalent cation binding on the secondary structure of ACP and thermostability. In the past, both CD and DSC have been used to study Ec-ACP.^9,14 In these studies, unfolding appeared to be a cooperative two state process without observable intermediates, with an increase in the midpoint of temperature-induced unfolding (T_m) and ΔH_cal upon addition of CaCl₂.¹⁴ We repeated these experiments to verify the reversibility of denaturation by taking wavelength scans of Ec-ACP before and after thermal denaturation (Figure 2). Protein unfolding appears to be mostly reversible, with an 85–90% recovery in signal following denaturation. We observed a similar trend upon addition of CaCl₂, with an increase in T_m (ΔT_m=13.8 °C) and an increase in ΔH of unfolding (ΔΔH=30 kcal/mol) (Figure 2). Experiments were conducted on two variants of Ec-ACP, with two His tags or one His tag, though results were within error of each other (see Supporting Information for details).

Figure 2. Holo-Ec-ACP thermal unfolding is fully reversible in absence and presence of CaCl₂.

A. T_m values were obtained upon fitting the fractional change of unfolded Ec-ACP as a function of temperature to a N↔D model with ΔCp = 0 in the absence of CaCl₂ (black squares) and in the presence of 10 mM CaCl₂ (blue circles). Best fits to the data are presented as solid lines. B. The full CD spectra of Ec-ACP pre-T_melt (solid lines) and post-T_melt (dashed lines) reveal that the CD signal is fully recovered and therefore unfolding is considered reversible.

Thermostability of Type II PKS ACPs.

Wavelength scans of the ACPs prior to thermal denaturation show that all ACPs display strong negative absorbances at 222 nm and 208 nm, which confirms the ACPs are folded and primarily alpha helical. Addition of CaCl₂ to the ACPs did not cause a significant difference in secondary structure (data not shown). Thus, binding of CaCl₂ does not seem to cause a conformational change detectable by CD.

The type II PKS ACPs display a remarkable range in T_m, with the least stable ACP (ACP_chry) unfolding at more than 40 °C lower than the most stable ACP (ACP_whiE) (Figure 3, Figure S7, Table 1, Table S5). These data highlight that despite inferred structural similarities, the physical properties of ACPs, such as thermostability, can be highly variable. For instance, ACP_act and Ec-ACP have similar folded structures as shown by NMR^41,42, yet our results indicate that their T_ms differ by almost 20 °C (in the absence of Ca²⁺). Overall sequence similarity does not seem to result in similar T_ms either, as ACP_ctc and ACP_lan are 52% identical and yet their T_m s differ by 22 °C. Using the sequences of each ACP, we calculated the aliphatic index and pI for each ACP, and plotted them against the respective T_m, and found that neither of the calculated parameters appear to be sole predictors of ACP stability (Figure S8). Taken together, these data suggest that there is a diverse range of chemical reasons for the observed differences in ACP thermostability.

Figure 3. ACPs show increase in T_m in the presence of CaCl₂.

The overlaid thermal denaturation profiles of WT ACPs and Ec-ACP mutants in absence of CaCl₂ (A and B) and presence of CaCl₂ (C and D) measured by CD (solid lines represent best fit to a two-state model with ΔC_p = 0). All ACPs were either exclusively in the holo-form or in a majority holo-form (see Figure S4 and Tables S3 and S4). A summary of thermodynamic parameters obtained from these experiments are presented in Tables S5 and S6.

Table 1.

Summary of the observed loaded states (apo versus holo) and thermostability of ACPs studied.

ACPs	Expressed as (apo/holo)	Post Sfp reaction	Tmelt (-CaCl2)	Tmelt (+CaCl2)	ΔTm	ΔΔ (H)
Ec -ACP (2H)	50/50	100 holo	59.5 ±0.2	74.3 ± 0.2	14.8	21.2
Ec-ACP (1H)	50/50	100 holo	57.5±0.3	71.3±0.3	13.8	30.6
Ec-ACP E41A	50/50	100 holo	64.3±0.4	76.0±0.2	11.7	30.1
Ec-ACP E47A	90/10	100 holo	48.8±0.4	65.8±0.2	17.0	21.6
Ec-ACP E47K	90/10	100 holo	45.7±1.9	61.9±0.4	16.2	28.1
ACPact	50/50	100 holo	39.7±1.7	65.6±1.2	25.9	6.8
ACParm	100 holo	-	47.8±0.3	56.8±0.5	9	6
ACPben	100 holo	-	50.3±0.4	61.4±0.4	11.1	9.8
ACPctc	50/50	100 holo	62.5±0.3	69.5±0.4	7.0	8.2
ACPchry	50/50	50/50	25.5±1.0	49.1±0.5	23.6	15
ACPfrn	50/50	100 holo	34.5±1.5	49.1±0.5	14.6	12
ACPlan	60/40	50/50	39.3±0.4	58.8±0.4	19.5	5.5
ACPmed	80/20	50/50	48.8±0.2	58.5±0.5	9.7	14.0
ACPwhiE	80/20	100	69.3±1.8	74.3±1.1	5	7.4

We observed a general increase in ACP stability upon addition of CaCl₂, with a wide variation in ΔT_m (Figure 1, Table 1, Tables S5 and S6). The effect of divalent cations on ΔH of unfolding for Ec-ACP prompted us to look at ΔΔH (ΔH_+CaCl2-ΔH_{no salt}) across all ACPs (Table 1). We noticed that the distribution of charged residues on the ACP appear to relate to the observed ΔΔH; an increase in overall negative charge for the H_II-loop-H_III region is associated with an increase in ΔΔH (Figure 4). ACPs with net charges below −6 are paired with a small ΔΔH of 7–10 kcal/mol, while Ec-ACP, the ACP with the largest negative charge of −11.3, also has the largest ΔΔH of 30 kcal/mol.

Figure 4. The change in enthalpy of unfolding in the absence versus presence of CaCl₂ (ΔΔH) correlates with the calculated negative charge value of helix II and III.

The ΔΔH plotted against the overall charge of helix II and III (D35-E60 based on Ec-ACP sequence numbering, see Figure 1) were fitted to a Boltzmann nonlinear equation: $y\text{=}A\text{2}\text{+}\text{(}A\text{1}-A\text{2) /(1}\text{+}{\left[\left(exp\right)\right]}^{\wedge }(((x-x0)/dx)).$ R² = 0.87.

Mutational analysis of Ec-ACP.

Comparison of ACP sequences (Figure 1) with the observed thermostability (Figure 2) led us to hypothesize that certain changes in single amino acid residues might influence ACP thermostability in predictable ways. For example, we observe that the ACPs with the highest T_ms (ACP_ctc, ACP_whiE, and ACP_fran), have an uncharged residue at position 41 (Asn, Gln, Gly, respectively) whereas other ACPs harbor acidic residues at this position. In the Ec-ACP crystal structure (PDB: 1T8K), we observe that E41 is a solvent-exposed residue positioned in the middle of H_II (Figure 5A). We postulated that slight reduction in electrostatic repulsion in H_II for ACPs with neutral residues at this position could confer these ACPs with a slightly higher thermostability. In support of this hypothesis, we observed that the E41A Ec-ACP mutant displays a T_m of 64.3, which is 5 °C higher than the WT Ec-ACP (Figure 3, Table 1).

Figure 5. Structure of apo-Ec-ACP co-crystallized with Zn²⁺ (sphere) highlights the role of E41 as a solvent-exposed residue.

Mutation of wild-type apo-ACP (PDB 1T8K; panel A) to the E41A-variant (modeled in panel B) resulted in an increase in T_m of unfolding of ~ 5 °C in the absence of CaCl₂.

Conversely, the glutamic acid at position 47 of Ec-ACP is involved with hydrogen bonding with the backbone of the loop connecting H_II and H_III (Figure 6A). We hypothesized that mutating this residue to a neutral (alanine) or basic (lysine) residue would lead to a greater change in thermostability as a result of removing the stabilizing interaction between H_II and the loop leading into H_III. While E47 is also a cation binding site, we did not anticipate a single mutation would have a notable impact on Ca²⁺ binding, as past studies have shown that mutation of all the residues in site A and site B is required to significantly influence cation binding.²¹ To test this, we made two non-conservative mutants to E47 (E47A and E47K). Wavelength scans indicated that these mutations did not affect secondary structure (Figure S9). The mutants display T_ms 8.7–11.8 °C less than WT Ec-ACP (Figure 3, Table 1), supporting the notion that disruption of an internal hydrogen bond destabilized the ACP structure, and had a greater effect than mutating the solvent-exposed residue E41. Both E47 mutants still exhibited an increase in stability upon addition of CaCl₂, paired with a large increase in ΔΔH (Table 1), consistent with previous observations that mutations to a single cation binding site do not significantly alter the stabilizing effect of divalent cations.²¹

Figure 6. Structure of apo-Ec-ACP co-crystallized with Zn²⁺ (sphere) highlights the role of E47 as a solvent-excluded residue involved in internal hydrogen bonding with I54.

Mutation of wild-type apo-ACP (PDB 1T8K; panel A) to the E47A-variant (modeled in panel B) and E47K variant (modeled in panel C) resulted in a decrease in T_m of unfolding of 8.7–11.8 °C in the absence of CaCl₂. Most other ACPs studied harbor a glutamine at this position (modeled in panel D; see Figure 1 for MSA). The models suggest that the internal hydrogen bond is conserved in the E47Q variant, but is not present in E47A. For the E47K mutant, the calculated distance between hydrogen bond donor and acceptor is too great for strong hydrogen bonding.

Implementation as a CURE.

By incorporating the study of type II PKS ACPs in the context of a CURE, the impact of our work expanded into pedagogical spaces. Students developed their skills as biochemists while working as a team on an original and relevant research problem. Throughout the 14-week course, pairs of students worked to clone, express, purify, and characterize a unique ACP. In addition to “wet lab” skills, students developed their in silico skills by conducting phylogenetic analyses, building homology models, and conducting multiple sequence alignments. Networking opportunities, as well as opportunities to develop oral and written scientific communication skills, were woven into the course, in order to provide a holistic training in biochemistry. Course evaluations revealed the effectiveness of this approach (see Supporting Information), which mirror the global success of Biochemistry CUREs.^23–25 Thus, when we think of the implications of this work for the bioengineer, we also think of training future scientists and, more generally, scientifically-literate citizens of society.

Discussion

The central role of ACPs in natural product biosynthesis coupled with the rapid increase in ACP sequence information make connecting ACP primary sequence information to structure and function attributes an important task. In an effort to contribute to this knowledge-base, we studied a suite of previously uncharacterized type II PKS ACPs as part of a biochemistry CURE. The ~40 °C range in T_ms of these ACPs is consistent with previous observations that despite considerable sequence similarities, the thermostability of ACPs is highly variable.^16,17 This phenomenon is highlighted by the comparison of Vh-ACP and Ec-ACP: even though these proteins share 86% sequence similarity, Vh-ACP is intrinsically disordered under physiological conditions, whereas Ec-ACP is folded.⁸ Interestingly, wild type Vh-ACP is unfolded at neutral pH in the absence of divalent cations, while the A75H Vh-ACP mutant is folded in the same conditions.¹⁰ We note that ACP_chry also harbors a histidine at position 75, yet this is the least stable type II PKS ACP studied in the current work. Thus, it appears that while the A75H mutation added significant stability to Vh-ACP, H75 is not a universal stabilizing feature of ACPs.

It has been reported in past literature that neutralizing electrostatic repulsion in H_II results in increased thermostability. For instance, mutating the glutamic acid residue at position 41 of Vh-ACP into a positively charged lysine (E41K) results in increased helical structure in the absence of divalent cations.⁴³ Similarly, we found that the ACPs with the highest thermostability had uncharged residues at position 41 (Figure 1, Figure 3, and Table 1), and that mutating E41 into an alanine in Ec-ACP resulted in increased stability of 5 °C. Another potentially important residue is the proline at position 55, found in the loop between H_II and H_III. Proline residues in the loop structures of proteins are believed to play an important role during protein folding.⁴⁴ ACPs without this proline residue (ACP_chry, ACP_mtm, and ACP_lan) have negatively charged residues in this position, and display low T_ms (Figure 1, Figure 3, Table 1). In addition to losing a proline that could be important for protein folding, introducing a charged residue in this region could destabilize these ACPs by changing electrostatic interactions. However, site directed mutagenesis studies would be required for any conclusive statements on the importance of this proline.

We observed that mutation of a single cation binding residue, E47, is not enough to impact ΔΔH. This is in agreement with past mutational studies of Vh-ACP, where mutation to multiple residues in each cation binding site were required to impact binding.⁸ More interestingly, mutation of this residue to either an alanine or lysine caused a significant decrease in thermostability. Upon closer inspection of the Ec-ACP NMR structure, we observe that E47 likely forms a hydrogen bond with the backbone of the loop leading into H_III (Figure 6). E47A and E47K can no longer make this interaction, resulting in an overall destabilization in the loop between H_II and H_III (Figure 6B, C). When looking at the sequences of the other type II PKS ACPs, we see that ACP_chry, the ACP with the lowest T_m, has a glycine at position 47. Conversely, all other ACPs have either a glutamic acid or a glutamine (Figure 1), and are therefore capable of forming a stabilizing intramolecular hydrogen bond with the backbone (Figure 6D). These results suggest that the identity of the amino acid at position 47 could play an important role in the stability of ACPs.

All the type II PKS ACPs studied were stabilized by the presence of Ca²⁺, though by varying degrees. As previously reported, we observed that denaturation of Ec-ACP in the presence of divalent cations resulted in a larger ΔH than without any salt present (Table 1).¹⁴ In the literature, this finding was originally coupled with the fact that although thermostability of Ec-ACP also increased in the presence of monovalent cations, ΔH of unfolding did not change compared to the no salt condition.¹⁴ These data suggested that monovalent and divalent cations stabilize ACP in different ways, such that monovalent ions act by nonspecifically balancing the negative charge of ACPs in a similar fashion regardless of whether the ACP is in the native or denatured state, while divalent ions act by binding to sites present only in the native state.¹⁴ Our data suggest that the distribution of charged residues across H_II and H_III correlate to the ΔΔH observed in the ACP unfolding in the absence versus presence of Ca²⁺ (Figure 4). ACP_chry, ACP_frn, and Ec-ACP are the ACPs with the largest negative net charge in H_II and H_III and the largest ΔΔH. Presumably, Ca²⁺ coordinates specifically to these ACPs in addition to alleviating charge repulsion, resulting in large enthalpic contributions. In contrast, it is possible that non-specific charge stabilization by presence of Ca²⁺ dominates for ACPs with lower net charges, resulting in a lower ΔΔH. Our results highlight the potential dual role of Ca²⁺ in ACP stabilization: a global neutralization effect (resulting in a small ΔΔH), and specific binding (indicated by a large ΔΔH).

It has been proposed that divalent cations could exert control over PKSs and FASs under physiological conditions, because they can act as both activators and inhibitors in a relatively narrow concentration range.⁴⁵ E. coli fatty acid biosynthesis functions most efficiently in the presence of 5 – 10 mM MgCl_2,, which has been suggested to be due to two effects: i) cation-binding to ACP can facilitate the binding of ACP to other enzymes in the synthase; and ii) the reduced repulsion between negatively charged molecular cargo and ACP can increase the reaction rate.⁴⁵ Several ACP-partner enzymes, such as AcpS (a phosphopantetheinyl transferase) and AcpH (a phosphodiesterase), also require divalent cations for activity, suggesting that local concentrations of these cations could be high.^46,47 At the same time, the binding affinity of Ec-ACP for divalent cations seems to vary depending on the loaded state of the ACP. For example, Mn(II) EPR studies suggest that two high-affinity binding sites in Ec-ACP bind much more weakly to divalent cations once the Ppant arm is acylated with a C8 chain (K_d/site = 1 mM versus 80 μM).⁴⁸

Molecular dynamics simulations have suggested that the conformational plasticity of H_III is key to its function as a gate-keeper for the chain-flipping mechanism.⁴⁹ H_III is the most dynamic part of the ACP, and constraining H_III to its natural helical conformation doubles the free-energy barrier for substrate “jack-knifing.”⁴⁹ Thus, structural stability is linked to function of the ACP and its ability to deliver a substrate. However, the relationship between the conferred stability as a result of Ca²⁺ binding and ACP functionality, such as chain-flipping, has not been directly studied. One possible connection is that when ACP is bound to Ca²⁺, H_III may be constrained in its helical conformation, in turn increasing overall stability of the ACP.

These results remind us that although charge neutralization can occur by addition of cations or removal of an acidic residue, it can also occur by intramolecular interactions or intermolecular interaction with positively-charged residues on partner enzymes. ACP-enzyme interactions are thought to be driven by electrostatic complementarity between acidic residues on H_II and H_III of the ACP and positive residues in the active site of partner enzymes.⁴⁹ The recognition H_II of Ec-ACP has been found to utilize a ubiquitin interacting motif (UIM)-like surface to bind to its partners, which provides some insight into how a single protein can interact with so many different enzymes, again emphasizing the importance of both electrostatic interactions and hydrophobic interactions.⁵⁰ The importance of charge distribution in H_II and H_III and the relationship to cation binding found in this study becomes more relevant when coupled with the previously established importance of H_II and H_III in ACP-partner interactions.

Though over a half-century has passed since ACPs were first discovered, there remain many fundamental questions about how these proteins function. Despite their small size, we know very little about how sequence relates to functionally-relevant structural characteristics. Our studies suggest a complicated relationship between ACP primary sequence and biophysical characteristics. While some negatively charged residues increase electrostatic repulsions and decrease the overall stability of ACP, other charged residues are essential for intramolecular stabilizing bonds. We observed a potential relationship between H_II and H_III charge distribution and enthalpy of unfolding, and by extension specificity of cation binding. We hope that these clues will be useful to the community as we continue to establish the “ground rules” for ACP activity carrying out their remarkable tasks within bacterial synthases.

Abbreviations

act

Actinorhodin

ACPs

acyl carrier proteins

AUC

analytical ultracentrifuge

arm

arimetamycin

ben

benastatin

CD

circular dichroism

ctc

chlorotetracycline

chry

chrysomycin

CUREs

course-based undergraduate research experience

DTT

dithiothreitol

Ef

Enterococcus faecalis

E. coli, Ec

Escherichia coli

FAS

fatty acid synthase

fran

frankiamicin

frn

frenolicin

Hp

Helicobacter pylori

IPTG

isopropyl β-D-1-thiogalactopyranoside

lan

landomycin

med

medermycin

Me

Micromonospora echinospora

mtm

mithramycin

MWCO

molecular weight cut off

PKS

polyketide synthase

pdm

pradimicin

SV

sedimentation velocity

SDS

sodium dodecyl sulfate

tcm

tetracenomycin

TMED

tetramethylethylenediamine

Vh

Vibrio harveyi