Development of a conditional localization approach to control apicoplast protein trafficking in malaria parasites

Abstract

Secretory proteins are of particular importance to apicomplexan parasites and comprise over 15% of the genomes of the human pathogens that cause diseases like malaria, toxoplasmosis and babesiosis as well as other diseases of agricultural significance. Here, we developed an approach that allows us to control the trafficking destination of secretory proteins in the human malaria parasite Plasmodium falciparum. Based on the unique structural requirements of apicoplast transit peptides, we designed three conditional localization domains (CLD1, 2, and 3) that can be used to control protein trafficking via the addition of a cell permeant ligand. Studies comparing the trafficking dynamics of each CLD show that CLD2 has the most optimal trafficking efficiency. To validate this system, we tested whether CLD2 could conditionally localize a biotin ligase called Holocarboxylase Synthetase 1 (HCS1) without interfering with the function of the enzyme. In a parasite line expressing CLD2-HCS1, we were able to control protein biotinylation in the apicoplast in a ligand-dependent manner, demonstrating the full functionality of the CLD tool. We have developed and validated a novel molecular tool that may be used in future studies to help elucidate the function of secretory proteins in malaria parasites.

Background

Plasmodium falciparum is the most lethal species of parasite that causes malaria in humans. After infecting a host, P. falciparum travels to the liver where it undergoes a period of development in hepatocytes. Morphological changes occur in the liver that allow the parasite to enter the blood stream, where it begins repeated cycles of asexual division in red blood cells ¹. Both the liver and blood stages of the P. falciparum life cycle rely on the apicoplast to provide essential metabolites for cell survival and organelle maintenance. During the liver stage, the apicoplast produces fatty acids which are necessary for progression to the blood stage ^2,3. In the blood stages, the apicoplast generates isoprenoid precursors ⁴ and iron-sulfur clusters that are required for cell survival ⁵. Despite the importance of the apicoplast at multiple stages in the parasite lifecycle, current knowledge of essential apicoplast biochemistry is largely based on predictions of which nuclear-encoded proteins are required for apicoplast function ⁶. Increasing our knowledge of these essential pathways will improve our understanding of parasite biology and could provide insight into new drug targets for malaria treatment. Research into the molecular biology of the apicoplast, however, is hindered by a lack of tools available to probe the function of specific apicoplast-targeted proteins. To this end, we have designed a Conditional Localization Domain (CLD) that can be used to mislocalize specific apicoplast proteins, allowing researchers to observe the consequences of their loss.

Currently there are only a few options for molecular tools to investigate apicoplast-targeted proteins. These options include genetic knockouts, which can only be applied to non-essential proteins, and conditional degradation domain tags ^7,8. Degradation domains used in P. falciparum have been successful in some cellular contexts, such as the cytosol or nuclear compartment ^9–11 but have had limited success in the secretory pathway, and in particular the apicoplast ^12,13. This observation of context dependent success of the degradation domain has also been reported in mammalian cells ¹⁴ and may be a generalizable feature of this tool. Recently, a translation control tool was designed for use in P. falciparum that requires introduction of an aptamer sequence at the 3’ and 5’ ends of a target mRNA to achieve maximal effect ¹⁵. This methodology may be difficult to implement with apicoplast-targeted proteins because introduction of an aptamer at the 5’ end of the mRNA, if translated with the leader peptide, could interfere with proper trafficking to the apicoplast. Given the small number of molecular tools available to control protein levels in Plasmodium, and the limitations of the current tools, the goal of this study was to develop a conditional localization domain that will allow us to control the localization of apicoplast-targeted proteins.

Previous studies in our laboratory on the targeting motifs required for apicoplast trafficking were instrumental in the conceptualization of the CLD designed in this study ¹⁶. Nuclear-encoded proteins that are trafficked to the apicoplast must contain both an N-terminal signal sequence and transit peptide motif ^17–19. The signal sequence directs the protein into the endoplasmic reticulum, where it is cleaved to reveal the transit peptide ¹⁸. The transit peptide then directs the protein to its final destination in the apicoplast ¹⁹. Previous studies in our laboratory showed that transit peptides must be unstructured to traffic to the apicoplast, and that the formation of secondary structure in the transit peptide region blocks apicoplast import ¹⁶. The CLD builds on this understanding of the structural requirements of transit peptides, and is designed to N-terminally tag a protein of interest and replace the transit peptide motif. In the absence of an effector ligand the CLD functions as a transit peptide and is unstructured to allow the protein to traffic to the apicoplast. When the effector ligand, Shield1, binds to the CLD, it stabilizes the structure of the CLD and causes the protein to be secreted from the cell as shown in Figure 1.

Figure 1. Model of the conditional localization system.

Diagram of how the conditional localization domain can be used to control the localization of a secretory protein. Ap = Apicoplast, PV = Parasitophorous Vacuole, ER = Endoplasmic Reticulum, Nu = Nucleus, SS = Signal Sequence, CLD = Conditional Localization Domain, POI = Protein of Interest, Gg = Golgi.

Results

Design of the conditional localization domain

When designing the CLD to mimic an apicoplast trafficking motif, we first reviewed the characteristics of verified transit peptides. Generally, transit peptides lack a conserved sequence motif, maintain a net positive charge near the N-terminus, and are unstructured during apicoplast import ^16,20,21. Additionally, studies from our lab showed that artificially generated structure in the transit peptide region – through point mutations to form an α-helix – abrogates apicoplast trafficking ¹⁶. Based on the role that structure plays in the function of transit peptides, we reasoned that we could control the localization of secretory proteins if we could modulate the structure of the N-terminal region. Our approach was to replace the native transit peptide with a protein domain with the following properties: 1) small, 2) positively charged at the N-terminus, 3) unstable enough to permit recognition by cell sorting machinery as an unfolded transit peptide, and 4) stable enough to tightly bind a cell permeant, non-toxic ligand. E. coli dihydrofolate reductase (DHFR) and human FK506- and rapamycin-binding protein (FKBP) were evaluated to determine whether they could fulfill these requirements since both proteins have previously been successfully used as molecular tools in P. falciparum ^7,8,22.

Successful CLDs were ultimately derived from FKBP. We began with the F36V mutant of FKBP which creates an artificial pocket allowing synthetic ligands that possess a corresponding “bump” to specifically bind to FKBP_F36V rather than other FKBP proteins ²³. We then introduced the L106P mutation shown by Banaszynski and coworkers to destabilize FKBP, allowing it to be used as a degradation domain tag in conjunction with the ligand Shield1 ²⁴. This degradation domain (dFKBP = FKBP_F36V,L106P) has also been successfully used in P. falciparum, albeit without much success for apicoplast proteins ^7,13. We modified dFKBP with lysine mutations to increase the net positive charge near the N-terminus. We introduced lysine mutations at residues three and five; dFKBP with these mutations – dFKBP_{Q3K, E5K} – is from here on referred to as CLD0 (Supplementary Figure 1 A). CLD0 trafficked to the apicoplast both in the presence and absence of the effector ligand and thus could not be used as a CLD (Supplementary Figure 1 B). In vitro studies of the stability of CLD0 showed that it was less stable than dFKBP, indicating that the positive charge mutations we introduced at the N-terminal region further destabilized the protein (Supplementary Figure 1 C). This suggested that CLD0 is unable to change localization in vivo because it does not bind Shield1 well enough to significantly stabilize its structure. In an effort to address this issue and increase the stability of our CLD constructs, we reverted the dFKBP L106P mutation back to the leucine residue found in the native FKPB sequence for all subsequent CLDs that we evaluated.

We developed three CLDs that can be used independently to control trafficking. As mentioned previously, an important characteristic of natural transit peptides is that they have a net positive charge near the N-terminus ²⁵. To meet this requirement of transit peptides, varying combinations of the same lysine mutations introduced in the original CLD0 protein were made in the FKBP sequence to generate CLD1, 2, and 3. CLD1 is mutated at residue three to convert an uncharged glutamine to a positively charged lysine (FKBP_Q3K); this increases the overall positive charge near the N-terminus +1. CLD2 is mutated at residue five to convert a negatively charged aspartic acid to lysine (FKBP_E5K) and increases the overall charge +2. Finally, CLD3 is mutated at both residues three and five (FKBP_{Q3K, E5K}) to increase the net positive charge +3. A summary of FKBP mutations made to generate CLD1, 2, and 3 is shown in Figure 2 A. As previously described, each CLD also has the F36V mutation allowing the CLD to bind the synthetic ligand Shield1 with higher affinity than endogenous ligands ²³. We generated three transgenic parasite lines to study the capacity of each CLD to traffic a simple cargo protein. These parasites express a verified signal sequence from the Acyl Carrier Protein (ACP) fused to CLD1, 2 or 3 and super folder GFP (SFG) at the C-terminus to track subcellular localization (Figure 2 A) ^26,27.

Figure 2. Design and expression of CLD1, 2, and 3 in P. falciparum.

(A) Test constructs for evaluation of the CLD in P. falciparum include a verified signal sequence fused to CLD1, 2 or 3 and SFG. The N-terminal sequence of CLD1–3 varies as shown. Lysine mutations in blue increase positive charge. The overall change in charge near the N-terminus is listed in the column to the right. ACP Signal = Signal Sequence from Acyl Carrier Protein, SFG = Super Folder GFP (B) Live images of transgenic parasite lines expressing CLD1, 2, or 3. The first two rows on the top show CLD1, the middle two rows show CLD2 and the bottom two rows show CLD3. The rows labeled “Sh-” show cells that have not been treated with Shield1 and rows labeled “Sh+” show cells that have been treated with 500 nM Shield1 for 72 hours. Mitotracker and DAPI mark the location of the mitochondria and nuclei, respectively, while intrinsic fluorescence (SFG) shows the location of the CLD proteins. Additional images for CLD1, 2 and 3 showing different parasite developmental stages are shown in Supplementary Figures 2, 3 and 4. All images are 10 microns wide by 10 microns long.

Expression of CLD1, CLD2, and CLD3 in P. falciparum

We tested the response of each CLD to the addition of the binding ligand, Shield1 (Figure 2 B and Supplementary Figures 2, 3 and 4). The “Sh-” rows in Figure 2 B show trafficking of CLD1, 2 and 3, with no Shield1 added to the culture. The pattern of fluorescence observed in these rows is consistent with previous reports of apicoplast trafficking. In the trophozoite stage, the apicoplast appears as a branched organelle that is distinct from the mitochondrion ²⁸. The “Sh+” rows in Figure 2 B show trafficking of the CLD after cells were treated with 500 nM Shield1 for 72 hours. When Shield1 is added to cells, it stabilizes the CLD structure, and blocks apicoplast import. The pattern of fluorescence in these rows is consistent with the protein being secreted. Proteins that are secreted from the cell accumulate in the parasitophorous vacuole space that separates the parasite from the red blood cell cytosol ²⁸. This experiment shows that all three CLDs meet the basic requirement of trafficking to the apicoplast or secreted compartment under the control of Shield1.To get an understanding of how effectively the CLDs traffic to each compartment in our next set of experiments, we analyzed immunofluorescence images of each CLD and compared them to control trafficking constructs.

Analysis of protein trafficking by the conditional localization domains

We confirmed the trafficking patterns observed in Figure 2 B by co-staining fixed cells with antibodies specific for the apicoplast marker ACP and the CLD trafficked SFG. We observed co-localization between the ACP and SFG signal in cells that had not been treated with Shield1 (Figure 3 A, “Sh-” rows). This co-localization shows that all 3 CLDs traffic SFG to the apicoplast, although CLD1 appears to traffic some protein to the apicoplast and secrete protein simultaneously, which is consistent with a leaky trafficking phenotype. We also stained cells that had been treated with 500 nM Shield1 for 72 hours with the same antibody combination to confirm the change in localization when SFG is secreted (Figure 3 A, “Sh+” rows). In these images, little to no co-localization was observed between SFG and ACP. The SFG signal accumulated around the cell or in the secretory pathway, en route to be secreted from the cell.

Figure 3. Co-localization analysis of protein trafficking by CLD1, 2, and 3.

(A) Immunofluorescence images of transgenic parasite lines expressing CLD1, 2, or 3. The first two rows on the top show CLD1, the middle two rows show CLD2 and the bottom two rows show CLD3. The rows labeled “Sh-” show cells that have not been treated with Shield1 and rows labeled “Sh+” show cells that have been treated with 500 nM Shield1 for 72 hours. The apicoplast is marked by antibodies specific for the Acyl Carrier Protein (αACP), DNA is marked with DAPI, and the CLD proteins are detected with antibodies specific for GFP (αGFP). All images are 10 microns wide by 10 microns long. (B and C) M₁ values were normalized to the average M₁ values calculated from control trafficking constructs (Supplementary Figure 6). M₁ values were calculated using ImageJ software ²⁹. (B) In the “-” Shield1 condition, cells were imaged without addition of Shield1 (CLD1: n = 41 cells, CLD2: n = 31 cells, CLD3: n = 29 cells). Cells in “+” Shield1 condition were treated with 500 nM Shield1 for 72hours (CLD1: n = 14 cells, CLD2: n = 16 cells, CLD3: n = 38 cells). Data were combined from at least two independent conditional localization experiments for each condition. (C) Cells from at least two independent immunofluorescence experiments were imaged without addition of Shield1 (n = 31 cells) or with 500 nM Shield1 added for 6 (n = 20 cells), 12 (n = 21 cells), 24 (n = 24 cells) or 72 hours (n = 28 cells).

For a more quantitative analysis of the efficiency of CLD1, 2, and 3 trafficking, we generated Mander’s overlap coefficient (M₁) values for each CLD from the immunofluorescence images collected as described above. In our analysis, M₁ describes the fraction of total SFG intensities that co-localize with ACP ²⁹. We also generated transgenic parasite lines that express SFG that is either constitutively apicoplast trafficked or constitutively secreted to use as controls to compare with the CLD trafficked protein (Supplementary Figure 5). Figure 3 B shows that M₁ values for CLD1, 2 and 3 were higher for cells not treated with Shield1, indicating that most of the SFG intensities are co-localized with ACP. M₁ values declined in cells treated with Shield1, because SFG is secreted, and thus no longer co-localizes with the apicoplast marker. The percent decline in the average M₁ when Shield1 is added to cultures is about 88 %, 100 %, and 94 % for CLD1, 2, and 3, respectively. CLD2 and CLD3 traffic to the apicoplast at levels similar to the apicoplast trafficking control, while CLD1 exhibits a more leaky apicoplast trafficking phenotype, confirming our qualitative observations from the images in Figure 3 A. While the localization of all three CLDs changes in response to Shield1, CLD2 appears to be the most responsive. We further analyzed the kinetics of localization change when Shield1 is added to CLD2. In Figure 3 C, Shield1 was added at the start of the experiment and samples of the cell culture were taken at 6, 12, 24 and 72 hours after the addition of Shield1 for colocalization analysis. These results demonstrated that most of the change occurs within the first 6 hours after addition of Shield1.

Thermal stability of the conditional localization domains

We hypothesized that the leaky apicoplast trafficking phenotype displayed by CLD1 (Figure 3 A and B), is caused by a higher stability of the CLD1 protein compared to CLD2 or 3. We reasoned that a more stable CLD1 protein would be less able to mimic the unstructured feature of transit peptides, leading to less apicoplast trafficked protein in the permissive (no Shield1) state. To investigate this hypothesis, we expressed CLD1, 2 and 3 proteins and the Shield1 binding FKBP (sbFKBP = FKBP_F36V) in E. coli and determined the relative stabilities of the pure recombinant proteins using thermal shift assays. Our first observation from this experiment was that the destabilized CLD2 and CLD3 proteins could not be purified at concentrations high enough for thermal shift assays without adding Shield1 during protein purification. This suggests that CLD2 and 3 are less stable than CLD1 in the absence of a binding ligand. In order to compare the relative stabilities of the three CLDs, all three were purified with bound Shield1. Table 1 shows that the melting temperatures of CLD2 and 3 are at least 3.8 °C lower than that of sbFKBP, while CLD1 is only 0.3 °C lower. This observation is consistent with our hypothesis that CLD1 is more stable than CLD2 or 3. This analysis also revealed that the charge reversal mutation at residue five (FKBP_E5K) has a more destabilizing effect than the mutation at residue three (FKBP_Q3K) since T_m values for sbFKBP and CLD1 are approximately equal and adding the Q3K mutation to CLD2 – to generate CLD3 – does not significantly change the melting temperature of the CLD.

Table 1.

Thermal stability analysis of CLD proteins

	sbFKBP	CLD1	CLD2	CLD3	dFKBP
− Shield1	54 ± 2.1 °C	–	–	–	–
+ Shield1	77.5 ± 0.4 °C	77.2 ± 0.9 °C	73.7 ± 0.7 °C	73.6 ± 1.2 °C	62.9 ± 0.2 °C
ΔTm	–	−0.3 °C	−3.8 °C	−3.9 °C	−14.6 °C

Thermal shift assays were used to determine the melting temperature of purified proteins. Melting temperatures are shown with standard deviations calculated from triplicate biological replicates, each conducted with triplicate technical replicates.

The change in melting temperature (ΔT_m) was calculated by subtracting the T_m of CLD1, 2, 3 or dFKBP from the T_m of sbFKBP.

sbFKBP = FKBP_F36V

CLD1 = FKBP_{Q3K, F36V}

CLD2 = FKBP_{E5K, F36V}

CLD3 = FKBP_{Q3K, E5K, F36V}

dFKBP = FKBP_{L106P, F36V}

We were interested in estimating whether the destabilization of CLD2 and CLD3 could make them facile substrates for degradation. Although the CLD constructs traffic through the secretory pathway, we expect that they could still be available to the proteasome for degradation after export to the cytosol via the Endoplasmic Reticulum Associated Degradation (ERAD) pathway. We compared the melting temperatures of the CLDs to that of the published degradation domain dFKBP ⁷. We purified dFKBP in the presence of Shield1 (it also required stabilization with this ligand during purification) and conducted thermal shift assays. The difference in T_m between dFKBP and sbFKBP (14.6 °C) was much larger than for any of the three CLDs (Table 1). This suggests that the CLDs are much more stable than the dFKBP degradation domain and less likely to be degraded.

Analysis of sensitivity of the conditional localization domains to Shield1

Next we set out to investigate the limit of sensitivity for each CLD. We titrated the concentration of Shield1 at 5, 25, 125, or 500 nM in cultures for 72 hours before live imaging analysis. Figure 4 A shows that CLD1 was secreted at the lowest concentration of Shield1 added (5 nM Shield1). This is consistent with the thermal stability analysis from Table 1, showing that CLD1 is the most stable domain, and thus the most competent to bind Shield1 in vivo. CLD2 and 3 have similar levels of sensitivity to Shield1 and are secreted at concentrations as low as 25 nM (Figure 4 B and C). This is also consistent with thermal stability data showing that CLD2 and 3 have similar melting temperatures. CLD3, however, exhibits a partially secreted phenotype at 25 nM Shield1 while CLD2 secretion at this concentration is more complete. This slight difference between the responsiveness of CLD2 and 3 could be because of the difference in net positive charge near the N-terminus of the two proteins. Since CLD3 has a more positively charged N-terminus, it may be more inclined to traffic to the apicoplast in low concentrations of Shield1.

Figure 4. Analysis of Shield1 sensitivity of CLD1, 2, and 3.

Live fluorescence images of CLD1 (A), CLD2 (B), and CLD3 (C) with different concentrations of Shield1 added to the culture media. Shield1 was added at various concentrations for 72 hours. Images are 10 microns wide by 10 microns long.

CLD2 exhibited the largest change in localization, and was responsive to low concentrations of Shield1. Based on this analysis, we chose CLD2 for subsequent validation studies to demonstrate use of the CLD to control the localization of an active parasite enzyme.

Validation of CLD2 by tagging the biotin ligase: HCS1

HCS1 (PlasmoDB PF3D7_1026900) is a biotin ligase that is located in the cytosol and is predicted to biotinylate only one protein in malaria parasites - the Acetyl-CoA Carboxylase (ACC). ACC however, is an apicoplast resident protein ³⁰, raising the possibility that HCS1 may have to change subcellular compartments in order to biotinylate and activate ACC. Recent studies show that the biotinylation of apicoplast proteins by HCS1 is critical for liver stage development, but that apicoplast proteins are not biotinylated in blood stage parasites ³¹. We hypothesized that expression of HCS1 in the apicoplast of blood stage parasites would result in the same protein biotinylation observed in liver stage parasites. To test this idea, we conditionally localized an exogenous copy of the P. falciparum HCS1 to the apicoplast using CLD2 (Figure 5 A), to determine whether simply changing the localization of HCS1 in blood stage parasites could control its activity. Figure 5 B and C show that CLD2 traffics HCS1 to the apicoplast in the absence of Shield1, and addition of Shield1 caused HCS1 to be secreted. When HCS1 is trafficked to the apicoplast we detected biotinylated protein in the apicoplast by staining cells with streptavidin FITC (fluorescein isothiocyanate) and co-localizing this signal with the apicoplast marker ACP. When Shield1 was added to the parasite culture the biotinylation activity of HCS1 in the apicoplast was gone (Figure 6). These data suggest that localization of HCS1 could play a role in controlling protein biotinylation in malaria parasites.

Figure 5. Analysis of protein trafficking by CLD2 tagged HCS1.

(A) Transgenic expression construct designed to conditionally localize the biotin ligase enzyme (HCS1) from P. falciparum. (B) Live cell images of CLD2 trafficked HCS1 protein. The rows labeled “Sh-” show cells that have not been treated with Shield1 and rows labeled “Sh+” show cells that have been treated with 500 nM Shield1 for 72 hours. Mitotracker and DAPI mark the location of the mitochondria and nuclei, respectively, while intrinsic fluorescence (SFG) shows the location of the CLD2-HCS1 protein. (C) Immunofluorescence images of cells treated as described in (B). The apicoplast is marked by antibodies specific for the Acyl Carrier Protein (αACP), DNA is marked with DAPI, and the CLD2-HCS1 proteins are detected with antibodies specific for GFP (αGFP). All images in (B) and (C) are 10 microns wide by 10 microns long.

Figure 6. Analysis of HCS1 activity in the apicoplast.

Immunofluorescence images of cells with no Shield1 added (“Sh-”) or 500 nM Shield1 added for 72 hours (“Sh+”). The bottom row shows the same image in the second row with brightness enhanced in FITC channel. Images are 10 microns wide by 10 microns long. SA = streptavidin

Discussion

The Conditional Localization Domain can be added to the small number of molecular tools available to probe the function of proteins in malaria parasites. Our first attempt to design the conditional localization domain involved converting the destabilized DHFR from E. coli into a CLD ⁸. Destabilized DHFR was mutated to increase its positive charge near the N-terminus and expressed in P. falciparum parasites (Supplementary Figure 6 A). However, modified DHFR did not traffic to the apicoplast or change localization when its ligand (trimethoprim) was added to culture medium and thus could not be used as a CLD (Supplementary Figure 6 B). The modified DHFR protein appears to be retained in the secretory pathway and is not recognized as an apicoplast trafficking motif (Supplementary Figure 6 B). In retrospect, DHFR may have been a poor choice for design of the CLD because of the N-terminal structure of the protein. Starting with the second amino acid, the N-terminus forms a central beta strand locked in a beta sheet that forms the core of the DHFR protein structure. Complete unfolding of DHFR might be required to make this region available for recognition by the apicoplast trafficking machinery ³².

These studies led us to choose a different destabilized protein (dFKBP) as the basis for the design of our next candidate CLD. CLD1, 2 and 3 are derived from FKBP and have a structural advantage over the DHFR protein because the N-terminus forms a peripheral beta strand that is not an integral part of the FKBP protein fold ³³. With the appropriate level of protein destabilization, it is easy to imagine this region being frequently available for interaction with the apicoplast trafficking machinery. The challenge with designing a CLD is achieving a metastable state that allows for frequent protein unfolding to expose the trafficking determinants, but also allows frequent sampling of the native state so that the CLD can bind its ligand when present. Surprisingly, relatively little destabilization of FKBP was needed to satisfy these requirements. As shown in Supplementary Figure 1, the mutations in CLD0 were too destabilizing (ΔT_m = −20.3 °C), resulting in a protein that was trafficked to the apicoplast, but presumably did not sample the native state frequently enough to efficiently bind Shield1. On the opposite side of the spectrum, CLD1 is only slightly destabilized (ΔT_m = −0.3 °C), resulting in a protein that is not efficiently trafficked to the apicoplast, presumably because it did not unfold frequently enough to expose the trafficking determinants. For FKBP, only modest destabilization (ΔT_m ≈ −4 °C) was needed to obtain the optimal properties found in CLD2 and 3. These CLDs are significantly more stable than the dFKBP mutant (ΔT_m = −14.6 °C) used in degradation domain approaches ⁷. Consequently, SFG tagged with CLD2 remains an intact fusion protein, even in the absence of Shield1 (Supplementary Figure 7).

CLD1, 2 and 3 are functionally similar domains that traffic to the apicoplast under permissive conditions and are secreted when a binding ligand is added to the parasite culture (Figure 2 B). Although we chose to use CLD2 in our validation studies to localize the HCS1 biotin ligase, the differences in the efficiency of trafficking among the three CLDs (Figure 3) could make CLD1 or 3 more attractive. For example, because of its leaky apicoplast trafficking phenotype, CLD1 may be optimal for the expression of proteins that perturb or interfere with apicoplast functions. The high stability of CLD1 would limit the trafficking of the toxic protein to the apicoplast in the presence of Shield1, facilitating generation of the transgenic parasite line. Conversely, CLD3 traffics to the apicoplast more efficiently than it is secreted and thus could be suitable for projects that investigate the function of a resident parasitophorous vacuole (PV) protein, since removal of the protein from the PV space would be most complete with the CLD3 tag. CLD2 exhibits the largest change in localization when Shield1 is added and traffics to the apicoplast and secreted compartments at levels similar to the trafficking controls. For most projects, CLD2 should be ideal to tag proteins when efficient trafficking to both compartments is desired. Although we validated the CLDs using the ligand Shield1, other ligands such as rapamycin are cost effective and non-toxic ³⁴ and are known to bind tightly to the F36V mutant of FKBP ³⁵.

In addition to the design and characterization of CLDs, we also used a CLD to control the localization of a parasite biotin ligase (HCS1). HCS1 controls the first and rate-limiting step in the Fatty Acid Synthesis pathway by biotinylating the ACC. During the erythrocytic cycle HCS1 and ACC are trafficked to separate compartments and are not predicted to interact at any point during trafficking³¹. Given this observation, it makes sense that blood stage malaria parasites don’t require biotin for growth ³¹ and that fatty acid synthesis is not essential during this stage of parasite development ^2,3. Our data show that HCS1 can be active in the apicoplast of blood stage parasites, and that this activity can be removed by changing its localization. This suggests that compartmentalization of HCS1 and ACC plays a key role in controlling HCS1 activity. This regulation may also be a factor in controlling fatty acid synthesis in blood stage parasites.

Methods

Generation of plasmid constructs

The human FKBP gene with the Shield1 binding mutation ⁷ was amplified using forward primers that contained nucleotide changes to produce lysine mutations at residues 3, 5, or both (to generate CLD1, 2 or 3 respectively). NdeI.FKBPq3k.for was used to generate CLD1, NdeI.FKBPe5k.for was used for CLD2, and NdeI.Fkkfor was used for CLD3. One reverse primer (FKBP.P106L.BglII.rev) was used to amplify the FKBP gene in all three reactions. To generate CLD0 NdeI.Fkkfor forward primer was used with FKBP.L106P.BglII.rev to amplify the FKBP gene. FKBP inserts were then digested using NdeI and BglII and ligated into cloning vectors (GeneArt) that contained the synthesized signal sequence from the ACP gene and the super folder GFP sequence (Supplementary Figure 8). The CLD sequences were then ligated in to a modified pLN ³⁶ vector for parasite transfection that contained the lower strength ribosomal L2 protein promoter ³⁷ instead of the calmodulin promoter using Quick Ligase (New England BioLabs).

Plasmid constructs used to generate parasites expressing control trafficking proteins (SS_ACP-SFG and SS_ACP-TP_ACP-SFG) were generated by ligating synthesized targeting sequences (from GeneArt) into the modified pLN vector ⁵ described above.

Plasmids used for protein expression in E. coli were generated by PCR amplifying the human FKBP gene with the Shield1 binding mutation ^7,23 using forward primers that contained nucleotide changes to produce lysine mutations at residues 3, 5, or both (to generate CLD1, 2 or 3 respectively). Fk3e5.EcoRI.LIC.for was used to generate CLD1, Fq3k5.EcoRI.LIC.for was used for CLD2, and Fkk.EcoRI.LIC.for was used for CLD3. Construct sbFKBP was amplified was amplified using the Fqe.EcoRI.LIC.for primer which does not generate any mutations at the N-terminus. One reverse primer (FKBP.P106L.HindIII.LIC.rev) was used to amplify the FKBP gene to produce these four constructs (CLD1, CLD2, CLD3 and sbFKBP). Construct dFKBP was amplified using the Fqe.EcoRI.LIC.for forward primer with the FKBP.P106.HindIII.LIC.rev reverse primer and construct CLD0 was amplified using the Fk3e5.EcoRI.LIC.for forward primer with the FKBP.P106.HindIII.LIC.rev reverse primer. For all six constructs, PCR amplicons were then inserted into the pMALcHT ³⁸ E. coli expression vector using ligase independent cloning with T4 DNA polymerase.

The HCS1 gene (3D7_1026900) was amplified from plasmid DNA using BsrG1.HCS10.for and BsiWI.HCS10.rev and digested with BsiWI and BsrGI. The previously described parasite expression vector containing CLD2 was then also digested with BsiWI. The HCS1 gene was ligated into the expression vector for parasite expression using Quick Ligase (New England Biolabs). All DNA sequences were confirmed by sequencing after insertion. The sequences of the primers used to generate plasmid constructs described in this section are listed in Supplementary Table 1.

Parasite transfection and culture

Parasites were cultured at 2 % hematocrit in RPMI 1640 medium containing 25 mM HEPES, 0.375 % sodium bicarbonate, 12.5 μg/ml hypoxanthine, 5 g/L Albumax II and 25 μg/ml gentamicin. Transfections were done using the Bxb1 mycobacteriophage integrase system in a Dd2 strain of parasites that contain an attB site for recombination ³⁶. Uninfected red blood cells were preloaded with transfection plasmids and electroporated using the protocol from Spalding and coworkers ³⁹. Electroporated red blood cells were then mixed with parasite culture and after two days of growth transgenic parasites were selected with 2.5 μg/ml Blasticidin. Integration was confirmed using primers that bind in the pLN plasmid and endogenous locus (Supplementary Figure 9).

For experiments in which parasites were synchronized, a homemade magnet with field strength of about 8,000 G was used to separate schizonts from mixed stage cultures. Briefly, a MACS LS Column (Miltenyi Biotec) was inserted into the magnet and infected red blood cell culture was pipetted into the top of the column. When the column is attached to the magnet schizonts stick to the magnetized beads while trophozoite and ring stage parasites flow through. Schizonts were eluted from the column by removing it from the magnetic field and running an additional 5 mL of media through the column.

Protein expression and purification

Plasmids pMALcHT-sbFKBP, pMALcHT-dFKBP, pMALcHT-CLD0, pMALcHT-CLD1, pMALcHT-CLD2, pMALcHT-CLD3 were transformed into BL21-Star (DE3) cells and co-transformed with the pRIL plasmid isolated from BL21-CodonPlus-RIL cells ⁴⁰. These cells produce a protein product fused to an amino-terminal Maltose Binding Protein (MBP) tag followed by a Tobacco Etch Virus (TEV) protease cleavage site and a six-histidine tag. CLD expression cells were grown to an OD₆₀₀ of 3 in TB (Terrific Broth) medium shaking at 37 °C. Protein expression was induced with 0.4 mM IPTG (Isopropyl β-D-1-thiogalactopyranosid) for 3 hr. Cells were harvested by centrifugation and resuspended in 20 mL of lysis buffer (20 mM HEPES pH 7.4, 500 mM NaCl, 1 mM DTT, 2.5 µg/mL DNAse I, and 1 mg/mL Lysozyme) per liter of cell culture. Resuspended cells were lysed by sonication and the lysate was clarified by centrifugation. The resulting supernatant was loaded onto an MBPTrap HP column (GE Life Sciences) and eluted with 100 mM maltose in loading buffer (20 mM HEPES pH 7.4, 500 mM NaCl, 1 mM DTT). MBP fusion proteins were cleaved with TEV protease at room temperature for 2–4 hours in the presence of 1 mM DTT, 0.5 mM EDTA and 40 µM Shield1 (when required). The cleavage product was dialyzed into 20 mM HEPES pH 7.4 and 500 mM NaCl and purified from the MBP tag using a HiTrap Chelating HP column (GE Life Sciences). Appropriate fractions were collected and further purified by size exclusion chromatography with a HiPrep 26/60 Sephacryl S-100 HR column (GE Life Sciences) equilibrated in 20 mM HEPES pH 7.4 and 100 mM NaCl.

GFP was produced to generate affinity purified antibodies (see below). Primers GFP.pMAL.EcoRI.For and GFP.pMAL.HindIII.Rev were used to amplify GFP from plasmid pLNα62¹⁶ for insertion into the pMALcHT E. coli expression vector. Plasmid pMALcHT-GFP was transformed into BL21 Star (DE3) cells (Invitrogen) and cotransformed with the pRIL plasmid and a plasmid encoding the TEV protease ³⁸. After in vivo cleavage of the amino-terminal MBP fusion protein, these cells produce GFP with a six-histidine tag. GFP expression cells were grown to an OD₆₀₀ of 0.8 in LB (Luria-Bertani) medium shaking at 37 °C. Protein expression was induced with 0.4 mM IPTG for 10 hr at 20 °C. Cells were harvested by centrifugation and resuspended in 20 mL of lysis buffer (10 mM HEPES pH 7.5, 200 mM NaCl, 2.5 µg/mL DNAse I, and 1 mg/mL Lysozyme) per liter of cell culture. Resuspended cells were lysed by sonication and the lysate was clarified by centrifugation. The resulting supernatant was loaded onto a HiTrap Chelating HP column (GE Life Sciences) and eluted with 10 mM HEPES pH7.5, 100 mM Imidazole. The eluate was applied directly to a HiTrap Q FF anion exchange column (GE Life Sciences) and eluted with a gradient to 1 M NaCl.

Live cell imaging

Parasite cultures were stained with DAPI and Mitotracker Red CMX Ros (Invitrogen). 100 μl of parasite culture was incubated for 30 minutes in 1 μg/mL DAPI and 30 nM Mitotracker Red CMX Ros at 37 °C. Samples were then washed three times in culture media and pipetted onto microscope slides. A coverslip was placed over the slide and sealed with wax (2 parts paraffin, 1 part Vaseline). Samples were then taken immediately to the Zeiss AxioImager M2 microscope for imaging.

Immunofluorescence assay

Microscope slides were prepared for immunofluorescence assays by drawing wells on the slide with a Super Pap Pen Liquid Blocker (Ted Pella, Inc.). A 0.01 % poly-L-Lysine solution (Sigma-Aldrich) in water was added to each well and allowed to dry for at least 30 minutes. 300 μL of parasite culture was then spun down and resuspended in an equal volume of fixative (4 % paraformaldehyde and 0.0075 % glutaraldehyde in PBS). Cells were added to each well on the slide and then incubated for 30 minutes at room temperature. After incubation, fixed cells were permeabilized by incubation in 1 % Triton for ten minutes. The samples were then reduced by incubation in 100 μg/mL NaBH₄ in water for 10 minutes. Next the cells were incubated in blocking solution (3 % BSA in PBS) for two hours. Before applying primary antibodies, cells were washed in PBS and then incubated with appropriate antibodies overnight at 4 °C [rat polyclonal αACP 1:2,000, raised against the P. falciparum antigen ⁴¹; rabbit polyclonal αGFP 1:10,000 raised against recombinant GFP protein (see below)]. The next day, the cells were washed three times in PBS and then once in 3 % BSA. Appropriate secondary antibodies [goat αRabbit AlexaFluor 594 1:1,000 (Life Technologies); donkey αRat AlexaFluor 488 1:3,000 (Life Technologies)] or streptavidin-FITC reagent 1:50 (Sigma) were added to cells and incubated for 2 hours in the dark at room temperature. Finally cells were washed in PBS three times and sealed with ProLong Gold antifade reagent with DAPI (Life Technologies) under a coverslip sealed with nail polish. Slides were incubated overnight at room temperature before imaging analysis.

Generation and purification of anti-GFP mouse antibodies

Pure recombinant GFP was used to generate rabbit antiserum using the custom antibody service of Cocalico Biologicals Inc. Briefly, 250 µg of GFP mixed with Complete Freund’s Adjuvant was used for the initial inoculation followed by boosts of 125 µg of antigen 2, 3, and 7 weeks later. Final exsanguination was performed on day 56. Specific antibodies were purified from antiserum using a GFP affinity column using methods similar to those previously described ⁴¹. A 1 mL NHS-activated HP column (GE Healthcare) was activated with 1 mM HCl according to the manufacturer’s instructions. Immediately after activation, 10.8 mg of GFP in 4 mL of reaction buffer (100 mM NaCO₃H, 500 mM NaCl pH 8.3) was pumped through the column at a constant rate of 0.1 mL/min. GFP was circulated through the column for 2 hr at room temperature at which point the release of N-hydroxysuccinimide reaction product was quantified by absorbance at 260 nM (ε=8600 M⁻¹cm⁻¹). The column was subsequently washed and blocked according to the manufacturer’s directions. Rabbit antiserum (13.5 mL) was circulated over the affinity column at room temperature for 2 hours at 0.5 mL/min. Nonspecific proteins were washed from the column with 10 mL of PBS followed by elution in 4 mL of 50 mM glycine pH 1.9. A total of 18.6 mg of αGFP IgG was concentrated to 3.1 mg/mL and stored at −80 °C in storage buffer (PBS, 40% glycerol, 0.02% NaN₃).

Thermal shift assay

Stability of the different FKBP constructs was determined using a thermal shift assay as previously described ⁴² with minor modifications. RT-PCR tube strips (Eppendorf) were used to hold 30 μL mixtures containing final concentrations of 40 μM (0.5 mg/mL) sbFKBP or FKBP mutant and 200 μM Shield1. Assay samples were generated by first adding Shield1 dissolved in ethanol and evaporating the solvent at 37 °C. Shield1 was re-suspended in buffer (HEPES pH 7.4 and 100 mM NaCl) followed by addition of sbFKBP or FKBP mutant and 1 μL of Sypro Orange (Sigma, product no. S-5692). The samples were incubated in a RT-PCR machine (Applied Biosystems, Step One Plus Real-Time PCR System) for 2 min at 20 °C followed by an increase in temperature of 0.2 °C per 10 s up to a final temperature of 80 °C. Fluorescence was monitored in the Step One Plus Real-Time PCR system using a TAMRA filter in which an increase in Sypro Orange fluorescence (excitation: 480 nm, emission: 568 nm) was observed upon thermal denaturation of the protein sample. Temperature and melt curve data were exported from the StepOne v2.3 software program and analyzed to determine the melting temperature. All thermal shift assays were done with triplicate technical replicates for the indicated number of biological replicates given in figure legends.

Statistical Analyses

In experiments where only two groups were compared, we conducted the Mann-Whitney U test for analysis of non-normal, unpaired data sets. Comparisons of three or more groups were conducted using the Kruskal-Wallis one-way analysis of variance for non-normal, unpaired data sets. For both of these analyses, tests were considered statistically significant if the P-value was less than 0.05.

Supplementary Material

Supp info

Supplementary Table 1. Primer sequences used for cloning. This table corresponds to the methods section on “Generation of plasmid constructs” and provides sequences for primers listed in this section.

Supplementary Figure 1. Design and analysis of CLD0. This figure corresponds to the results section on “Design of the conditional localization domain” and provides information on our original design and results from our first attempt to use FKBP as a CLD.

Supplementary Figure 2. CLD1 localization at each stage of parasite development in the red blood cell. This figure corresponds to Figure 2 B and shows that CLD1 localization is consistent at all stages of parasite development in the red blood cell.

Supplementary Figure 3. CLD2 localization at each stage of parasite development in the red blood cell. This figure corresponds to Figure 2 B and shows that CLD2 localization is consistent at all stages of parasite development in the red blood cell.

Supplementary Figure 4. CLD3 localization at each stage of parasite development in the red blood cell. This figure corresponds to Figure 2 B and shows that CLD3 localization is consistent at all stages of parasite development in the red blood cell.

Supplementary Figure 5. Apicoplast and secreted trafficking controls. This figure corresponds to Figure 3 in the main text and provides information (including images and average M₁ values) on the constitutively trafficked expression constructs that were used as controls for the analysis shown in Figure 3.

Supplementary Figure 6. Design and analysis of CLD:EcDHFR. This figure corresponds to the discussion section and provides information on our attempts to use DHFR as a CLD, including the CLD design and images of the CLD:EcDHFR expressing parasite line.

Supplementary Methods. Generation of CLD:EcDHFR plasmid construct. These supplemental methods support supplementary Figure 3 and the discussion section in the main paper. This section provides primer sequences and the harmonized DHFR gene sequence that was expressed in P. falciparum.

Supplementary Figure 7. Anti-GFP western blot of CLD2-SFG parasites cultured in the absence of the stabilizing ligand Shield1. Three independent parasite samples were collected and analyzed by western blot to assess protein degradation (lanes 2–4). A major band consistent with the mass of intact CLD2-SFG was observed (~42 kDa) with two minor degradation bands, including a band consistent with truncated SFG (~28 kDa). For comparison, lysate from positive control parasites constitutively expressing apicoplast SFG is shown in lane 1.

Supplementary Figure 8. DNA sequence for the Super Folder Green protein expressed in P. falciparum parasites. This figure corresponds to figures 2, 3, 4, 5, and 6 in the main paper and supplemental figures 1 and 2. It provides the DNA sequence of the SFG tag appended to the C-terminus of each of the transgenic proteins expressed in this study.

Supplementary Figure 9. Integration confirmation PCR analysis for transgenic parasite lines used in this study. This figure corresponds to figures 2, 3, 4, 5, and 6 in the main paper and supplemental figures 1, 2, and 3. It provides integration PCRs for all of the parasite lines described in this study, including primer sequences and a diagram of the integrated locus with primer binding sites.

Synopsis.

Here, we present the development and implementation of a novel molecular tool to control protein trafficking in malaria parasites. We have designed and evaluated three conditional localization domains (CLD1, CLD2, and CLD3) that can be used to control protein trafficking via the addition of a small cell permeant ligand. To validate this conditional localization approach we used the CLD to conditionally localize a parasite biotin ligase while retaining the enzymatic activity of the enzyme.