Using SLI (Birnbaum et al., 2017), parasites were genetically modified to be resistant to G418 if expressing a targeted var gene (Figure 1A), permitting selection of a population of parasites expressing the desired PfEMP1. In addition, the chosen PfEMP1 obtains a C-terminal 3xHA tag to specifically detect it (Figure 1A). We first aimed to generate two different 3D7 parasite lines: in the first, we targeted PF3D7_0809100 (3D7var0809100, the predominant var gene in our 3D7 wildtype Figure 1—figure supplement 1A), and in the second, PF3D7_1200600, the 3D7 VAR2CSA-encoding var gene. Cell lines with the expected genomic modification were obtained in both cases (3D7var0809100-HA^endo and 3D7var2csa-HA^endo parasites) and the HA-tagged PfEMP1 was detected at the Maurer’s clefts in the host cell (Figure 1B and C), indicating that the cells in the culture expressed the desired PfEMP1 and that it could conveniently be detected via the epitope tag. qPCR showed predominant expression of the activated var genes (Figure 1E and F). This was confirmed by RNA-Seq (Figure 1—figure supplement 1B, Source data 1), which showed high read coverage across the desired var gene, whereas transcripts of all other var genes were negligible (example in Figure 1—figure supplement 1C, Source data 1). As PfEMP1 surface exposure is not typically detected using standard immunofluorescence assays, we conducted trypsin digestion assays with intact infected RBCs (Waterkeyn et al., 2000) which showed a protected fragment indicative of surface exposure of the HA-tagged PfEMP1 (Figure 1D). When we lifted G418 pressure for 3 weeks, the dominance of the targeted PfEMP1 declined in favor of a more heterogeneous var expression profile, indicating that switching to other vars was still possible in these parasites (Figure 1E and F).

Previous work indicated that SLI to select parasites expressing a specific var gene can influence transcription of neighboring genes that are oriented head-to-tail (Omelianczyk et al., 2020). We did not observe activation of head-to-tail oriented rifs in our 3D7var0809100-HA^endo and 3D7var2csa-HA^endo SLI lines, as the corresponding rifs showed no or negligible transcription in RNA-Seq (Figure 1—figure supplement 1D). To further look into co-activation, we used SLI to select for parasites expressing a var gene that shares a promoter region with a rif gene in a head-to-head orientation (PF3D7_0425800: cell line 3D7var0425800-HA^endo) (Figure 1G, Figure 1—figure supplement 1E). Correct integration of the plasmid into the genome and expression of the tagged PfEMP1 was confirmed (Figure 1G). RNA-Seq showed predominant expression of the activated var gene (Figure 1G) but also transcription of the neighboring rif gene (Figure S1F) with ~40% of all rif transcripts belonging to this rif gene (PF3D7_0425700; Figure 1—figure supplement 1G).

We also activated two rif genes (PF3D7_0425700: cell line 3D7rif0425700-HA^endo and PF3D7_1254800: cell line 3D7rif1254800-HA^endo; Figure 1—figure supplement 1H). The two resulting cell lines had the correct genomic modification and IFAs indicated expression of the HA-tagged RIFINs (Figure 1—figure supplement 1H). qPCR showed that in both lines the activated rif gene was the most expressed (~65% and ~40% of rif transcripts; Figure 1—figure supplement 1I). While in the 3D7rif1254800-HA^endo parasite line, the var expression profile looked similar to the 3D7 parent with predominant expression of var PF3D7_0809100 located on a different chromosome (Figure 1—figure supplement 1I), activation of 3D7rif0425700 (which, in contrast to 3D7rif1254800, has a var gene in head-to-head orientation) led to co-activation of the neighboring var gene (PF3D7_0425800; Figure 1—figure supplement 1I). Overall, the data with activated vars and rifs suggests that neighboring genes can be co-activated if in head-to-head orientation, likely due to a shared promoter region affected by the epigenetic changes resulting in the expression of the SLI-targeted var gene. While SLI-activation of rif genes also led to the dominant expression of the targeted rif gene, other rif genes still took up a substantial proportion of all detected rif transcripts, speaking against a mutually exclusive expression in the manner seen with var genes.

Previous work described a 3D7 line expressing multiple var genes in single infected RBCs (Joergensen et al., 2010), likely due to a defective var regulation system, here designated 3D7^MEED (for ‘mutually exclusive expression defective’). We used SLI to obtain parasites expressing PF3D7_0425800 in the 3D7^MEED parasites (3D7^MEEDvar0425800-HA^endo; Figure 1H). In contrast to standard 3D7 with the same SLI modification to express 3D7var0425800, 3D7^MEED showed elevated levels of transcription of multiple var genes in addition to the activated one, both by qPCR (Figure 1—figure supplement 1E) and by RNA-Seq (Figure 1H and I). Anti-HA IFAs showed that the 3D7^MEED parasite nevertheless expressed the activated PfEMP1 (all trophozoites were HA-positive in IFAs (n=82 parasites from four independent experiments)) (Figure 1H), indicating that individual parasites expressed multiple var genes. This line might therefore be an interesting tool to study mutually exclusive expression, silencing, and switching mechanisms.

In an attempt to find changes that may cause the MEED phenotype, we compared all differentially expressed transcripts (161 down- and 93 up-regulated using a log2 fold change of >2 and Padjusted of <0.05 as cut off) of the 3D7 vs 3D7^MEED parasites (Figure 1J, Figure 1—source data 6). This confirmed the upregulation of most var genes in the 3D7^MEED parasites. Members of some other gene families encoding exported proteins and genes of other exported proteins were also upregulated. Many of these may be co-regulated with the vars, as for instance many of the upregulated rif gene loci are close to upregulated var loci (Figure 1—source data 6). However, a few genes encoding exported proteins were also downregulated. Concentrating on non-exported proteins to identify potential changes responsible for the MEED phenotype, we noticed that the transcripts encoding the ApiAP2 protein SIP2 (Flueck et al., 2010) were down ~eightfold (pAdjusted ~0.025; Figure 1J, Figure 1—source data 6). SIP2 was previously shown to bind heterochromatin in subtelomeric and telomeric regions, including certain var promoters (Flueck et al., 2010). Its downregulation might result in changes to chromosome end biology influencing var silencing. The other potentially causal change was an upregulation of the non-coding RNA ruf6 for which overexpression has been shown to impair monoallelic var gene expression (Guizetti et al., 2016). While both sip2 downregulation and ruf6 upregulation are possible explanations for the relaxed silencing of var genes in the 3D7^MEED parasites, independent experiments are needed to confirm that any of these changes are reasons for the MEED phenotype.

Next, we tested if SLI would permit obtaining parasites all expressing a modified PfEMP1 to track and study its transport. Limited overlap of PfEMP1 with PTEX components had raised the question whether it is transported via PTEX or not (McMillan et al., 2013; Riglar et al., 2013). While ablating PTEX function blocks PfEMP1 transport, indicating PTEX-dependent PfEMP1 transport (Beck et al., 2014; Elsworth et al., 2014), this may also be an indirect effect as inhibiting PTEX function also blocks the transport of other exported proteins essential for PfEMP1 transport. To directly assess PfEMP1 transport through PTEX, we used SLI to obtain parasites expressing VAR-0809100 (PF3D7_0809100) or VAR2CSA (PF3D7_1200600) and at the same time tagged them with mDHFR-3xHA (Figure 2A, Figure 2—figure supplement 1). The folding of the mDHFR domain can be stabilized by addition of WR99210 (WR) which prevents transport through translocons requiring unfolding (Eilers and Schatz, 1986) and can be used to assess PTEX passage of soluble and transmembrane proteins in P. falciparum parasites (Gehde et al., 2009; Grüring et al., 2012; Figure 2A). Both mDHFR-fused PfEMP1s were efficiently exported to the Maurer’s clefts but not blocked when WR was added (Figure 2A). While this suggests that PfEMP1 might not be transported via PTEX, we previously noted that the length of the region between the transmembrane and mDHFR domain influences its blocking properties in exported transmembrane proteins (Mesén-Ramírez et al., 2016) and it therefore cannot be fully excluded that this protein still uses PTEX for transport. To circumvent this problem, we exploited the property of proteins that – when fused to mDHFR – can be conditionally arrested in the PTEX translocon, preventing the passage of other exported proteins (Mesén-Ramírez et al., 2016). Blocking PTEX in that manner (using SBP1-mDHFR-GFP conditionally arrested in PTEX) prevented PfEMP1 transport (Figure 2B), suggesting the need of PTEX function for PfEMP1 transport. However, similarly to previous work inactivating PTEX components (Beck et al., 2014; Elsworth et al., 2014), this does not exclude that PfEMP1 trafficking was prevented due to other exported proteins needed for PfEMP1 that were themselves prevented from export through PTEX. We therefore expressed the PTEX blocking construct later in the cycle (using the crt promoter) in an attempt to block PTEX passage only after the PfEMP1 trafficking proteins had already reached the host cell (Figure 2C). Using this strategy, the early expressed PfEMP1-transport protein REX1 was still exported while a later-stage episomally expressed mScarlet-tagged KAHRP reporter could be blocked and was used to monitor clogging of PTEX after the parasite ring stage (Figure 2C and D). We then inspected REX1 and PfEMP1 transport in the cells where the KAHRP-mScarlet reporter showed a late block of PTEX (Figure 2D). REX1 was exported in most of the cells with a blocked KAHRP reporter, indicating that proteins needed for PfEMP1 trafficking were not hindered in reaching the host cell (Figure 2D). However, PfEMP1, which is later expressed, showed accumulation around the parasite in the majority of cells (Figure 2C). This indicated that clogging PTEX later in the cycle prevented PfEMP1 transport, supporting that PfEMP1 passes through PTEX. While we monitored only REX1, most PfEMP1-trafficking proteins show a similar early expression (McMillan et al., 2013; Grüring et al., 2011; Marti et al., 2004), indicating the effect may be direct, favoring the idea that PfEMP1 passes through PTEX. This would also mean that the mDHFR-based translocation block is not effective in the PfEMP1-mDHFR fusion construct.

Next, we tested whether SLI PfEMP1-expressor parasites are useful for PfEMP1 binding studies. Initial experiments using 3D7 parasites with activated vars showed no or only minimal binding of infected RBCs to receptors (see below). We therefore moved to the FCR3/IT4 clone generally considered a cytoadhesion-competent parasite line (Bourke et al., 1996; Udeinya et al., 1983) and used SLI to generate parasites expressing PfIT_040025500 (IT4var66), predicted to encode a CD36-binding PfEMP1 (Hsieh et al., 2016) with a similar domain composition to 3D7var080910, as well as PfIT_120006100 (IT4var2csa), encoding the IT4 VAR2CSA variant (Figure 3A and B). Also, in IT4, SLI was effective to obtain parasites expressing the targeted vars, and based on IFAs, the HA-tagged PfEMP1 was expressed in the corresponding parasite lines (Figure 3A and B). RNA-Seq showed predominant expression of the activated var genes (Figure 3C and Source data 1), and trypsin assays that the expressed PfEMP1 was presented on the RBC surface (Figure 3D). For binding studies, we developed and validated a semi-automated pipeline to score the number of bound infected RBCs in binding assays to permit the unbiased and higher throughput scoring of bound infected RBCs and increase the number of fields that can be analyzed per assay (Figure 3—figure supplement 1). Both IT4 lines showed the expected receptor binding. The IT4var2csa-HA^endo infected RBCs bound both decorin-coated slides and endothelial cells expressing CSA (HBEC-5i; Dörpinghaus et al., 2020), and binding was inhibited by soluble CSA (Figure 3E and F). RBCs infected with IT4var66-HA^endo bound to CHO-CD36 but not CHO-GFP or CHO-ICAM-1 cells (Figure 3G). In contrast, the 3D7 parasites expressing VAR2CSA (3D7var2csa-HA^endo) or 3D7var0809100 (3D7var0809100-HA^endo, predicted to bind CD36 Hsieh et al., 2016) did not or only poorly bind their respective receptors (Figure 3E–G) despite the expressed PfEMP1 being detectable on the surface (Figure 1D). While the binding properties of PfEMP1 are difficult to compare between strains, VAR2CSA was chosen because it is well-conserved between isolates, permitting a comparison of binding efficiencies between 3D7 and IT4. Hence, these findings indicate that IT4 is a better cytoadhesion binder than the 3D7 used in our lab.

VAR2CSA is assumed to be the only PfEMP1 encoded in the genome that binds CSA. No binding to CSA was observed with the parental IT4 parasites (Figure 3E and F) in agreement with the low levels of var2csa transcripts in these parasites (Figure 1—figure supplement 1J). This indicated that SLI targeting var2csa selected this binding phenotype from undetectable levels. Overall, we conclude that SLI generated PfEMP1 expressor lines in IT4 can be used to study binding of specific PfEMP1 and that 3D7 - at least the version from our lab - is less suitable.

In order to extend the repertoire of cell lines to study PfEMP1 binding, we selected two known or suspected CD36- and ICAM-1-binders (PfIT_060021400; cell line IT4var01-HA^endo and PfIT_120024500; cell line IT4var16-HA^endo; Howell et al., 2008; Janes et al., 2011; Metwally et al., 2017) and an EPCR-binder (PfIT_010005000: cell line IT4var19-HA^endo; Turner et al., 2013). Correct generation of the cell lines and expression of the desired var gene was confirmed (Figure 4A–C). Trypsin assays showed surface exposure of the IT4var01 and IT4var16 but not IT4var19 encoded PfEMP1 (Figure 4D). Both RBCs infected with IT4var01-HA^endo and IT4var16-HA^endo bound to CHO-CD36 and CHO-ICAM-1 cells but not GFP-expressing CHO cells, in agreement with the expected binding properties (Figure 4E). IT4var16-HA^endo parasites showed a higher binding capacity to ICAM-1 than IT4var01-HA^endo parasites, whereas IT4var01-HA^endo parasites showed proportionally more binding to CD36 than IT4var16-HA^endo parasites (Figure 4E). In contrast, the IT4var19-HA^endo parasites showed no significant binding to EPCR, CD36, or ICAM-1 (Figure 4F). However, after five rounds of panning against EPCR-expressing CHO cells, the panned IT4var19-HA^endo parasites exhibited significant binding to EPCR and to a lower degree to ICAM-1 (Figure 4F) even though the var transcript profile was not noticeably altered compared to the unpanned IT4var19-HA^endo parasites and still showed predominant expression of IT4var19 with similar overall var transcript levels (Figure 4G; Figure 1—figure supplement 1K). Trypsin assay of the panned parasites showed surface expression of IT4VAR19-HA (Figure 4H), contrasting with the unpanned parasites (Figure 4D). RNA-Seq indicated only a few genes differently transcribed in the panned compared to the unpanned IT4var19-HA^endo parasites (Figure 4I). Interestingly, this included the two paralogs of ptp3 that both were upregulated after panning. The single ptp3 gene in 3D7 encodes a protein needed for PfEMP1 surface display in 3D7 (Maier et al., 2008), suggesting low ptp3 expression as the reason for failure of the unpanned IT4var19-HA^endo parasites to bind. As the two ptp3 loci in IT4 are more than 10 genes apart, including 3 genes with comparable expression levels (to the ptp3 genes) that were not differentially expressed in the panned parasites (PfIT_140083600, PfIT_140084100, PfIT_140084200), the initially low expression likely was not due to a genomic deletion but plastically altered transcription of ptp3 genes.

In summary, we obtained parasites binding to the most common receptors although in the case of EPCR, we had to pan the parasites and detect binding to ICAM-1 in addition to EPCR, which was unexpected (Avril et al., 2012; Nunes-Silva et al., 2015; Adams et al., 2021).

To further exploit the SLI-generated PfEMP1 expressor parasites, we generated a second SLI plasmid system with different selection markers termed SLI2 to modify the genome of parasites already carrying a SLI modification (Figure 5A). To test SLI2, we used the IT4var01-HA^endo line and applied SLI2 to disrupt PTP1 (Maier et al., 2008), a Maurer’s clefts located PfEMP1 trafficking protein. Integration of the SLI2 plasmid into the correct genomic locus and perpetuation of the first genomic modification was confirmed by diagnostic PCR (Figure 5B). Live cell imaging showed a GFP signal in the food vacuole and faint dispersed signal in the host cell, confirming successful inactivation of PTP1 (Figure 5C). IFAs showed that in the PTP1-TGD parasites, SBP1 and PfEMP1 were found in many small foci in the host cell that exceeded the average number of ~15 Maurer’s clefts typically found per infected RBC (Blancke Soares et al., 2025; Figure 5D). This phenotype resembled the previously reported Maurer’s clefts phenotype of the PTP1 knockout in CS2 parasites (Rug et al., 2014). PfEMP1 was still transported into the host cell in the PTP1 disruption parasites (Figure 5D) but PfEMP1 was no longer surface exposed (Figure 5E) and the parasites failed to bind to CD36 and ICAM-1 (Figure 5F), indicating that the IT4var01-HA^endo parasites with the PTP1-TGD had lost their ability for cytoadhesion. While we did not detect the failure of PfEMP1 transport into the host cell, the binding phenotype agrees with previous work (Rug et al., 2014; Maier et al., 2008). Hence, SLI2 permits the study of other proteins in SLI generated PfEMP1 expressor lines, for instance, to study trafficking and binding of PfEMP1.

Next, we assessed whether SLI could be used to obtain proxiomes (proximal proteins and interactors) of PfEMP1 in living parasites by generating parasites expressing a PfEMP1 fused with the promiscuous biotin ligase BirA* to carry out BioID (Roux et al., 2012). For this, we chose IT4var01 and generated three SLI cell lines with BirA* in different positions of that PfEMP1 (IT4var01-BirA*Pos1^endo, –2^endo and –3^endo) (Figure 6A-D, Figure 6—figure supplement 1). In position 1, BirA* was C-terminal to the ATS, in position 2 between transmembrane domain and ATS, and in position 3 directly upstream of the transmembrane domain (Figure 6A–D; note that IT4var01-BirA*Pos3 lacked the intron (Data S1)). The resulting cell lines predominantly expressed the modified PfEMP1 (as judged by IFA and RNA-Seq), the PfEMP1 was on the surface and the infected RBCs showed the expected binding pattern (Figure 6B–F). Parasites expressing the position 3 PfEMP1 construct showed less binding, suggesting partial impairment of placing BirA* into the extracellular part of PfEMP1. Nonetheless, overall the BirA* modified PfEMP1 was functional (Figure 6F). BirA* in the PfEMP1 was active as judged by streptavidin blots which showed biotinylation with all three cell lines but not with the IT4 parent (Figure 6G). Next, we carried out BioID experiments with these cell lines, analyzing enrichment of biotinylated proteins over IT4 in two sequential protein extracts: first the proteins extractable by mild detergents (Figure 6—figure supplement 1, Figure 6—source data 6) and the proteins requiring extraction with SDS to release more structurally connected proteins (Figure 6H-J, Figure 6—figure supplement 1, Figure 6—source data 6). In all experiments, the tagged PfEMP1 (but no other PfEMP1) was highly enriched due to self-biotinylation. Comparably few proteins were enriched in the mild detergent fraction (Figure 6—figure supplement 1), but the SDS-fraction contained many proteins known to be important for PfEMP1 trafficking, including for instance SBP1, MAHRP1, REX1, and several PTPs (1, 2, 4, 5, 7), indicating efficient detection of PfEMP1 trafficking factors (Figure 6H-J, Figure 6—figure supplement 1). In addition, other exported proteins were detected. This, for instance, included proteins of the MSRP6 complex found at the Maurer’s clefts, which is involved in anchoring the clefts but has no role in PfEMP1 transport (Blancke Soares et al., 2025), several PHISTs (Sargeant et al., 2006), and exported proteins with unknown function here termed EMP1 interacting candidate 1–6 (EMPIC1-6; Figure 6H-J, Figure 6—source data 6).

Interestingly, the repertoire and relative enrichment of the proteins detected in the BioIDs with the three constructs was remarkably similar (Figure 6H-J, Figure 6—figure supplement 2A, Figure 6—source data 6), including position 3 where BirA* is located on the C-terminal side of the transmembrane domain. This supports the hypothesis that PfEMP1 is not transported as an integral membrane protein (Batinovic et al., 2017; Marti and Spielmann, 2013; Papakrivos et al., 2005; Petersen et al., 2016) as BirA* in the N-terminal part appears to have had access to biotinylate the same proteins as BirA* in the C-terminal part. While there was little evidence for topology-specific interactors, several of the detected PHISTs (PfIT_120058000, PfIT_040006400; PfIT_130076100) as well as GEXP10 and the less efficiently enriched PTEF are known to be RBC membrane localized (Birnbaum et al., 2017; Chan et al., 2017; Dantzler et al., 2019; Davies et al., 2023; Hermand et al., 2016; Tarr et al., 2014), indicating that the proxiome also included hits from surface exposed PfEMP1. The hits obtained with the 3 PfEMP1 BirA*-fusion constructs were also more similar to each other than to a general Maurer’s clefts proxiome or that of the MSRP6 Maurer’s clefts binding domain (Blancke Soares et al., 2025), suggesting specificity for PfEMP1 (Figure 6—figure supplement 2B). Further, hits from other structures than Maurer’s clefts, such as the tether protein MAHRP2 (Pachlatko et al., 2010) and the known (Külzer et al., 2012; Zhang et al., 2017) and likely (Schulze et al., 2015) J-dot proteins HSP70x, GEXP18, and PHIST P2 (PfIT_120006500; Figure 6—source data 6) further supported that the PfEMP1 proxiome covered hits beyond its dominant location at the Maurer’s clefts.

One notable difference between the hits of the 3 PfEMP1 BirA*-fusion constructs was that the position 1 construct detected 7 PHISTs, whereas the position 2 and 3 constructs only detected 3 and 2 PHISTs, of which one was not detected with position 1 (Figure 6H-J, Figure 6—figure supplement 2A, Figure 6—source data 6). In addition, some of the PTPs (PTP1 and PTP7) appeared to be differentially enriched in the 3 positions. Finally, position 1 detected the largest number of enriched proteins, possibly because the larger distance from the transmembrane domain permitted more efficient labeling or because this part of the construct is in proximity to a larger number of proteins.

Taken together, these experiments detected most of the proteins previously implicated with PfEMP1 transport and surface display, indicating these proxiomes give a valid representation of proteins in contact with PfEMP1.

We selected several proteins from the PfEMP1 proxiomes (Figure 6H-J, Figure 6—source data 6) that previously had not been connected with PfEMP1 transport and used SLI2 to generate full-length tagged versions (Figure 7—figure supplement 1A) as well as disruptions (Figure 7—figure supplement 1B) by targeting the corresponding genes in the IT4var01-BirA*Pos1^endo parasites to assess whether they could be needed for PfEMP1-mediated cytoadhesion. This included PfIT_020007200 that we had previously identified as a PEXEL negative exported protein (PNEP) exported to the host RBC periphery (Birnbaum et al., 2017) and that was implicated in VAR2CSA translation and named P. falciparum translation enhancing factor (PTEF) (Chan et al., 2017). We also included TryThrA (PfIT_080035200), a PNEP (Heiber et al., 2013) that was a prominent hit in the BioIDs with all positions, including in the Triton fraction as well as EMPIC3 (PfIT_070007400), also a PNEP (Heiber et al., 2013) detected with all 3 PfEMP1 BioID constructs with intermediate to high enrichment (Figure 6—source data 6; Figure 7A). In addition, we included PeMP2 (PfIT_050006400), a member of the MSRP6 complex (Blancke Soares et al., 2025), not previously tested for its function in PfEMP1-mediated cytoadhesion.

Full-length endogenously Ty1-tagged TryThrA, EMPIC3, and PeMP2 showed IFA patterns consistent with the reported localizations of these proteins in the host cell (Figure 7—figure supplement 1A; Birnbaum et al., 2017; Blancke Soares et al., 2025; Heiber et al., 2013). For TryThrA-Ty^endo and EMPIC3-Ty^endo, we observed cells where the Ty1 signal appeared as circles that likely corresponded to the Maurer’s clefts periphery and only partially overlapped with the HA signal of PfEMP1 (Figure 7—figure supplement 1A). This staining pattern is reminiscent of the subcompartmentalization of different proteins at the Maurer’s clefts previously observed by superresolution microscopy (McMillan et al., 2013).

Next, we examined the parasites wherein the selected candidates had been disrupted. While the SLI2-based disruptions of PTEF and PeMP2 did not result in loss of parasite binding to CD36 and ICAM1, the TryThrA- and EMPIC3-TGDs resulted in markedly reduced binding (Figure 7B), indicating TryThrA and EMPIC3 are novel proteins needed for cytoadhesion. Interestingly, the TryThrA-TGD parasites showed atypical localization of PfEMP1, SBP1, REX1 (dispersed signal in the host cell in addition to foci; disproportionally strong foci) but not KAHRP, pointing to a defect of the Maurer’s clefts or protein transport to these structures, while in the EMPIC3-TGD parasites PfEMP1, SBP1, and REX1 showed a localization typical for Maurer’s clefts with clearly defined foci and absence of a dispersed pool (Figure 7—figure supplement 1B).

In order to better define the phenotype in the TryThrA-TGD parasites, we transfected this and the EMPIC3-TGD parasite line with an episomal plasmid mediating expression of mCherry tagged SBP1, permitting analysis of live, unfixed parasites. As we had already used four selection markers to generate these parasites, we employed a plasmid encoding a mutated version of the lactate transporter FNT (Wu et al., 2015, Marchetti et al., 2015, Golldack et al., 2017) that confers resistance to the chemical BH267.meta (Walloch et al., 2020), to transfect these parasites and episomally express SBP1-mCherry (Figure 7—figure supplement 2A). While in the EMPIC3-TGD parasites, SBP1-mCherry was found in foci in the host cell typical for Maurer’s clefts (Figure 7—figure supplement 2B), the majority of the TryThrA-TGD parasites showed an additional soluble pool of SBP1-mCherry in the host cell (Figure 7—figure supplement 2C). Some TryThrA-TGD parasites also contained foci of increased intensity, suggesting enlarged or aggregated Maurer’s clefts (Figure 7—figure supplement 2C). As the IFAs showed a similar phenotype for REX1 and PfEMP1, the TryThrA-TGD parasites appear to have a defect in the localization of multiple Maurer’s clefts proteins and possibly the morphology of the Maurer’s clefts. These changes likely are responsible for the cytoadhesion defect in the TryThrA-TGD. In contrast, no obvious changes explaining the binding phenotype were detected in the EMPIC3-TGD parasites.

To ensure the cytoadhesion defect was not due to unrelated changes that had occurred during generation of the TGDs, we sequenced the genome of the TryThrA and EMPIC3 TGD lines. No major changes compared to the cytoadherent IT4var01-BirA*Pos1^endo parent were detected that would explain the cytoadhesion defect (Source data 2). Surface trypsin treatment assays showed that in both cases, some PfEMP1 was still surface exposed (Figure 7C), indicating that transport to the surface was either merely reduced or that an impairment of the correct presentation of PfEMP1 on the surface caused the binding defect. To ensure this was not due to limitations of the trypsin assay, we also disrupted PTP7, which also was a prominent hit in our BioIDs (Figure 6I) and is a well-characterized PfEMP1 trafficking protein that results in loss of surface transport when disrupted (Carmo et al., 2022). We confirmed the cytoadhesion defect in the parasite with a disrupted PTP7 (Figure 7B), and in contrast to the TryThrA and EMPIC3 TGD cell lines, trypsin assays indicated that there is no IT4VAR01-HA on the surface in that cell line (Figure 7C).

In conclusion, we identified two novel proteins (TryThrA and EMPIC3) needed for PfEMP1 function. Given that many of the known PfEMP1 trafficking proteins were detected in the proxiomes and testing some of the others revealed more such proteins, we assume that the BioID experiments give a relevant representation of the protein environment of PfEMP1 and likely contain further proteins important for cytoadhesion.

For genome integration constructs, homology regions encoding the C-terminus of target genes (for C-terminal tagging) or a region in the N-terminal part (for TGDs) were PCR amplified from 3D7 or IT4 gDNA purified with the Monarch gDNA Purification Kit, NEB (T3010), or QIAGEN DNA extraction kit (56304) and cloned into pSLI (Birnbaum et al., 2017) or pSLI2 using Gibson assembly (Gibson et al., 2009) or T4 ligase. Plasmids, including the SLI2 plasmids and the FNT resistance plasmid for episomal expression of SBP1-mCherry, are shown in Supplementary file 1. For the Position 1 BirA* fusion plasmids, the targeting region followed by a 7xGGGS linker, a previously used sequence encoding BirA* (Birnbaum et al., 2020) and a 3xHA-tag was cloned into pSLI. For position 2 and position 3 plasmids, the part of PfEMP1 encoded C-terminal to BirA* was synthesized with a different codon usage (GenScript) to prevent integration into the genome in that region and was cloned together with the targeting region (Position 2: until amino acid 2415; Position 3: until amino acid 2376), the region encoding BirA* flanked by short linkers and a 3xHA-tag into pSLI. For the episomal early stage blocking construct SBP1-mDHFR-GFP was cloned into pARL2 containing a mal7 promoter (Grüring et al., 2011). For the tagging of PfEMP1 with mDHFR, homology regions encoding the C-terminus of the target var genes were cloned into pSLI with a mDHFR domain between the targeting region and a 3xHA-tag. To ensure there were no undesired mutations, all cloned inserts were sequenced by Sanger sequencing (Microsynth).

P. falciparum parasites (3D7 Walliker et al., 1987 and IT4 Jensen and Trager, 1978) were cultured using standard procedures (Trager and Jensen, 1976). The parasites were maintained in RPMI1640 supplemented with 0.5% Albumax (Life Technologies, 11021) and human 0+ erythrocytes (University Medical Center Hamburg-Eppendorf (UKE), Germany) at a hematocrit of 5% at 37 °C under an atmosphere consisting of 1% O₂, 5% CO₂, and 94% N₂.

Late schizont stage parasites were purified using Percoll as described (Rivadeneira et al., 1983), using 60% Percoll for 3D7 and 64% for IT4 parasites. Fifty μg of purified plasmid DNA (QIAGEN, 12143) were transfected using the Amaxa system (Lonza Nucleofector II AAD-1001N, program U-033) following previously described protocols (Moon et al., 2013). Transfectants were selected with either 4 nM WR99210 (Jacobus Pharmaceuticals; pSLI) or 2 μg/ml blasticidin S (Life Technologies, R21001; pSLI2). SLI for selection of parasites with the plasmid integrated into the genome was done as described (Birnbaum et al., 2017) by adding 400 μg/ml G418 (Sigma Aldrich, A1720; pSLI) or 0.9 μM DSM1 (Merck, 5.33304.0001; pSLI2) to the culture. After the parasitemia recovered under drug selection, genomic DNA was isolated and correct genomic integration of the plasmid in the knock-in parasites was verified by PCR as described (Birnbaum et al., 2017). For transfection of the plasmid harboring the mutated gene encoding PfFNT (amino acid change G 107 S; Golldack et al., 2017), BH267.meta (Walloch et al., 2020) was used at 5 µM until parasites appeared, after which the concentration of drug was dropped to 2.5 µM to maintain the culture.

IFAs were performed as described (Spielmann et al., 2003). Briefly, pelleted parasites (2000 × g for 5 min) were washed with 1 x PBS and applied at a hematocrit of 1–2.5% to 10-well glass slides, air-dried, and fixed in acetone for 30 min. Wells were rehydrated with 1 x PBS, then washed five times with 1 x PBS. Antibodies were applied in 1 x PBS containing 3% BSA. Primary antibodies were rat anti-HA (Roche, 11867423001), 1:2000; rabbit anti-HA (Cell Signaling Technology, 3724), 1:1000; rabbit anti-SBP1-C (Mesén-Ramírez et al., 2016), 1:2500; rabbit anti-KAHRP (kind gift of Prof. Brian Cooke), 1:500; mouse anti-EXP2 (European Malaria Reagent Repository) used 1:500; rabbit anti-REX1 (Mesén-Ramírez et al., 2016), 1:10,000; mouse anti-Ty1 (Thermo, MA5-23513), 1:20,000; rat anti-RFP (Chromotek, 5f8 – 100), 1:1000; mouse anti-GFP (Roche, 11814460001) used 1:1000 and rabbit anti-GFP (Thermo, A-6455), 1:1000. For secondary antibodies anti-rabbit conjugated with Alexa Fluor-488 (A-27934), Alexa-Fluor546 (A-10040), or Alexa Fluor-647 (A-21244), anti-mouse conjugated with Alexa Fluor-488 (A-11001), goat anti-rat conjugated with Alexa Fluor-488 (A-10041), or Alexa Fluor-594 (A-11007) (all Invitrogen) were used (all 1:2000). Secondary antibodies were applied together with 4’,6’-diamidine-2’-phenylindole dihydrochloride (DAPI, 10236276001; 1 µg/ml) or Hoechst (50 ng/ml; Cayman, K9061; as indicated in the figure legends) for staining of parasite nuclei. For the Streptavidin-fluorescence assay, streptavidin coupled to Alexa Fluor-594 (Invitrogen, S32356) was added (1:2000) together with the secondary antibody. Slides were mounted with Dako (Sigma Aldrich, S3023) and covered with a cover slip.

Live or fixed parasites were imaged with a Zeiss AxioImager M1 or M2 equipped with a Hamamatsu Orca C4742-95 camera using a 100×/1.4-numerical or a 63×/1.4-numerical aperture lens. AxioVision software (version 4.7) was employed to capture the images. Live cell imaging of parasites expressing fluorescent proteins was performed as previously described (Grüring and Spielmann, 2012). To stain the parasites' DNA, parasites were incubated with either 1 μg/ml of DAPI (Roche, 10236276001) or 50 ng/ml Hoechst 33342 (Cayman, K9061) (as indicated in the figure legends) in parasite medium for 10 min at 37 °C. Images were processed in Corel Photo-Paint (version 2021) and arranged in Corel Draw (version 2021).

Parasite cultures with 5–10% parasitemia were synchronized for rings using sorbitol (Lambros and Vanderberg, 1979) and then grown for 12 hr at 37 °C. The resulting trophozoite stage parasites were isolated with a Percoll gradient as described (Heiber and Spielmann, 2014) for 3D7 cell lines. For IT4 parasites, an adjusted gradient with 80%, 64%, and 40% Percoll was used. The purified infected erythrocytes were washed and split into two samples. One sample was incubated with 50 µg/ml TPCK-treated Trypsin (Sigma Aldrich, 4352157) in 1 x PBS at 37 °C for 30 min while the other sample (control) was incubated in 1 x PBS alone. Thereafter, trypsin inhibitor from soybean (Sigma Aldrich, 10109886001) was added (1 mg/ml final concentration), and the samples were incubated on ice for 15 min. The cells were washed in 1 x PBS, then lysed in 100 µl lysis buffer (4% SDS, 0.5% Triton X-100 in 0.5 x PBS), containing 1 mg/ml trypsin inhibitor, 1 mM PMSF (Thermo Fisher Scientific, 36978), and 1 x complete protease inhibitor cocktail (Roche, 11697498001). Extracts were immediately subjected to SDS-PAGE or frozen at –20 °C until needed.

For binding assays, Chinese Hamster Ovary (CHO-745 or CHO-K1) cells that express CD36, ICAM-1, GFP ⁶⁶ or EPCR (in CHO-K1) (Avril et al., 2016), or human brain endothelial cells HBEC-5i cells (American Type Culture Collection (ATCC), Manassas, VA, USA; no. CRL-3245) were seeded two (1x10⁵ cells/ml) or three (2x10⁵ cells/ml) days before the binding assay into a 24-well plate containing coverslips (0.5 ml/well). For binding assays against decorin (chondroitin sulfate proteoglycan from bovine articular cartilage previously used for VAR2CSA binding Dahlbäck et al., 2011), the coverslips in 24-well plates were incubated overnight at 4 °C with decorin solution (5 µg/ml in PBS), thereafter washed with 1 x PBS, blocked with 1% BSA in 1 x PBS for 2 hr and washed with 1 x PBS twice (Renn et al., 2021). Knobby parasites of the tested cell lines were enriched using 1% gelatin in glucose-free RPMI (16.4 g/l RPMI-HEPES (Applichem, A1538), 0.05 g/l hypoxanthine, 30 ml/l NaHCO₃ (7.5 %) and 250 µl/l gentamycin (Ratiopharm, 3928180) in H₂O, pH 7.2) as described (Goodyer et al., 1994). After washing in binding medium (16.4 g/l RPMI-HEPES and 20 g/l glucose in H₂O, pH 7.2), number of erythrocytes/ml (Neubauer counting chamber) and % infected erythrocytes (Giemsa smears) were determined and the suspension adjusted to 2x10⁶ infected erythrocytes/ml in binding medium. The wells with the CHO or HBEC-5i cells were incubated with binding medium for 30 min before the parasite suspension was added to the wells (500 µl/well). Per experiment, three wells per parasite cell line and receptor were used. In the binding assays with decorin and HBEC-5i, the parasites' suspension was split and either incubated with soluble CSA (100 µg/ml) or soluble BSA (100 µg/ml) (control) for 30 min at 37 °C before adding the infected erythrocytes to the wells. The plates were then incubated for 60 min at 37 °C for binding, with careful shaking every 15 min. The coverslips were washed six times by carefully dunking them into binding medium and blotting excess medium on paper after every dunk. The coverslips were then laid face-down parallel to the table in a washing plate that was angled at 45° (with the face-side hanging free in the binding medium) and incubated for 30 min at room temperature. Immediately after, the coverslips were fixed in 1% glutaraldehyde in 1 x PBS for 30 min and stained with filtered 10% Giemsa (Merck, 1092040500) in 1 x PBS for 15 min. The stained coverslips were washed in water and glued with CV-Mount (Leica, 14046430011) face-down onto glass slides. Five images per coverslip (per experiment 15 images per parasite line and condition) were captured with a Thermo Fisher EVOS xl (75% light intensity at ×40 magnification).

The evaluation of images of binding assays was automated using Ilastik v1.3.3post3 (Berg et al., 2019) and CellProfiler v4.2.1 (Stirling et al., 2021; Figure 3—figure supplement 1). First, the images of the binding assays were processed with a trained Ilastik model for the segmentation of the foreground (infected erythrocytes) and background (CHO/HBEC-5i cells and plastic). For the training, the pixel classification module was manually trained with 20 microscopy images representing different shapes of infected erythrocytes, backgrounds, and artefacts. All the color/intensity, edge, and texture features were enabled for training. The resulting processed images were exported as probability images with pixel intensities from 0.0 to 1.0 for the probability of a foreground pixel (regression values; 1.0=100% probability for foreground pixel). Ilastik pre-processed images were then fed to a CellProfiler pipeline (Figure 3—figure supplement 1) using the ‘IdentifyPrimaryObjects’ module to identify and count roundish objects with a diameter of 15–35 pixel units. Robust background thresholding and de-clumping by shape was selected. The number of counted infected erythrocytes scored per image was given out as a spreadsheet. To show the reliability of the automated pipeline in comparison to the manual scoring, statistical tests between the two methods were conducted as shown in Figure 3—figure supplement 1.

Synchronous ring-stage parasites with a parasitemia of 3–5%, from 10 ml of culture were pelleted at 800  × g and dissolved in five pellet volumes of Trizol (Thermo Fisher, 15596018), thoroughly mixed, incubated for 5 min at 37 °C and immediately stored at –80 °C until RNA isolation. To purify the RNA, the Trizol sample was thawed, 1/5 volume of chloroform added, thoroughly mixed, and centrifuged at 16,000  × g for 30  min at 4 °C. The resulting clear supernatant was transferred to a new tube and processed using the Qiagen miRNeasy Mini Kit (217004) according to the manufacturer’s instructions. RNA integrity was assessed using the Agilent 2100 bioanalyzer system with the RNA 6000 Pico Kit (Agilent, 5067–1513). All samples had a RIN >8, our cutoff for inclusion.

Ribosomal RNA was removed using QIAseq FastSelect RNA Removal Kit (QIAGEN, 333390). Libraries were prepared with the QIASeq Stranded mRNA Library Kit (QIAGEN, 180440) and sequenced on an Illumina NextSeq 550 system with NextSeq 500/550 Mid Output Kit v2.5 (Illumina, 20024906; 150 cycles). Raw reads were mapped with hisat2 (version 2.2.1) to the respective reference genomes sourced from PlasmoDB (Amos et al., 2022; IT4: Release 58; 3D7: Release 62). Mapped reads were sorted and indexed with samtools (version 1.17). Reads mapped to genomic features were counted using featureCounts (version 2.0.4). For var genes, only reads mapping to exon 1 were considered; for rifs, reads to the entire coding region were included. The data have been deposited in NCBI’s Gene Expression Omnibus (Edgar et al., 2002) and are accessible through GEO Series accession number GSE267413. Python3 (version 3.11.4) and bioinfokit (version 2.1.2) were used to normalize the reads to transcripts per million (TPM) as well as to create the coverage plots with matplotlib (version 3.7.2). A volcano plot was done in GraphPad Prism. Differential gene expression analysis for panned against unpanned parasites was performed in R with the DESeq2 (version 1.42.0) package.

Quantification of var and rif transcript levels was measured relative to internal control gene seryl-tRNA synthetase by real-time quantitative PCR using primers specific to each 3D7 var or rif gene as previously described (Wang et al., 2009).

Assays assessing the transport of PfEMP1 fused directly to mDHFR (Eilers and Schatz, 1986) were done as described (Grüring et al., 2012) with some modifications: schizont stages of the corresponding lines were purified with Percoll and allowed to invade for 8 hr, followed by synchronization with 5% sorbitol (Lambros and Vanderberg, 1979) to obtain ring stages with an age of 0–8 hr post invasion, the culture split into one with 4 nM WR and one without (control) and grown for 24 hr before analysis by IFA. For co-blocking assays (Mesén-Ramírez et al., 2016), where transport through PTEX was assessed indirectly by conditionally clogging it with another exported protein fused to mDHFR, the parasite cultures were synchronized using Percoll to obtain schizonts as described (Rivadeneira et al., 1983) and grown for 24 hr in the presence or absence of 4 nM WR followed by analysis of export by live cell imaging or IFA. For the late stage PTEX block, the pARL2-SBP1-mDHFR-GFP-2A-KAHRP-mScarlet plasmid was utilized (Mesén-Ramírez et al., 2016).

For proximity biotinylation, biotin (Sigma Aldrich, B4639; 50 µM final) was added to asynchronous parasites expressing the BirA*-PfEMP1 fusion constructs as well as to IT4 parent parasites (5% parasitemia, 150 ml per condition and experiment) and cultured for 24 hr with one exchange of medium with fresh biotin after 12 hr. Thereafter, the parasites were washed twice with DPBS before they were subjected to saponin lysis (0.03% saponin in DPBS) on ice for 10 min, followed by five washes in DPBS before lysis in 2 ml lysis buffer (50 mM Tris-HCl pH 7.5, 500 mM NaCl, 1% Triton-X-100, 1 mM DTT, 1 mM PMSF and 1 x protease inhibitor cocktail) and storage at –80 °C. For isolation of proteins, the samples were thawed and frozen two times before centrifugation at 16,000 × g for 10 min. The supernatant (Triton-extract) was saved and the pellet frozen, thawed, and once more extracted using 4% SDS in 50 mM Tris-HCl pH 7.5, 500 mM NaCl, 1% Triton-X-100, 1 mM DTT (SDS-extract). The SDS extract was transferred to a fresh tube and cleared by centrifugation at 16,000 × g for 10 min. For the purification of biotinylated proteins, both extracts (Triton and SDS) were diluted 2:1 in 50 mM Tris-HCl and incubated with 50 µl Streptavidin Sepharose (GE Healthcare, 17-5113-01) overnight at 4 °C while rotating. The beads were washed twice in lysis buffer, once in H₂O, twice in Tris-HCl pH 7.5, and three times in 100 mM Triethylammonium bicarbonate buffer. The proteins on the beads were digested as described (Hubner et al., 2015). Briefly, the beads were treated with 50 μl elution buffer (2 M Urea in 100 mM Tris pH 7.5 containing 10 mM DTT) at room temperature, shaking for 20 min. Subsequently, iodoacetamide (IAA) was added to a final concentration of 50 mM and the samples were further incubated in the dark, shaking for 10 min. The proteins were then treated with 0.25 μg Trypsin/LysC (Promega, V5072), while shaking at room temperature. After 2 hr, the supernatants containing eluted proteins were collected and the beads were immersed with an extra 50 μl of elution buffer for 5 min at room temperature. The supernatant was pooled with the previous elution, and the final 100 μl of eluted proteins was supplemented with 0.1 μg of Trypsin/LysC and treated overnight while shaking at room temperature. The protein samples were then desalted on Stagetips using C18 membranes (Rappsilber et al., 2007) and eluted in 80% acetonitrile, 0.1% Formic acid.

The acetonitrile was evaporated in a SpeedVac, and the concentrated sample was then reconstituted to a final volume of 12 μl with 0.1% Formic acid. To analyze the sample by mass spectrometry, 5 μl of sample was analyzed during a 60 min run on an Easy-nLC 1000 (Thermo Fisher Scientific) with a 30 cm C18-reverse phase column coupled on-line to an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific). Data was acquired in top 20 mode with a dynamic exclusion of 45 s.

Raw mass spectrometry data were processed using MaxQuant (Cox and Mann, 2008; version 1.6.6.0). Parameters were set to default except for the following: Deamidation (NQ) was added as a variable modification together with oxidation (M) and acetyl (N-term). Match-between-runs and re-quantify options were enabled with default parameters and iBAQ values were calculated. Mass spectra were compared to peptide masses from the Plasmodium falciparum IT4 annotated proteome (PlasmoDB v64). The ‘proteinGroups’ file from MaxQuant output was analyzed using the Perseus software package (Tyanova et al., 2016; version 1.4.0.20). The data were filtered against peptides assigned as ‘only identified by site’, ‘reverse’ and/or ‘potential contaminant’ hits in the datasets. IBAQ values were transformed to log₂ values and missing values were imputed following a normal distribution. Data obtained from Triton extraction and SDS extraction were analyzed separately. Significant outliers were identified at each position by using the two-sided Benjamini-Hochberg test with an FDR cut-off of 0.05. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifier PXD052297.

Western blots were conducted as described (Heiber and Spielmann, 2014). In brief, preparation of extracts from the BioID experiments or the trypsin cleavage assays was centrifuged at 16,000 × g and the supernatant was mixed with 4 x Laemmli sample buffer. Samples were incubated for 10 min at 90 °C before they were applied to 10% polyacrylamide gels for sodium dodecyl sulfate polyacrylamide gel electrophoresis. The proteins separated on the gels were transferred to nitrocellulose membranes (Amershan Protran membranes, GE Healthcare, GE10600002) using transfer buffer (0.192 M Glycine, 0.1% SDS, 25 mM Tris and 20% methanol in H₂O). For the detection of proteins by antibodies, membranes were blocked in 5% skim milk in 1 x TBS (50 mM Tris and 150 mM NaCl in H₂O) for 2 hr at room temperature, washed three times with 1 x TBS with 1% Tween, and incubated in 1 x TBS with 3% skim milk with the first antibody rolling overnight at 4 °C. First antibodies were rat anti-HA (Roche, 11867423001; 1:1000); rabbit anti-SBP1-N (1:4000; Mesén-Ramírez et al., 2016) or rabbit anti-aldolase (1:4000; Mesén-Ramírez et al., 2016). Secondary antibodies were horseradish peroxidase (HRP)-conjugated anti-rat (Dianova, 112035003; 1:2000) or HRP-conjugated anti-rabbit (Dianova, GtxRb-003-DHRPX; 1:2000) and were applied in 1 x TBS with 3% skim milk and incubated rolling for 2 hr at room temperature. For the detection of biotinylated proteins, HRP-conjugated streptavidin (Thermo Fisher Scientific, A-11001) was used in 5% BSA in 1 x TBS as described (Cui and Ma, 2018) and incubated by rolling overnight at 4 °C. After secondary antibody or HRP-conjugated streptavidin incubation, the membrane was washed three times in 1 x TBS with 1% Tween, then 5 ml ECL solution A (0.025% luminol (Sigma Aldrich, A8511) in 0.1 M Tris-HCl in H₂O, pH 8.6) was mixed with 500 µl ECL solution B (6.7 mM p-Coumaric acid in DMSO) and 1.5 µl H₂O₂ and applied to the membrane before the ECL signal was detected with a ChemiDoc XRS imaging system (Bio-Rad).

Genome sequencing was done essentially as described (Behrens et al., 2024). The NEB Monarch Genomic DNA Purification Kit (T3010) was used to prepare genomic DNA from 50 ml cultures of the TryThrA and EMPIC3 TGD parasites (both generated in the IT4var01-BirA*Pos1^endo background) and from the parent (IT4var01-BirA*Pos1^endo). BGI TECH SOLUTIONS (Hong Kong) carried out DNBSEQ PE100 sequencing and bioinformatic analysis. This included calling of SNP, InDel, SV, and CNV compared to IT4 reference. The data was deposited at GEO (Accession number GSE275671) which also includes technical details on sample preparation and filtering. All SNPs leading to a stop or potential splice mistake, all INDELs leading to frame shifts, all SVs and CNVs indicating gene or partial gene loss in the Var01-TGD parasites that were not present in the parent (IT4-Var01 parasites) were manually assessed by inspecting the reads in that region. Only changes affecting exported proteins were considered and were manually re-assessed in all three lines by analyzing the individual reads. In addition, known PfEMP1 trafficking genes were manually checked for differences.

P values are indicated in the figure and p<0.05 was considered as significant. All error bars shown are standard deviations. Statistical significance was determined by unpaired t-test. A ratio-paired t-test was used for the comparison between the individual images of the binding assays evaluated by manual scoring and the automated pipeline. Statistical analysis was done in GraphPad Prism (version 9). Intraclass correlation coefficient (ICC) was calculated using Excel (Microsoft); Two-factor ANOVA without replication was applied; ICC was calculated with the variations of the ANOVA; ICC = MS_Row-MS_Error/MS_Row +df_ColumnxMS_Error + (df_Column +1)x(MS_Column-MS_Error)/(df_Row +1). Graphs were done in GraphPad (version 9) and transferred to CorelDraw (version 2021) with adjustments to style without altering the data. Corel Draw (version 2021) was used to prepare the figures.