The attachment phenotype of genetically similar merozoites from P. falciparum 3D7 and NF54 strain are significantly different

We used optical tweezers to quantitate attachment interactions between newly egressed P. falciparum merozoites and erythrocytes, as outlined in more detail in Methods and first established in [31]. In brief, purified schizonts were monitored using video microscopy to identify schizonts in the process of egress; then optical traps were used to manipulate a newly egressed merozoite and position it between two erythrocytes. After positioning, the cells were left for a period of between 2–5 s to allow for attachment to occur, and then one erythrocyte was pulled with the optical trap, S1 Video. As the erythrocyte is pulled, it stretches until the pulling force overcomes the strength of attachment between merozoite and erythrocyte and the cell-cell connection is broken. The stiffness constant of erythrocytes is well-characterised [37], meaning that the cells themselves can be treated like force meters, i.e. the force that was exerted to detach the cells can be estimated from the extent to which the erythrocyte stretches at the point of detachment and hence the strength of the attachment between merozoite and erythrocyte can be calculated Fig 1A. The frequency of attachment was also determined by dividing a) the number of merozoites that successfully attached (defined as an attachment strong enough to cause a visual deformation of the erythrocyte when pulled with the optical trap) after positioning between two erythrocytes by the b) the total number of times merozoites were positioned between two erythrocytes from each egressed schizont. This frequency was then averaged across all egresses for a given strain or condition to create a mean attachment frequency Fig 1B. All measurements were taken within 180 s of egress to maintain a balance between experimental feasibility and merozoite viability; previous studies with video microscopy have shown that the average time between egress and invasion is 38 s [4] and it has been shown that very few invasions occur after 180 s post-egress [5].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Optical tweezers show that genetically similar wild-type P. falciparum NF54 and 3D7 merozoites have different attachment strengths.

(a) Video-microscopy images from the optical tweezer assay used throughout this work, with a cartoon representation underneath. The orange arrows show the direction of the forces applied to the cells with the optical traps represented as a star. ON/OFF indicates if the optical traps are active during the step. During manipulation, one of the erythrocytes was often attached to the slide, in which case only one erythrocyte was pulled, as in these images. If neither erythrocyte was attached to the slide, then one erythrocyte was held still with an optical trap, and a second trap was used to pull the other erythrocyte. To calculate the force of attachment, the length of the erythrocyte is measured before it is moved (L₀), and the maximal stretch of the erythrocyte is measured in the frame immediately before detachment occurs (L_max). The change in length is calculated as ΔL = L_max-L₀. The force of detachment is calculated based on F = kΔL [31]. The erythrocyte stiffness constant (k) is assumed to be 20 pN/μm [37]. Scale bar = 5 μm. (b) Cartoon to show how the attachment frequency is measured, as described in the text. (c-e) Data sets that were significantly different (t-test) have a line between them with stars showing the level of significance *** = ≤0.001. (c) Box plot showing the detachment force of merozoite-erythrocyte attachments. NF54 –mean detachment force 24 ± 2 pN (n = 85). 3D7 –mean detachment force = 36 ± 3 pN (n = 23). The difference between strains is significant (p = 0.0007, t-test); both data tests are normally distributed (Anderson-Darling test). The central bold line shows the median, with the top and bottom of the box at the 25^th and 75^th percentiles and the whiskers showing the total range of the data. At least three biological repeats were completed for both data sets with blood from two different donors. The number (n) of individual interactions measured for each strain is shown under the plot. The attachment behaviour of 3D7 has been previously measured in our lab [35,36]. The detachment force measured previously for 3D7 (43 ± 3 pN), [35], was not significantly different to that measured here (t-test p = 0.16). (d) The bar chart shows the frequency of erythrocyte-merozoite-erythrocyte positions that lead to attachment of the merozoite to both erythrocytes. NF54 mean 18.6 ± 2.6% (552 positions) and 3D7 31 ± 9% (92 positions). The number of individual interactions measured for each line is shown above each bar. Error bars show the standard error of the mean (SEM). (e) Box plot showing the growth rate measured in static culture represented as the Parasite Erythrocyte Multiplication rate (PEMR), calculated by dividing the parasitemia of the ring stage culture by the parasitemia of the same culture 24 h previously. PEMR was averaged over 6 invasion cycles for three technical repeats per cycle, more details provided in Methods. NF54 mean PEMR = 4.1 ± 0.3 and 3D7 mean PEMR = 8.7 ± 0.8. 3D7 and NF54 were significantly different (p = ≤ 0.0000, t-test). The central bold line shows the median, with the top and bottom of the box at the 25^th and 75^th percentiles and the whiskers showing the total range of the data.

https://doi.org/10.1371/journal.ppat.1012041.g001

The majority of the experiments in this study were performed using P. falciparum strains NF54 or 3D7, or genetically modified lines made using one of these strains as background, so we first set out to establish the attachment interactions of these wildtype lines. NF54 originates from a patient in the Netherlands, and genomic analysis suggests is likely of West African origin [38]. 3D7 is a clone of NF54 that was used for the P. falciparum genome project [39] but the two strains have been cultured separately for over 30 years, meaning they have diverged in some aspects of their biology [40–42] while remaining genetically very similar [43]. The optical tweezer assay established that NF54 merozoites had a mean detachment force of 24 ± 2 pN, Fig 1C and an attachment frequency per egress of 19 ± 3%, Fig 1D (mean ± standard error of the mean is used throughout). By contrast, 3D7 merozoites had a detachment force of 36 ± 3 pN and an attachment frequency of 31 ± 9%. The detachment force of 3D7 was significantly higher than NF54 (t-test p = 0.0007), whereas the attachment frequency was not significantly different (t-test p = 0.13). Given the differences in attachment strength between these two closely related strains, we were interested in whether their growth rates were also different. We set up technical triplicates with tightly synchronised parasites from both lines and measured their growth rates over 6-invasion cycles, see Methods for details. The growth rate was calculated by dividing the parasitemia of the newly invaded rings over the parasitemia measured the day before, giving the parasite erythrocyte multiplication rate (PEMR) as defined by [44]. The PEMR of 3D7 was 2.1 times higher than NF54 (p = ≤0.0000), with NF54 mean PEMR = 4.1 ± 0.3 and 3D7 mean PEMR = 8.7 ± 0.8, Fig 1E, reinforcing the fact that there are biological differences between these strains, despite their genetic similarity.

As Fig 1C emphasises, there is considerable variability in detachment force between individual merozoite-erythrocyte interactions, even within the same strain. We explored several sources of potential variation within the assay, S1–S3 Figs. Firstly, we showed that time since schizont egress had no impact on detachment force, S1 Fig. In addition, when sequential detachment forces were measured for the same merozoite, forces could be either higher or lower than previous measurements, S1F Fig. We also looked at examples of when two forces were measured for a single detachment using two different erythrocytes; there was only a weak positive correlation between the measurements, indicating a high level of variability. This could be due to measurement error, technical variation such as differences in pulling speed or biological variation, most notably differences in stiffness between individual erythrocytes, S2 Fig. Merozoites are polar cells and invasion ligands are secreted at the apical (wide) end, meaning there is likely an asymmetrical distribution of ligands across the surface. To explore whether this meant that there was a preference in which end detachment occurred, we generated a genetically modified line with a fluorescent apical marker, PfAMA1. However, it was not possible to visualise merozoite orientation and detachment at the same time due to the limited resolution of the microscope and the small size of the merozoites, S3A–S3D Fig. Reasoning that attachment might happen preferentially at the apical end; we analysed the wild-type NF54 data to test whether detachment occurred more often from the first or second erythrocyte the merozoite attached to. Detachment did occur slightly more often (68%) from the second erythrocyte the merozoite attached to, but there was no difference in force calculated from either the first or second erythrocyte, S3E Fig. These investigations concluded that the assay is robust and quantifiable enough to detect large differences in detachment force (such as between 3D7 and NF54, as in Fig 1C) but that multiple potential sources of variability limit the ability to detect more subtle differences.

Dissecting key Plasmodium falciparum determinants of attachment strength

Having established the key parameters of the optical tweezer assay, we next used it with a panel of inhibitors and genetically modified P. falciparum parasite lines, each of which disrupts specific steps of the invasion process, as summarised in Fig 2A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Impact of disrupting different stages of invasion on merozoite attachment strength.

(a) Schematic summarising the key stages of a merozoite invading an erythrocyte, as well as the stage that each interaction and the inhibitors used to disrupt them are thought to act on. (b-e) Data sets that were significantly different (t-test) have a line between them with stars showing the level of significance * = ≤0.05, ** = ≤0.01, and *** = ≤0.001. Box plots showing the detachment force of merozoite-erythrocyte attachment. The central bold line shows the median, with the top and bottom of the box at the 25^th and 75^th percentiles and the whiskers showing the total range of the data. At least three biological repeats were completed for each condition with blood from at least two different donors. The number (n) of individual interactions measured for each line is shown under the plot. The bar chart shows the frequency of erythrocyte-merozoite-erythrocyte positioning attempts that lead to attachment of the merozoite to both erythrocytes. Error bars show the SEM. (b, c, e) were carried out using conditional knock-out lines. DMSO–inactivated control or Rap–rapamycin activated conditional knockout. The number of individual interactions measured for each line is shown above each bar. (b) Measurements were carried out using cΔMSP1. (c) Measurements were carried out using cΔGAP45. (d) Comparison of wild-type NF54 untreated or in the presence of 10 μg/ml of anti-GYPC (BRIC 4), anti-GYPA (BRIC 256) or anti-Basigin (MEM-M6/6); or with neuraminidase treatment 66.7 mU/ml erythrocytes or 20 μM R1 peptide that inhibits PfAMA1-PfRON2 binding. (e) Measurements were carried out using cΔPfAMA1.

https://doi.org/10.1371/journal.ppat.1012041.g002

PfMSP1

PfMSP1 is one of the most abundant proteins on the merozoite surface [45]. Evidence for its precise role is conflicting, with several largely indirect pieces of evidence suggesting it is involved in the initial attachment of merozoites to erythrocytes [10,11], perhaps through interaction with band 3 [28] or GYPA [46], while more recent genetic evidence provides strong evidence for PfMSP1 primarily playing a role in egress [27,47]. PfMSP1 is essential for blood-stage growth, so we used a conditional DiCre recombinase-mediated gene excision approach to test its role during merozoite attachment. This conditional system involves integrating two loxP sequences into the target gene in a manner which, when integrated, does not change gene expression but can be recombined to disrupt gene expression upon the addition of rapamycin, which activates DiCre recombinase [48]. A conditional PfMSP1 line has been made previously; however, in this line, PfMSP1 is only truncated [47]. To ensure no fragments of PfMSP1 were left after disruption, which could potentially contribute to binding, we made a new PfMSP1 conditional knock-out line, which after activation, left a modified locus with only a ~16 kDa fragment of PfMSP1, S4A and S4C Fig. An excision efficiency of 94% was measured using IFA in rapamycin-treated samples, S4D Fig. The efficiency of gene deletion during imaging was assessed by genotyping PCR, which also showed a high efficiency of rapamycin-induced disruption, although some unedited loci were detected in the rapamycin-treated samples. Giemsa-stained smears post-rapamycin treatment showed a mean 2.3-fold reduction in parasitemia compared to DMSO-treated controls, S5A Fig, in keeping with previously published data [47]. There was also, as expected, a clear defect in egress characterised by poor release of merozoites seen as clumping of the merozoites (84% in rapamycin compared to 7% in DMSO control), S3 Video and S5B Fig, again in keeping with previous data [47]. Overall, this suggests a high efficiency of the activation of the conditional knock-out. Optical tweezer measurements of cΔPfMSP1 showed no significant difference in attachment strength between control DMSO treated (30 ± 2 pN) and rapamycin-activated lines (25 ± 2 pN; t-test p = 0.15), Fig 2B. The attachment frequency was also not significantly different (rapamycin-treated 22 ± 3%, DMSO-treated control 18 ± 4%; t-test p = 0.42, Fig 2B). We also plotted the data using only merozoites that were released from normal egresses for the DMSO-treated sample and the abnormal egresses for the rapamycin-activated knock-out, where we can be confident that rapamycin-induction has deleted the PfMSP1 gene; there was still no significant difference in detachment force or attachment frequency between the conditions S5B Fig. This data, therefore, suggests for the first time that PfMSP1 plays no role in determining merozoite-erythrocyte attachment strength.

cΔPfGAP45

The glideosome complex, an actinomyosin-based motor complex [22–24], is involved both in the process of erythrocyte entry and in the deformation of the erythrocyte that occurs immediately after initial merozoite attachment [21,25]. We used a conditional knock-out of PfGAP45 (cΔPfGAP45), which has previously been shown to disrupt the assembly of the glideosome and produce merozoites that egress normally but do not deform the erythrocyte membrane on contact and cannot successfully invade [25]. Genotyping PCR after rapamycin treatment confirmed efficient gene excision, in the rapamycin-treated samples, only the (0.7 Kb) bands were seen for the excised locus and no band (1.4Kb) was visible to indicate the presence of unedited parasites and mean parasitemia after reinvasion should have occurred was reduced to 1.4% of the level of the DMSO treated control, S5C Fig, in keeping with previous data [25]. Despite this impact on invasion, loss of PfGAP45 caused no effect on the detachment force (DMSO mean– 31 ± 2 pN compared to rapamycin 31 ± 3 pN, not significant p = 0.83) but did significantly reduce attachment frequency (25 ± 2% for DMSO-treated to 8 ± 2% for rapamycin-treated, t-test p = ≤0.0000), Fig 2C. This indicates that the glideosome plays a role in parasites attaching/remaining attached for the time of the measurements but does not affect the strength of the subsequent attachment.

Disrupting PfEBA and PfRH ligand-receptor interactions

After initial attachment has occurred, the merozoite reorientates so that its apical end faces the erythrocyte membrane, and it begins to deform the erythrocyte. This process depends on multiple interchangeable members of the PfEBA and PfRH family of ligands, which bind to an array of erythrocyte receptors, some known and some unknown [15]. We initially used enzymes and antibodies to disrupt known erythrocyte receptor partners to explore the role of PfEBA and PfRH ligands in attachment strength. We confirmed that the antibodies bound to erythrocytes using flow cytometry, S6A Fig and replicated their previously reported effects on parasite growth using growth inhibition assays with the NF54 strain S6B–S6D Fig.

Neuraminidase cleaves sialic acid residues of glycoproteins and glycolipids on the surface of the erythrocyte. The interactions of GYPA-PfEBA175, GYPC-PfEBA140, GYPB-PfEBL1 (not relevant in NF54 due to a missense mutation in PfEBL1 [49]) and the binding of PfEBA181 and PfRH1 to unknown receptors have all been shown to be sensitive to the neuraminidase treatment [50–52]. Neuraminidase treatment reduced invasion by nearly half in NF54 strain parasites (60 ± 4%), S6D Fig, as expected. There was also significantly reduced attachment frequency for neuraminidase-treated erythrocytes (1.4 ± 0.5%), which was significantly different to NF54 (rank-sum p = 0.0001, Fig 2D); in fact, only 4 attachments could be measured across 28 egresses and 282 erythrocyte-merozoite-erythrocyte positionings when using neuraminidase treated erythrocytes. For these 4 attachments, the mean detachment force was much lower than for the detachment from untreated blood, but the difference was not significant (13 ± 5 pN, t-test p = 0.14, Fig 2D), likely due to the small number of data points that could be achieved due the very low attachment frequency. This strongly suggests that sialic acid-dependent interactions play a major role in merozoite-erythrocyte attachment.

PfEBA175 binds to GYPA on the erythrocyte surface [53,54]. Anti-GYPA antibodies have previously been shown to reduce the invasion efficiency of P. falciparum [55], and in our assays, at 10 μg/ml of anti-GYPA invasion was 56 ± 7% that of untreated NF54, S6B Fig. The presence of anti-GYPA also caused a significant reduction in the mean detachment force to 14 ± 2 pN, (t-test p = 0.025), Fig 2D. As with neuraminidase, the mean attachment frequency per egress 5 ± 2%, was also significantly lower than wildtype NF54 (ranksum p = 0.0002), Fig 2D. Inhibition using anti-GYPA therefore has a significant effect on the attachment behaviour, pointing towards PfEBA175-GYPA binding being a key attachment interaction in the P. falciparum NF54 line.

Erythrocyte protein GYPC is the receptor for PfEBA140 [56], anti-PfEBA140 antibody led to a ~20% reduction in invasion in 3D7 [52,56]. Anti-GYPC (BRIC 4) at 10 μg/ml reduced invasion to 89% ± 7% of levels observed in the absence of the antibody, S6B Fig, but the same concentration had no significant effect on either detachment force (19 ± 3 pN; p = 0.15, Fig 2D), or attachment frequency (16 ± 5%, t-test p = 0.47, Fig 2D). This is in keeping with previous evidence that suggested that PfEBA175 is a more critical invasion ligand than PfEBA140 in 3D7, and presumably by inference in NF54, the parent of 3D7 [52]. We also tested anti-CR1 antibodies, as CR1 is the receptor for PfRH4 [57], but saw no significant effect on invasion efficiency, detachment strength or attachment frequency at the concentrations tested, S6B, S6E and S6F Fig, potentially because the specific antibodies used do not productively interfere with the PfRH4-CR1 interaction.

Anti-Basigin

As invasion progresses after strong deformation has occurred, PfRH5 binds to Basigin (also known as CD147) [19,30,58]. PfRH5 is the only essential member of the PfRH family [19,59,60] and forms part of the PfPTRAMP–PfCSS–PfRipr–PfCyRPA–PfRh5 (PCRCR) complex [19] which has been proposed to be required for the discharge of the rhoptries and calcium signalling during invasion [17,30]. In keeping with previous studies [58,61], anti-Basigin antibodies at 10 μg/ml reduced invasion to 23 ± 4% of untreated NF54, S6B Fig. Anti-Basigin also caused a significant (p = 0.037 t-test) increase in the measured detachment strength, with a mean detachment force of 31 ± 3 pN compared to 24 ± 2 pN for untreated NF54 Fig 2D. There was no effect on attachment frequency (16 ± 3% t-test p = 0.44), Fig 2D.

PfAMA1-PfRON2 interaction

The parasite irreversibly commits to invasion when PfRON2 is integrated into the erythrocyte membrane and bound by the PfAMA1 in the merozoite membrane, forming the tight junction [20]. We tested two approaches to disrupt PfAMA1-PfRON2 binding; using a small peptide inhibitor R1 that binds PfAMA1 [62] and a conditional PfAMA1 knock-out line (cΔPfAMA1) [27]. The R1 peptide at 20 μM reduced invasion to 27 ± 3% of untreated parasites S6C Fig, in keeping with previously reported figures [20]. As with anti-Basigin antibodies, 20 μM R1 peptide significantly increased detachment force (mean 33 ± 3 pN, t-test p = 0.0037) and it also increased attachment frequency (29 ± 4%, t-test p = 0.055), Fig 2D. However, when testing the conditional PfAMA1 knock-out line there was no difference in detachment force (27 ± 2 pN DMSO control vs 29 ± 2 pN rapamycin activated, t-test p = 0.55) or significant reduction in attachment frequency (32 ± 3% DMSO control and rapamycin 23 ± 5%, t-test p = 0.14). Both genotyping PCR and invasion efficiency, S5D Fig, showed a high level of editing in the rapamycin-treated samples. There were faint bands in the rapamycin-treated samples for two repeats, indicating that editing was not 100% efficient and there could still be a low percentage of unedited parasites. However, there was a considerable reduction in invasion after rapamycin treatment (mean parasitaemia of rapamycin-treated samples was 2.0% of the parasitaemia of DMSO control samples S5D Fig), confirming a high efficiency of successful deletion of PfAMA1 post-rapamycin induction.

Correlation between invasion rates and optical tweezer measurements

For conditions where we measured the parasite replication rate either through the growth rate assay (NF54 and 3D7) or through the growth inhibition assays (anti-GYPC, anti-GYPA, anti-Basigin, neuraminidase treatment erythrocytes or R1 peptide) we were interested in how these measurements correlated with the measurements of attachment, S6 Fig. When the invasion efficiency relative to NF54 and detachment force were compared there was only a weak positive correlation (correlation coefficient 0.33). However, when we excluded the conditions that disrupted late in invasion (anti-Basigin and R1 peptide), which are involved after the stage likely important to attachment strength, the positive correlation between invasion efficiency and detachment force was much stronger (correlation coefficient 0.982). The relationship between invasion rate and attachment frequency was also strong, with a correlation coefficient of 0.81 when including data from late-stage inhibitors and 0.95 when excluding them. This indicates that there is likely a positive relationship between both the strength of detachment and attachment frequencies with invasion rate, with stronger attachment leading to more productive invasion.

Understanding the contribution of individual PfEBA and PfRH ligands using single knock-out lines

Several inhibitors that disrupted the interactions of members of the PfEBA and PfRH invasion ligand families (neuraminidase, anti-GYPA and anti-GYPC) reduced the detachment strength in a manner that correlated to a reduction in the invasion rate, S5 Fig. The P. falciparum PfEBA family consists of PfEBA140, PfEBA175, PfEBA181, PfEBL1 and PfEBA165, although PfEBA165 is a pseudogene in all strains tested to date [63,64], and PfEBL1 is also a pseudogene in multiple strains, including the NF54 background used here [49]. The P. falciparum PfRH family consists of PfRH1, PfRH2a, PfRH2b, PfRH3, and PfRH4 (and PfRH5, although as noted above it is thought to operate at a later invasion step, and was not considered further here). Like PfEBA165, PfRH3 is also a pseudogene in all strains tested to date [65].

In order to test more systematically the role in attachment of specific proteins in the PfEBA and PfRH family we generated a panel of knock-out lines, Fig 3A. All PfEBA and PfRH knockouts have been individually made previously: 3D7ΔPfEBA175 [66,67], W2mefΔPfEBA175 [66], 3D7ΔPfEBA181 [67], 3D7ΔPfEBA140 [56,67], W2mefΔPfEBA140 [56], Tak994ΔPfRH1 [68], 3D7ΔPfRH1 [69], 3D7ΔPfRh2a [69], 3D7ΔPfRh2b [52,69] and W2mefΔPfRH4 [70], but these knockouts exist on a wide range of strain backgrounds and were generated using different experimental genetic approaches. To allow for a more complete comparison, we assembled a consistent set of knockouts constructed in the same manner in NF54 background. Two control knock-out lines were made using the same approach targeting PfP230P and Pfs25; both proteins are gametocyte specific, have very low-level expression in the asexual stage and have no known role in invasion [71–73]. Some knockouts (NF54ΔPfEBA140, NF54ΔPfEBA181, NF54ΔPfRH4, NF54ΔPfEBA175, Dd2ΔPfEBA175) were made as part of a larger systematic study of P. falciparum genes, while others (NF54ΔPfRH2a, NF54ΔPfRH1, NF54ΔPfP230P, NF54ΔPfs25) were made specifically for this study. All lines were cloned by limiting dilution and genotyped following cloning specifically for this study S8A and S8B Fig, see Methods for further details.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Characterisation of PfEBA and PfRH knock-out lines (a) Schematic of the regions disrupted in each knock-out line. Striped regions indicate the regions deleted and replaced by the drug resistance cassette. For all genes, this insertion was made within the predicted erythrocyte binding domain apart from in PfRH4, which was generated as part of another study, and PfRH2a, where the binding domain is identical between PfRH2a and PfRH2b, meaning a unique deletion must be targeted elsewhere. The pink regions show the predicted binding regions, which were predicted using https://www.ebi.ac.uk/interpro. The light blue regions show the predicted transmembrane domains and the dark blue region shows the cytoplasmic domains. (b, d-e) Data sets that were significantly different (t-test) have a line between them with stars showing the level of significance * = ≤0.05, ** = ≤0.01, and *** = ≤0.001. The central bold line shows the median, with the top and bottom of the box at the 25^th and 75^th percentiles and the whiskers showing the total range of the data. The red line across the graph shows the median and shading across the whole graph shows the 25^th to 75^th percentile range of NF54. All knock-out lines tested are in the NF54 background. (b) Box plot showing the growth rate measured in static culture represented as the Parasite Erythrocyte Multiplication Rate (PEMR), calculated by dividing the parasitemia of the ring stage culture by the parasitemia of the same culture measured 24 h previously. PEMR was measured as detailed in the Methods. At a 5% level of significance, the only lines that were different to NF54 mean PEMR = 4.5 ± 0.3 were NF54ΔPfs25 PEMR = 5.8 ± 0.2 (p = 0.0025) and NF54ΔPfEBA140 PEMR = 6.3 ± 0.4 (p = 0.0020). (c) Analysis of invasion-associated gene expression using qPCR. Error bars show the standard deviation between repeats. The expression in each sample is normalised to the housekeeping gene PfActin I and stars indicating significant changes at greater than a 5% significance level (t-test). The gene expression of each knock-out line is expressed as a relative fold change compared to NF54ΔPf230P. We monitored the expression of the PfEBA and PfRH family of proteins, as well as two essential invasion genes PfAMA1 and PfRH5. Note that ΔPfEBA175 was not included in this experiment as the line was not available at the time. (d) Box plots showing the detachment force of merozoite-erythrocyte binding. The central bold line shows the median, with the top and bottom of the box at the 25^th and 75^th percentiles and the whiskers showing the total range of the data. The red line across the graph shows the median and shading across the whole graph shows the 25^th to 75^th percentile range of NF54. At least three biological repeats were completed for each data set with blood from two different donors; the number (n) of individual interactions measured for each line is shown under the plot. (e) The bar chart shows the frequency of erythrocyte-merozoite-erythrocyte positions that lead to successful attachment of the merozoite to both erythrocytes. The attachment frequency of all the knockouts was not significantly lower than NF54 at p = ≤0.05 except NF54ΔPfRH1 which was nearly significant at a 5% level (t-test p = 0.058), data sets that were not significantly different (t-test) have a line between them with NS above. The number of erythrocyte-merozoite-erythrocyte positioned for each line is shown above each bar. Error bars show the SEM.

https://doi.org/10.1371/journal.ppat.1012041.g003

Effect of PfEBA and PfRH gene deletions on growth

We began by carrying out a growth assay on the knock-out lines, Fig 3B. We used tightly synchronised parasites in triplicates and sampled parasitemia daily using flow cytometry over multiple invasion cycles (see Methods for further details). The Parasite Erythrocyte multiplication rate (PEMR), [44], was then calculated by dividing parasitemia after invasion by parasitemia in the same culture the day before. Under static culture, there were significant differences in the mean growth rate relative to NF54 (mean PEMR = 4.5 ± 0.3) for NF54ΔPfs25 (PEMR = 5.8 ± 0.2) was 28% higher (p = 0.0025) and NF54ΔPfEBA140 (PEMR = 6.4 ± 0.4) was 38% higher (p = 0.0020). The fact that the growth rate of NF54ΔPfs25 was higher than NF54, despite Pfs25 having no known role in invasion, suggests that the invasion rate can change during the process of generating the knockout lines. Differences in invasion phenotypes of clones of the same line have been reported previously [74–77]. The growth rate of NF54ΔPfEBA140 was not significantly different to NF54ΔPfs25 (p = 0.28), so we cannot conclude that this difference is due to PfEBA140 deletion.

RT-qPCR shows few significant changes in invasion gene expression in PfEBA and PfRH knockout lines

PfEBA and PfRH proteins are known to have variable expression patterns which has been correlated with invasion phenotype changes [67,70,78–82]. Generation of a cloned knock-out line requires 3–6 months of culturing, so we assessed whether any of our PfEBA and PfRH knockout lines also carried transcriptional changes that could impact phenotype by using qPCR to compare transcript levels across all family members. PfEBA and PfRH proteins have highly stage-dependent expression, with expression peaking in late schizonts (~44–48 h) and merozoites [81,83,84]. Samples were therefore collected of tightly synchronised schizonts treated with Compound 2 (C2), which pauses the schizonts ~15 mins before egress [85]. We assessed whether the C2 treatment had any transcriptional impact by measuring the expression of the invasion genes in wild-type NF54 before and after 3.5 h of C2 treatment; no differences were observed, S9A Fig. For most of the PfEBA/PfRH knockout lines, there was a large decrease in transcription of the targeted gene (NF54ΔPfEBA140, NF54ΔPfEBA181, NF54ΔPfRH1, NF54ΔPfRH2a), as expected (Figs 3C and S9B). The only exception was NF54ΔPfRH4, where there was no significant change in PfRH4 expression relative to NF54ΔPFP230P. However, the qPCR primers amplified a region downstream of the integrated selectable marker for all knock-outs, except for NF54ΔPfRH4 where they amplified a region upstream of the integration, meaning they may detect a stable yet truncated transcript. Other than the expected decrease in expression of target genes, there were no consistent differences in expression patterns across lines, but expression levels of PfEBA181, PfEBA140 and PfRH1 did frequently differ between lines, even in the control knockouts, and in NF54ΔPfRH2a there was a small but significant 1.4-fold increase in the expression of PfRH2b (t-test p = 0.034). The ΔPfEBA175 line was not tested as it was not constructed when this experiment was carried out.

Effect of PfEBA and PfRH gene deletions on merozoite-erythrocyte attachment strength

Optical tweezer measurements were performed on all PfEBA and PfRH knock-out lines along with the control lines. Across all knockout strains, 267 egresses were observed, 2608 individual erythrocyte-merozoite-erythrocytes were positioned, and 251 attachments were measured, Fig 3D and 3E. The attachment frequency of all the knockouts (except NF54ΔPfRH1) was significantly lower than NF54, including both controls NF54ΔPfP230P (t-test p = 0.049) and NF54ΔPfs25 (t-test p = 0.045), Fig 3E. This indicates the process of constructing the knock-out lines impacted the resultant attachment frequency, but there was no difference in attachment frequency between knockout lines. By contrast, the detachment force of NF54ΔPfRH4 (mean 19 ± 3 pN) was significantly lower than both NF54ΔPfs25 (mean 26 ± 2 pN p = 0.0321) and ΔPfP230P (mean 26 ± 2 pN p = 0.044) control lines, Fig 3D, suggesting a role for PfRH4 specifically in attachment strength in NF54. None of the other knockout lines had significantly different distributions of detachment force compared to NF54 or the two control lines.

The lack of phenotype in the NF54ΔPfEBA175 line is very different from the significant decrease in detachment force after treatment with anti-GYPA, which targets the same PfEBA175-GYPA interaction albeit through a different inhibitory mechanism. This interaction is thought to be more important for invasion in the Dd2 strain than in NF54 because invasion in Dd2 is almost completely blocked by neuraminidase treatment, whereas in NF54 and 3D7 neuraminidase treatment only reduces invasion by around 50%, S6D Fig and [86–88]. We therefore also tested a ΔPfEBA175 line made in the Dd2 background. The optical tweezers attachment assay showed unexpectedly that Dd2ΔPfEBA175 had a significantly (p = 0.0001) higher detachment force (mean = 31 ± 2 pN) compared to the Dd2 wildtype control (mean = 23 ± 2 pN), S10B Fig, whereas the attachment frequency was significantly lower, in keeping with the knockouts made in the NF54 background, Fig 3E. While surprising, it is notable that PfRH4 has been shown to be significantly up-regulated when PfEBA175 is disrupted in the parental lines of Dd2 W2mef [70]–there is therefore consistency between the reduction in detachment force when PfRH4 is deleted, and an increase in detachment force when it is up-regulated, implying that PfRH4 may play a particularly important role in attachment. We next tested the Dd2ΔPfEBA175 line in the presence of anti-GYPA, which resulted in a significant reduction in detachment force (mean 16 ± 2 pN, p = 0.0001) compared to Dd2ΔPfEBA175. This indicates that some of the effect of anti-GYPA inhibition is independent of disruption of the PfEBA175-GYPA interaction, S10B Fig, either acting on other PfEBA interactions or in another indirect manner.

Ethics statement

Ethical approval for the use of human blood was obtained from National Health Service (NHS) Cambridge South Research Ethics Committee, 20/EE/0100, and the University of Cambridge Human Biology Research Ethics Committee, HBREC.2019.40, formal written consent was obtained at the point of collection and is held by the NHS Blood and Transplant (NHSBT).

Plasmodium falciparum culture

P. falciparum strains (either the background NF54, 3D7 or Dd2 lines, or genetically modified lines described below) were cultured in human erythrocytes (purchased from NHSBT, Cambridge, UK) at a 4% haematocrit (HCT) in RPMI 1640 medium (Gibco, UK) supplemented with 5 gl^-1 Albumax II, DEXTROSE ANHYDROUS EP 2 g l^-1, HEPES 5.96 g l^-1, sodium bicarbonate EP 0.3 gl^-1 and 0.05 gl^-1 hypoxanthine dissolved in 2 M NaOH. Culture was performed at 37°C in a gassed incubator or gassed sealed culture container under a low oxygen atmosphere of 1% O₂, 3% CO₂, and 96% N₂ (BOC, Guildford, UK).

Sorbitol synchronisation

Sorbitol synchronisation was done as described by [103]. Ring stage cultures were pelleted by centrifugation, the supernatant was removed, and the pellet was resuspended in ten times the pelleted volume of 5% D-sorbitol (Sigma-Aldrich), then incubated at 37°C for 5 mins, during which time later stage parasites (trophozoites and schizonts) rupture. The cells were centrifuged and resuspended in RPMI at twenty times the pelleted volume. The cells were pelleted again and then resuspended in complete medium to give 4% HCT.

Percoll synchronisation

Smears were checked to confirm predominantly schizonts and a few rings. Most of the media was removed and the infected blood was resuspended in 5 ml of media, then layered over 5 ml of 70% Percoll in a 15 ml tube (Percoll Merck P4937, 10% 10x PBS, 20% RPMI). The tube was then centrifuged at 1450 rcf for 11 mins with the break set to 1 and accelerator to 3. The band of late-stage parasites at the percoll-media interface was then added to a flask with media and blood at a 4% HCT and incubated in a gassed incubator at 37°C for 3.5 h unless indicated otherwise. The Percoll separation was then repeated as above but this time the bottom pellet containing ring stage parasites (which had invaded in the previous 3.5 h) was kept and then sorbitol synchronised as described above.

Optical tweezers imaging

The method used to carry out the optical tweezers analysis is described in [31]. Protocol available DOI: dx.doi.org/10.17504/protocols.io.5qpvo3zmzv4o/v1. Cultures were sorbitol synchronised at least three times in consecutive cycles before imaging; a 25 ml culture was split so 20% of the flask was diluted for continuous culture and the remainder was used for imaging. Schizonts were isolated for imaging using magnetic purification [104] with a magnetic cell fraction column (MACS, Miltenyi Biotec) and a MidiMACS separator. The eluted schizonts were pelleted by centrifugation and resuspended in 200–800μl of complete medium, depending on the pellet size. Uninfected blood was added to 200ul of schizonts to give a 0.025% HCT. If antibodies, S1 Table, or the R1 peptide were added, they were added at the indicated concentrations at this stage and were present for the imaging duration. Imaging was performed with the cells in a SecureSeal hybridisation chamber (SecureSeal; Grace Bio-Labs, Bend) and coverslips coated with 6 μl of poly(l-lysine)-graft-poly(ethylene glycol) (PLL [20]-g[3.5]-PEG [2]) (SuSoS AG, Dubendorf, Switzerland) at 0.5 mg/mL concentration to reduce adherence of the erythrocytes to the coverslip.

The temperature was kept at 37°C throughout the experiment using a custom-built temperature-control stage. Imaging was performed on an Eclipse Ti-E, Nikon inverted microscope and a 60x water objective (Plan Apo VC 60x 1.20 NA water objective, Nikon). The custom-built optical tweezers set-up, described in [105], consists of a solid-state pumped Nd:YAG laser (PYL-1-1064-LP, 1064 nm, 1.1 W (IPG Laser). The traps are steered with a pair of acousto-optical deflectors (AA Opto-Electronic, Orsay, France) controlled by custom-built electronics; multiple traps can be used with subnanometer position resolution. Videos are recorded in brightfield at 20 frames/s using model Grasshopper3 GS3-U3-23S6M-C (Teledyne FLIR) with a Physical pixel size of 5.86 μm resulting in a 0.0973 μm pixel size in the images. The cells were manipulated as indicated by Fig 1A. Three repeats were done for each condition on separate days using at least two different donors’ blood, to account for both human and parasite potential sources of biological variation.

Determining the attachment frequency

The frequency of attachment was established for each condition by calculating the frequency at which an adhered erythrocyte-merozoite-erythrocyte chain was formed when two erythrocytes were placed next to a merozoite and left for ~2–5 s Fig 1B. If the same combination of cells was positioned more than three times, they were no longer counted. Measurements were only recorded for erythrocyte-merozoite-erythrocyte chains positioned within 180 s of egress.

Determining the detachment force

Pulling is done in one of two ways, depending on whether one erythrocyte is attached to the glass slide. If one erythrocyte has attached to the slide, then only one trap is needed and is used to stretch the other erythrocyte (for NF54 this occurred 78% of the time). If neither erythrocyte is attached to the slide, two traps are used, one to hold one erythrocyte, while the other erythrocyte is pulled using a second trap. It has been shown that erythrocytes’ mechanical properties can be modelled as a linear spring [37]. This means there is a linear relationship between the force applied and the stretch of the erythrocyte along a single axis. A healthy human erythrocyte has a stiffness (k) of 20 pN/μm [37] with ~20% uncertainty due to erythrocyte variation. The elongation of the erythrocyte (ΔL = L_max-L₀) was calculated by measuring the length of each erythrocyte before it was moved (L_o) and subtracting it from the length in the frame before detachment I (L_max), Fig 1A. Measurements of length were performed using ImageJ software. The force of detachment is calculated based on F = kΔL [31]. Measurements were only taken if the attachment was by a single merozoite, and force measurements were not taken from erythrocytes attached to the glass slide. If one of the erythrocytes undergoes echinocytosis then it is not measured. If two erythrocytes were stretched, then one force was selected at random and included in the data set (so that two measurements for the same detachment were not included).

Growth rate assay

At the start of the assay all lines tested were defrosted at the same time; they were sorbitol synchronised in the first cycle when >0.5% rings and then delayed at room temperature to push the egress window to the intended start window. The percoll synchronisation described above was carried out with a 3 h invasion window. The parasitemia was then counted using Giemsa smears and the cultures diluted to a 0.2–0.5% parasitemia at a 4% HCT (the parasitemia was set as high as possible up to 0.5% based on the number of parasites available and the number of wells). For the duration of the assay, the lines were grown in 5 ml at 4% HCT in a 6-well plate. Every day at ~3 h after the initial invasion window the parasitaemia was measured using flow cytometry. 5 μl of each well was diluted in 45 μl of PBS in a 96-well plate. The samples were stained using SYBR Green I (Invitrogen, Paisley, UK) at a 1:5000 dilution for 45 mins at 37°C. The samples were then run on an Attune NxT acoustic focusing cytometer (Invitrogen) with SYBR green excited with a 488 nm laser (BL1-A) and detected by a 530/30 filter. The parasitemia was measured using the Attune NXT software; for each data set, a plot of SSC-A vs FSCA was used to gate for roughly the size of an erythrocyte, and then singlets were gated for in an FSC-H vs FSC-A. Next histograms for the BL1-A intensity were used to gate for the SYBR green positive erythrocytes; the percentage of positive erythrocytes represents the percentage of infected erythrocytes. Every 48 h, when the culture was ring stage, it was split to give ~0.5% parasitemia.

The Parasite Erythrocyte Multiplication Rate (PEMR) for each day was calculated by dividing the parasitemia of the ring stage culture by the parasitemia the day before. Custom Python scripts were used to do further analysis. Due to a negative correlation between the starting parasitemia and PEMR of the subsequent cycle, any cycles where the starting parasitemia was >1% were removed. We also manually went through the plots of the PEMR for each cycle by repeat and removed any data sets that had a very high parasitemia (typically greater than 5%), causing subsequent repeats to be very low (the culture had crashed). The measurements were taken over a total of 5–8 invasion cycles, the number of cycles was determined by when blood was available to allow the cultures to be split over the course of the assay using blood from 3 donors to account for potential human sources of technical variation.

Tight synchronisation

Defrosted lines were sorbitol synchronised in the first cycle when the culture was greater than 0.5% rings and then delayed at room temperature to push the egress window to ~9 am on a Tuesday or Thursday. The above percoll synchronisation was carried out with a 3.5 h invasion window. The infected blood was resuspended at 1.5% HCT in complete media. The flask was delayed at room temperature until 8 pm on Tuesday or if synchronisation was done on a Thursday until 8 pm on Friday to ensure the start of the egress window remained around 9 am on Tuesday and Thursday. This was done repeatedly every Tuesday and Thursday for the duration of the experiments.

Rapamycin activation and DMSO control treatment of conditional knockouts

All the conditional knockout lines used were made in the B11 background, a clone of 3D7 that expresses DiCre recombinase components [25,48]. When activating the conditional knock-out lines the above tight-synchronisation protocol was used, when the culture was at a parasitemia > 4% and in 1.5 ml of packed infected blood at the start of the Tuesday cycle, the above protocol was followed at the end of which 1 ml of the blood was taken for imaging and the remaining 0.5 ml was diluted with fresh blood to keep the total packed blood at 1.5 ml and kept in the synchronisation protocol. The 1 ml of blood was split into two flasks with 30 ml of media added to each and delayed until 8 pm on Wednesday at which point dimethyl sulfoxide (DMSO) was added to one flask and rapamycin (Sigma-Aldrich) resuspended in DMSO added to the other at a 1:10,000-fold dilution to give a final rapamycin concentration of 10 nM. The flask was returned to the incubator at 37°C for 14 h to allow rapamycin-induced DiCre-mediated gene deletion to occur. The culture was washed twice with complete media to remove the DMSO or rapamycin before returning to normal culture conditions.

Protein separation by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS PAGE) and Western blot analysis

DMSO (control) and Rapamycin (Rap) treated schizonts, prepared as described above, were solubilised 3:1 in 4X SDS loading buffer (0.2 M Tris-HCl, 277 mM SDS (8% w/v), 6 mM Bromophenol blue, 4.3 M glycerol) and boiled at 95°C for 5 mins. Proteins in samples were separated, alongside a protein standard molecular weight marker (SeeBlue Plus2, Invitrogen), on 4–15% gradient polyacrylamide gel (Mini-Protein TGX precast protein gels, BioRad) by gel electrophoresis at a constant voltage of 200 V in SDS running buffer (25 mM Tris, 192 mM glycine, 0.1% SDS, pH 8.3).

Following separation, gels were analysed by Western blot, probing for MSP1 (X509, anti-MSP1 p38/42, 1:1000) or SERA5 (pAb Anti-SERA5, 1:2000). The proteins separated in gels were transferred onto nitrocellulose membrane (pore size 0.2 μm, Biorad) via overnight wet transfer (Trans-Blot cell, Biorad) in transfer buffer (25 mM Tris, 192 mM glycine, 20% methanol (v/v), pH 8.3). Nitrocellulose was blocked for 1 h using 5% semi-skimmed milk in PBS 0.05% Tween 20, washed three times in PBS 0.05% Tween 20 and incubated at room temperature with the primary antibody diluted in 3% w/v Bovine Serum Albumin (BSA) for 1 h. The blots were then washed three times in PBS 0.05% Tween 20 and then incubated at room temperature with horseradish peroxidase-conjugated secondary antibody diluted in PBS 0.05% (For X509: Goat-anti-human (Abcam), 1:5000; for pAb anti-SERA5: Goat-anti-rabbit (Abcam), 1:3000) Tween 20 for 1 h. After incubation, blots were washed a further three times for 5 min in PBS 0.05% Tween 20, then horseradish peroxidase substrate (Immobulin, Millipore) was applied and then immediately imaged using a Biorad ChemiDoc imaging system.

Immuno-fluorescence assay (IFA)

After synchronisation and treatment with Rap or DMSO as described above, parasites were collected at schizogony by using centrifugation on a 70% Percoll cushion. Parasites were then washed in RPMI w/Albumax. Schizonts were placed in PKG inhibitor C2 (4-[7-[(dimethylamino)methyl]-2-(4-fluorphenyl)imidazo[1,2-α]pyridine-3-yl]pyrimidin-2-amine), LifeArc), 1 μM in fresh RPMI w/Albumax, and allowed to mature for 4 h. Schizonts were then smeared onto glass coverslips and allowed to dry completely. If not to be used immediately, dried coverslips were wrapped in tissue paper and stored at -80°C in water-tight sealed bags with silica beads (desiccant condition). Sealed containers were then warmed to 37°C for 1 h prior to use. Smears on coverslips were then fixed with 4% formaldehyde (Thermo scientific) in PBS for 1 h and then washed with 0.1% Triton X-100 in PBS for 10 min and then two times in PBS for 5 min. Slides were blocked with 3% w/v BSA (Sigma Aldrich) overnight and then washed three times in PBS. Slides were incubated with primary antibody in 3% w/v BSA (X509, 1/5000) in a humidified chamber at 37°C for 1 h and then washed three times PBS for 5 min. Incubation with secondary Alexa Fluor antibodies diluted in 3% w/v BSA (1/2000, Goat-anti-human-alexofluor647) was carried out in a humidified chamber at 37°C for 1 hr, shielded from light and then washed three times in PBS for 5 min. Slides were mounted in Vectashield containing DAPI (Vector laboratories) and images were acquired using a AxioVision 3.1 software on an Axioplan 2 microscope (Zeiss) with a Plan-APOCHROMAT 100×/1.4 oil immersion objective and processed using Fiji.

Imaging of conditional knock-outs

The conditional knock-out lines were treated as described above then on Friday imaging with the optical tweezers was carried out as described above. Imaging was performed at least three times for each line with the DMSO and rapamycin samples imaged on the same day in the same blood; different blood was used for each experiment, to account for potential human sources of biological variation. A sample of the culture was saved to make Giemsa smears to assess after the invasion window and a pellet of the culture was saved for genotyping on the day of imaging.

Generation of PfAMA-mNEON

The PfAMA1-mNEON tagged line was constructed by integrating a full-length PfAMA1-mNEON fusion, flanked by 5’ and 3’ regions of the PfAMA1 gene, into the PfP230P locus. The line also retained episomal copies of the repair template plasmid, meaning that PfAMA1-mNEON could be expressed from both the PfP230P locus and the episomal plasmid. The 5- and 3’ homology regions and guides used here and for the ΔPfP230P described below were based on the design used by [106]. All the sequences used for construction are detailed in, S4 Table. A repair template plasmid was constructed using Gibson assemble [107] (NEB Hifi assembly kit). The repair construct was made by joining the PfAMA1 coding region and ~1.5 Kb 5’ to the transcription start site amplified from 3D7 genomic DNA, a C-terminal mNeongreen tag and the ~1 Kb region 3’ to the end of the PfAMA1 transcript amplified from 3D7 genomic DNA; either side of this assembly were homology regions amplified from PfP230P. The repair template was inserted into the NotI/SacII sites of the pCC1 plasmid (PlasmoGEM Sanger) while removing the negative selection cassette. The guide was inserted into plasmid pDC2-cam-Cas9-U6-yDHODH (PlasmoGEM Sanger), pDC2 contained two bbsI sites in the guide RNA sequence, for each guide a pair of primers were designed which when annealed produce the guide sequence with overhangs matching those produced when pDC2 was digested with bbsI (New England Biolabs, USA), allowing for ligation to occur. pDC2 also contains the cas9 expression sequence.

Ethanol precipitation was used to prepare 60 mg of repair pCC1 plasmid and 30 mg of guide pDC2 plasmid in 60 μl of cytomix (120 mM KCl, 0.15 mM CaCl2, 2 mM EGTA, 5 mM MgCl2, 10 mM K2HPO4, 25 mM HEPES, pH 7.6). Ring stage sorbitol-synchronised culture (parasitemia 4–5% rings) was washed in cytomix. For each transfection 60 μl of DNA and 450 μl of infected erythrocytes were then electroporated at 310 V, Resistance: infinite, 950 μF with a Gene Pulser (BioRad, Watford UK). This was transferred into 6 ml of complete media to give a 4% HCT. The culture was then maintained on WR99210 (WR) (Sigma-Aldrich) until parasites could be seen in Giemsa smears. The selection was done with WR99210, which only selected for the repair template in this instance. The PCC1 repair template plasmid contains the hDHFR gene which gives resistance to otherwise toxic WR99210 (WR) [108]. The line was genotyped, as described in S4A–S4B Fig using primers detailed in S15 Table.

Antibody binding assay

Blood was diluted to 4% HCT and washed twice in complete media. For each condition, 100 μl of blood was added to a well of a 96-well plate. The cells were pelleted by centrifugation, the supernatant removed and replaced with 100 μl of the complete medium if indicated containing the primary antibody, detailed S1 Table. Control wells were run with just blood or just the secondary antibody. The plate was incubated for 60 min at 37°C and then washed three times in PBS. Then 100 μl of the second secondary antibody goat-anti-rabbit AlexFluor488-labelled Invitrogen (A32723) at a 1:1000 dilution (2 μg/ml) was added to the appropriate wells. Two washes were then done in PBS and then the samples were diluted to at 0.4% HCT in a round bottom 96-well plate. The sample was then analysed using an Attune NxT acoustic focusing cytometer (Invitrogen) with excitation using a 488 nm laser and detection using a 530/30 filter. 50,000 events were collected per sample. Analysis was done using a custom Python script that used the FlowCal library [109]. Initial gating was done as before to the isolate signal from single erythrocytes. Next histograms for the BL1-A intensity which detects the Alex Fluro plus 488 were normalised to allow comparison of data with different counts. We then averaged the histograms of the repeats.

Growth inhibition assay

Two-colour assays were set up as described [29]. The day before, the culture was synchronised using sorbitol as described above so that the culture was predominantly late-stage trophozoites at the time of assay set-up with a 1–2% parasitemia and 2% HCT. Uninfected erythrocytes were stained with 4 μM CellTrace Far Red Cell Proliferation kit (Invitrogen, UK) for 2 h at 37°C whilst rotating and then washed twice in complete media and resuspended at a 2% HCT. 50 μl of the labelled erythrocytes were mixed with 50 μl of the infected culture in the wells of a flat-bottomed 96-well plate. The erythrocytes were pelleted by centrifugation, the supernatant removed and replaced with 100 μl of the complete medium, containing the relevant antibody or inhibitor at the final concentration if indicated. The details of the commercial antibodies used are detailed in S1 Table, while the R1 peptide at 98% purity (Sequence: VFAEFLPLFSKFGSRMHILK,GenScript). The first anti-GYPC antibody we tested, BRIC 10, caused the erythrocyte to aggregate, so was not used further. In all assays, wells were set up with no antibodies used as controls to express relative expression, and wells containing only unstained uninfected erythrocytes were included to aid in correctly positioning gates. To obtain neuraminidase-treated erythrocytes, blood was diluted to 20% HCT in RPMI, neuraminidase (Vibrio Cholera N7885-1UN) was added at 66.7 mU/ml (the stock was assumed to be at 1.5 U/ml), and the mixture incubated at 37°C for 1 h on a rotator. After incubation, erythrocytes were washed twice with RPMI and stained with CellTrace Far Red as above.

Plates were incubated for 20 h under P. falciparum culture conditions described above. After incubation, infected erythrocytes were stained with the fluorescent DNA dye SYBR Green I (Invitrogen, Paisley, UK) at a 1:5000 dilution for 1 h at 37°C. The sample was then fixed by suspending in 2% paraformaldehyde and 0.2% glutaraldehyde in PBS and incubating at 4°C for 30 min. The sample was then resuspended at 0.4% HCT in PBS. The samples were then analysed either with a BD Fortessa flow cytometer (BD Biosciences, Oxford, UK) running BD FACS Diva software (BD Biosciences, Oxford, UK) or an Attune NxT acoustic focusing cytometer (Invitrogen). On the BD Fortessa, SYBR Green I was excited with a 488 nm blue laser and detected by a 530/30 filter and CellTrace Far Red was excited by a 640 nm red and detected by a 670/14 filter. On the Attune NxT, SYBR Green I was excited with a 488 nm laser and detected by a 530/30 filter, and Cell Trace Far Red was excited by a 637 nm laser and detected by a 670/14 filter. 50,000 events were collected per sample. Analysis was performed using FlowJo (Tree Star, Ashland, Oregon). Initial gating was done as before to isolate the signal from a single erythrocyte. The percentage of the CellTrace Far Red labelled erythrocytes that had been infected (ensuring only erythrocytes infected after the assay was started are measured) was determined using the plots of the blue laser (SYBR Green) vs red laser (Cell trace Far Red). Unless indicated otherwise, all experiments were carried out in triplicate.

When testing the R1 peptide, even at the highest concentration we still detected some invasion, we explored whether the residual level of parasites identified by flow cytometry represented real invasion events, or merozoites stuck on the outside of erythrocytes, which could happen after blocking PfAMA1 function, and which cannot be distinguished between using flow cytometry. Using fluorescence microscopy with both the untreated control and the sample treated with 40 μM R1 peptide (the highest concentration tested), we saw both rings (star-shaped DNA signals) in erythrocytes as well as merozoites that appeared to be stuck on the outside. This indicates that some of the parasitemia signal present with the 40 μM R1 peptide after invasion is rings; we did not, however, maintain the culture to see if the parasites could develop further.

Generation of CRISPR knock-out lines

Knockouts were generated using a CRISPR/Cas9 approach. Construct design was done in Benchling and gene sequences were obtained from the PlasmoDB website (https://plasmodb.org/plasmo/app). Details of all the sequences used for each line are summarised in S3 Table. Each knock-out was constructed by replacing a region of the gene with a selectable marker. Guide sequences were selected in the region to be replaced, Fig 3A, using the CRISPR tool on Benchling. The guide plasmids for the knock-out line construction were made in the same manner as described above for the PfAMA1-mNeonGreen line. Repair constructs were created using the pCC1 plasmid (PlasmoGEM Sanger) backbone, in which homology regions (500–800 bp) targeting the region of interest were assembled on either side of the human dihydrofolate reductase (hDHFR) positive selection maker and a barcode sequence unique to the knock-out; the plasmid also contained a negative selection gene for cytosine deaminase to allow the plasmid to be removed from the edited parasites [69]. The homology regions were amplified from NF54 DNA using primers specific to the homology region with 20 bp overhangs added to allow annealing to the neighbouring fragments. The hDHFR gene was amplified from the pCC1 plasmid. The plasmid backbone was made by digesting pCC1 with SpeI and EcoRI then gel purifying the cut band. The repair template was assembled and integrated into pCC1 using Gibson assembly [107] (NEB Hifi assembly kit). Colony PCR was used to check the plasmids had the correct inserts (F-93/R-91) and (F-92/R-94), sequences in S2 Table.

Transfection was done as described above. Once transfected the parasites were cultured in the presence of 2.5 nM WR99210 (Jacobus Pharmaceuticals) until the parasites could be seen in Giemsa smears. For the lines made as part of the pipeline [110] (ΔPfEBA140, ΔPfEBA181 and ΔPfRH4) the lines were grown from stocks frozen after transfection and positively selected by growth on WR99210 then treated like the other lines. The presence of WR99210 selects for both edited parasites where the repair template containing the hDHFR gene had been integrated into the target locus and wild-type parasites carrying the pCC1 plasmid as an episome. Negative selection was therefore carried out by culturing the parasites in the presence of 5-fluorocytosine (5-FC, Sigma-Aldrich, UK) at 48 mM and 2.5 nM WR99210 for a week. The cytosine deaminase gene on the pCC1 plasmids makes the 5-FC lethal [111], so any parasites still carrying the plasmid were killed. This ensures that the only parasites remaining are the ones with the hDHFR-positive selection gene integrated into the targeted genome region.

Finally, all the lines were cloned by limiting dilution using a plaque cloning assay as described by [112]. Wells containing a single colony were inspected using an EVOS microscope (Leica, Germany), 4X objective. Finally, PCR (CloneAmp HiFi PCR Premix, Takara-Bio Europe) genotyping was done with primers designed for each knock-out line that allowed for confirming either the presence of the wild-type (F-gt5/R-gt5WT and F- gt3WT/R-gt3) or presence of the knock-out gene (F-gt5/R-91 and F-92/R-gt3), as shown in S8A Fig. Sequence given in S2 Table.

We successfully made cloned knock-outs of PfEBA140 (PlasmoDB ID PF3D7_1301600), PfEBA181 (PlasmoDB ID PF3D7_0102500), PfRH1 (PlasmoDB ID PF3D7_0402300), PfRH4 (PlasmoDB ID Pf3D7_0424200), PfRH2a (PlasmoDB ID PF3D7_1335400), PfP230P (PlasmoDB ID PF3D7_0208900) and Pfs25 (PlasmoDB ID PF3D7_1031000). We were unable to make ΔPfRH2b (PlasmoDB ID PF3D7_1335300), 4 attempts were made at transfecting using 3 different guides. On several of these occasions, other transfections were done in parallel, resulting in successful edited parasites. The same guides and repair construct worked on the first attempt when the transfection was done into the 3D7 line and correct editing was shown by genotyping indicating the constructs worked. The efficiency of disrupting PfRH2b in NF54 seems very low, whereas ΔPfRh2b lines have been made in 3D7 multiple times previously [52,69]. Given all the other knockouts were made on the NF54 background, complicating comparisons, we did not use the 3D7 ΔPfRH2b for further studies.

Quantification of gene expression with RT-PCR

Protocol available DOI: dx.doi.org/10.17504/protocols.io.q26g7peo9gwz/v1. The qPCR set-up was guided by the protocol described [77]. All the lines were synchronised in parallel with the tight synchronisation protocol described above, using a 150 ml culture at a 1% HCT. On a Thursday morning, the late-stage schizonts were isolated with Percoll as described above, a third was left to reinvade and two-thirds of the schizonts were added to 40 ml media with PKG inhibitor Compound 2 ((4-[7-[(dimethylamino)methyl]-2-(4-fluorophenyl)imidazo [1,2-α]pyridine-3-yl]pyr- imidin-2-amine)) at 1 μM at 37°C for 3.5 h. Compound 2 arrests schizonts ~15 min prior to egress [85], ensuring the samples were all comparable in their developmental stage. After C2 treatment, the schizonts were then pelleted by centrifugation, washed in 5 ml ice-cold PBS, and resuspended in 1 ml ice-cold 0.01% saponin in PBS, left on ice for 10 min then pelleted again and resuspended in 1 ml of 0.01% Saponin, pelleted and resuspended in 200 μl PBS (100 μl if the pellet was small). 2000 μl of TRIzol LS reagent (Ambion ref 10296010) was added and heated to 37°C for 5 min. We collected samples for all the lines in parallel over 4 weeks in different blood each week, collecting one sample a week with two invasion cycles in between.

1000 μl of each sample in Trizol was then added to Phasemaker tubes (Invitogen A33248) and left for 5 min, then 200 μl of chloroform was added, left for 3 mins at room temperature and then centrifuged at 12,000 g for 5 min at 4°C. The RNA was then extracted from the aqueous phase using the RNA Clean and concentrator kit (Zymo R1015) following the manufacturer’s instructions and using the on-column DNase I treatment. cDNA synthesis was performed using 1 μg of RNA per sample with the superscript IV VILO MM with ezDNase enzyme (ThermoFisher 11756050) following the manufacturer’s instructions alongside a no reverse transcription (RT) control. For each qPCR reaction, the SYBR Power SYBR Green PCR Master Mix (Fisher scientific 4367659) was used with a 10 μl volume and a 650 nM primer concentration; the sequence for primers used is given in S3 Table. For every pair of primers, a set of standards of genomic NF54 DNA was run at (10, 2, 0.4 and 0.08 ng/μl) on every plate. We ran a minimum of triplicate wells for the samples and the standards, then one for the no-RT control and water controls. Samples were pipetted into the 384-well plates using an Opentrons OT2 pipetting robot with an Opentrons Gen 2 p20 pipette to minimise technical variation, keeping the 384-well plate cooled to 10°C. Each set of samples was run over two plates, with all lines on each plate and primers for 7 genes with Actin I (housekeeping reference) repeated on both plates. The plates were run using a BioRad 384 well RT PCR machine (CFX384) at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 55°C for 20 s, 60°C for 60 s.

Analysis was done using a custom Python script. Any outliers were eliminated by removing any points of the triplicate repeat that were more than 0.3 Cq away from the mean of the data points (the measurements for 47 wells out of the 1470 measured for the cDNA of the different lines were removed representing 3.2% of the measurements). For each set of primers, we performed a linear fit to the measurements of the standards. The Cq values were interpolated based on the standard curve. We confirmed visually that there was a good separation between the RT samples and the no-RT control. Each gene’s expression levels were then normalised to the housekeeping gene Actin I in the given sample. The gene expression of each knock-out line is expressed as a relative fold change compared to the wild-type NF54 line, ΔPfP230P or ΔPfs25. To determine which changes were significant we performed a one sample t-test with the null hypothesis that the true mean of the population equals 1 at a 95% level of significance, which would represent no change in gene expression relative to the corresponding reference line.

Statistics

Error bars used throughout show the standard error of the mean or the standard deviation as indicated in the legends of the figures. The Anderson-Darling test has been used to assess if the detachment force data sets are normally distributed [113]. The test assesses how well a set of data fits a normal distribution. The null hypothesis is that the data are normally distributed. The test gives a P value of how likely the null hypothesis is true. P = 0.05 was used to determine if the dataset fulfilled the null hypothesis. If both data sets were normally distributed, as assessed with the Anderson-Darling test, the significance of the difference between them was assessed with a t-test at a 5% significance level. The test assesses the null hypothesis that the two data sets being compared are normally distributed and come from independent random samples with equal means and equal but unknown variances [114]. If one or both data sets were not normally distributed, assessed with the Anderson-Darling test, then the difference’s significance was assessed with a Wilcoxon rank test at a 5% significance level. The non-parametric test assesses the null hypothesis that the two data sets are from a continuous distribution with equal medians. The test assumes that the two samples are independent [114].