Cryogenically preserved RBCs support gametocytogenesis of Plasmodium falciparum in vitro and gametogenesis in mosquitoes

The overall study design is described below, and in the supplement accompanying this manuscript (Additional file 1). Cryo-preservation was performed 3–4 days after collection as suggested elsewhere [27, 31]. To control for the variability inherent to SMFAs [9] and thus account for any technical/biological variation arising from unknown factors independent of RBC storage method and duration, the study was originally designed with the goal of meeting two important criteria. First, gametocyte flasks for comparing each treatment were initiated from the same asexual seed culture since parasite fitness can vary over generations, which in turn can introduce variation in induction of gametocytogenesis. Second, infectious cultures resulting from the treatments were offered to mosquitoes that were sorted from the same starting population and age post-emergence to minimize any variation in sporogony caused by inter-generational differences across mosquito cohorts. Therefore, each experimental block was composed of the following steps: (1) the same asexual seed culture was used to initiate gametocytogenesis, (2) rates of gametocytaemia monitored over 14–16 days in individual flasks (i.e., 1 flask/treatment) with refrigerated (4 °C) and/or cryo-preserved RBCs at various storage durations, and (3) sexually mature gametocytes from each flask/treatment offered to mosquitoes from the same cohort to assess the efficiency of gametogenesis and sporogony.

Although the experimental design was initially centered on replicating the outcomes across 4 unique donor RBC populations while fulfilling the criteria outlined above, a variety of unforeseen logistical constraints (e.g. in vitro contamination) prevented the accomplishment of these goals (Additional file 2). As such, the experimental design was “partially nested” wherein not all 4 donors were represented equally across all treatments of storage method or duration and to account for this, all statistical analyses were performed using generalized linear mixed-effects models (GLMMs) that are particularly well-suited for analysing such datasets, as suggested previously [32‐34] (see Additional file 2 for schematic showing donor-specific testing regime within experimental blocks).

All reagents and consumables described herein were purchased from Fisher Scientific (Hampton, NH) unless noted otherwise.

Parasite culture media was prepared and stored as described previously [15], with minor modifications. Briefly, incomplete media was prepared by dissolving pre-made RPMI-1640 powdered media in distilled water before adding 2% sodium bicarbonate (w/v) and 0.005% hypoxanthine (w/v, Sigma-Aldrich, St. Louis, MO). The incomplete media was then filter-sterilized with 0.2 µm filters under vacuum before moving to storage at − 20 °C in aliquot sizes of 450 or 900 mls. Complete media was prepared just prior to use by adding 50 or 100 mls of 10% A+ non-heat inactivated human serum (Valley Biomedical, Winchester, VA) to 450 or 900 mls to thawed, incomplete media respectively. Media was dispensed into 45 ml aliquots in 50 ml conical-bottom tubes and the air–liquid interface sparged for 5–10 s with a micro-aerophilic gas mixture of 5% CO₂, 5% O₂ and 90% N₂ (referred to herein as “tri-gas mixture”, Airgas LLC, Kennesaw, GA) prior to storage for 1 week. Since “quality” of serum is critical for culturing transmission competent parasites, serum samples were tested as described previously [15]. Serum from 14 individuals was pooled into pairs and tested for their ability to support gametocytogenesis as well as ex-flagellation of male gametocytes in case of P. falciparum isolate NF54 (unpublished observations). The same pool of serum donors was used to culture P. falciparum NF54 for the duration of this study.

Cryo-preservation of RBCs was performed 3–4 days after collection as suggested elsewhere [27, 31]. Additionally, it should be noted that due to practical constraints, the comparison between ≤ 1-week old refrigerated and cryo-preserved RBCs were performed after the latter had been stored for 2–3 days.

Procedures for freezing and thawing were adapted from Goheen et al. [28] and Sputtek [31] with minor modifications. Two-fold concentrated stocks of cryo-protectant were prepared in deionized water by warming a solution of 28% glycerol (v/v), 3% sorbitol (v/v from a 1 M stock) and 0.65% (w/v) sodium chloride (NaCl) prior to sterilization with a 0.2 µm filter and storage at room temperature. Whole blood (~ 500 ml units, Valley Biomedical, Winchester, VA or Interstate Blood Bank, Memphis, TN) was dispensed in 45 ml aliquots and either stored refrigerated at 4 °C until use (see next section) or allowed to warm to room temperature prior to cryo-preservation. To ensure reproducible recovery of viable RBCs following cryo-preservation, it is critical for the cryo-protectant and RBCs to be at the same temperature (e.g., room temperature), prior to use [31].

Aliquots of whole blood, no more than 3–4 days post-collection, were centrifuged at 1800×g for 10 min at room temperature at a low brake setting and plasma and white blood cell layers aspirated under vacuum before an equal volume of cryo-protectant solution (~ 20–25 mls) was added to the packed RBC pellet to achieve a final glycerol concentration of 14% (v/v) and a haematocrit of ~ 50%. The packed RBCs were then equilibrated with the cryo-protectant for 15–20 min with gentle intermittent mixing. Glycerolized RBCs were then dispensed in 2 ml aliquots and snap-frozen by immersing in liquid nitrogen for 2–3 min before transferring to the vapour phase for long-term storage. Cryo-preserved RBCs were thawed just before use by incubating cryogenic vials in a dry bath for 5 min at 37 °C before transferring the contents into a 15 or 50 ml conical bottom tube prior to de-glycerolization. Unless stated otherwise, all solutions for de-glycerolization were prepared in deionized water and sterilized with 0.2 µm filters before storage at room temperature. De-glycerolization was initiated with the addition of successive gradients of sodium chloride concentrations starting with 0.4 mls of 12% NaCl (w/v in distilled water, 0.2 volumes for every ml of RBC and cryoprotectant mixture) added dropwise under gentle agitation (~ 1 min). The mixture was then incubated at room temperature for 5 min with gentle intermittent mixing before dropwise addition of 10 mls (5 volumes) of sterile 1.6% NaCl (w/v in distilled water). The de-glycerolized RBC suspension was centrifuged at 1000×g for 3 min at 20–23 °C and low brakes and 20mls (10 volumes) of a salt-dextrose solution (0.9% NaCl + 0.2% dextrose, w/v in distilled water) added dropwise to the packed RBC pellet under gentle agitation. If haemolysis was still observed in the supernatant, the packed RBCs were washed once more with the same volume of salt-dextrose solution. Finally, the thawed RBCs were resuspended in complete media before use. Recoveries of ~ 0.8–0.9 mls of packed RBCs should be considered routine.

Plasmodium falciparum isolate NF54 was obtained from BEI Resources, NIAID, NIH (‘Plasmodium falciparum, Strain NF54 (Patient Line E), product number MRA-1000, contributed by Megan G. Dowler’). Asexual feeder cultures were routinely initiated at ring state parasitaemia of 0.1–1% using cryo-preserved RBCs pooled from 2 to 4 donors as substrate and final haematocrit of 5% in 5 or 15 mls of complete media for T25 or T75 flasks, respectively. The same volume of media was replaced every 24 h and then sparged gently at the air–liquid interface with the tri-gas mixture for 15–20 s. Parasitaemia was monitored every 1–2 days by transferring 50–100 µl of parasite culture to a 0.6 ml tube, centrifuged at 1800×g for 1 min and supernatants aspirated until final culture volume was ~ 15–20 µl. The RBC pellet was re-suspended with repeated pipetting before preparing smears with 2 µl of the concentrate on glass slides. Slides were dried on a slide-warmer for 20–30 s and then fixed by submerging in 100% Methanol for 10 s prior to Giemsa staining. Concentrated Giemsa (Sigma-Aldrich, St. Louis, MO) was filtered in aliquot sizes of 30–35 mls with a Whatman no. 1 filter paper and stored at room temperature. Just before use, the filtered stain was diluted to a ratio of 1:20 (v/v) in phosphate-buffered saline (pH 7.2) and fixed slides immersed in the stain in glass Coplin-jars for 10–15 min at room temperature. Stained sides were washed under running tap-water for 5–10 s with the smeared side facing downwards before being air-dried for microscopy. Smears were mounted directly in immersion oil (Cargille Labs, Cedar Grove, NJ) and examined at 1000× magnification with a Leica DM2500 upright microscope (Leica Microsystems, Buffalo Grove, IL). Parasitaemia was estimated from ~ 1000 RBCs and staged into early trophozoites (referred to herein as rings), late trophozoites and schizonts following guidelines described previously [13, 35]. Ring-stage parasitaemia expressed as a proportion of RBCs counted was used to prepare flasks for routine sub-culture and/or gametocytogenesis [15].

Flasks destined for gametocytogenesis were only initiated when ring-stage parasitaemia in the asexual seed flasks above showed exponential increase in growth rates (> 6- to 8-fold) relative to the previous sampling point. Flasks for gametocytogenesis were prepared essentially as described above with 1 flask representing each treatment (Additional file 5a). Briefly, freshly thawed or refrigerated RBCs resuspended in pre-warmed (37 °C) complete media at 5% haematocrit (e.g., 1 mls naïve RBCs in 20 mls complete media) were seeded with asexual cultures to achieve a final ring-stage density of 0.6–1% in the gametocyte flasks and returned to an incubator maintained at 37 °C. Media in flasks was replaced every 24 h with 15 mls of fresh pre-warmed (37 °C) complete media and sparged with tri-gas mixture for 10–30 s. All manipulations outside the incubator were performed by placing the flask on a slide-warming platform maintained at 38 °C (MedSupply Partners, Atlanta, GA). Gametocyte flasks were maintained for 14–16 days and parasitaemia monitored with Giemsa staining at 1-, 4-, 8-, 12- and 14-days post-seeding as suggested previously [15]. Asexual stages were staged as described above and gametocytes were further classified into stages II, III, IV and V male or V female gametocytes using guidelines established previously [13, 35] (Additional file 3a, b). Stage I gametocytes were excluded from the counts as they are virtually indistinguishable morphologically from late trophozoites, especially in asynchronous cultures [36]. Finally, maturation status of the culture was assessed from 12 days post-seeding by quantifying ex-flagellation of male stage V gametocytes in a Neubauer haemocytometer chamber as suggested previously [15]. Briefly, after adding fresh media, 10 µl of culture was pipetted directly into the chambers of a haemocytometer and incubated at 21–24 °C for 15–20 min before quantification (Additional file 3c). Imaging was performed at 400× magnification on a Leica DM2500 equipped with differential interference contrast (DIC) optics. A flask was deemed infectious once ex-flagellation was observed and offered to mosquitoes within 24–48 h.

Plasmodium falciparum Cambodian isolate CB132 (kind gift of Prof. Dennis E Kyle, University of Georgia, Athens, GA, USA) [30] was cultured under similar conditions to NF54 with the exception that flasks destined for gametocytogenesis were seeded at a density of 0.5%.

Anopheles stephensi mosquitoes were maintained in a level 2 Arthropod Containment Laboratory at the University of Georgia, which were initiated from eggs kindly provided by the Walter Reed Army Institute of Research ca. 2015 is a wild-type strain referred to as Strain “Indian” [10, 37]. Colonies were housed in a dedicated walk-in environmental chamber [chamber (Percival Scientific, Perry, IA)] at 27 °C ± 0.5 °C, 80% ± 5% relative humidity, and under a 12 h light: 12 h dark photo-period schedule. Adult mosquitoes were maintained on 5% dextrose (w/v) and 0.05% para-Aminobenzoic acid (w/v) and provided whole human blood offered in glass-jacketed feeders (Chemglass Life Sciences, Vineland, NJ) through parafilm membrane maintained at 37 °C to support egg production. Husbandry procedures were established according to guidelines as suggested elsewhere [38] with minor modifications. Briefly, eggs were rinsed twice with 1% house-hold bleach (v/v, final concentration of 0.06% sodium hypochlorite) before surface-sterilization for 1 min in the same solution at room temperature. Bleached eggs were washed with 4–5 changes of deionized water and transferred to clear plastic trays (34.6 cm L × 21.0 cm W × 12.4 cm H) containing 500 ml of deionized water and 2 medium pellets of Hikari Cichlid Gold fish food (HikariUSA, Hayward, CA) and allowed to hatch for 48 h. Hatched L1 larvae were dispensed into clear plastic trays (34.6 cm L × 21.0 cm W × 12.4 cm H) at a density of 300 larvae/1000 ml water and provided the same diet until pupation. The feeding regime consisted of 2 medium pellets provided on the day of dispensing (day 0) followed by the provision of a further 2, 4, 4 and 4 medium pellets on days 4, 7, 8 and 9 respectively. This regime allows > 85% larval survival and > 90% pupation of the surviving larvae within 11 days with a sex ratio of 1:1 adult males and females (unpublished observations).

All steps outlined below were performed with pre-warmed (38 °C) equipment, including tubes, pipettes, pipette tips and centrifuge rotor buckets. Cultures deemed infectious were concentrated by centrifugation at 1800×g for 1–2 min at room temperature in pre-weighed 15 or 50 ml conical bottom centrifuge tubes. Supernatants were aspirated under vacuum and the weight/volume of the concentrate estimated by weighing the tube and subtracting the value of the pre-weighed empty tube. 3–6 volumes of a freshly washed and pre-warmed mixture of RBCs resuspended in freshly thawed serum (30% haematocrit) was added to the infected RBC pellet to achieve a final haematocrit of 45–50% and mature gametocytaemia of ~ 0.1%. The volumes of naïve RBCs and serum mixture to be added was informed by the densities recorded in the gametocyte flasks on the day of infection. The mixture of naïve and parasitized RBCs was then fed immediately to 60–80 (3–7 day old) female An. stephensi mosquitoes for 15–20 min. To facilitate high blood feeding rates, mosquitoes were previously starved for 16–24 h in dedicated environmental chambers programmed to fluctuate a total of 9 °C around a daily mean of 24 °C with 80% ± 5% relative humidity and a 12 h light:dark photo-period schedule (Percival Scientific, Perry, IA) as described previously [10]. After the blood-feeds, feeding status was qualitatively ascertained and engorged mosquitoes selected for by extending the starvation period for a further 24–48 h to eliminate non-blood-fed mosquitoes [39].

To quantify parasite densities in the infectious feed, Giemsa-stained smears were prepared from a 2 µl aliquot of the infectious blood meal offered to mosquitoes, as described above. Prevalence and abundance of parasites in the midguts of mosquitoes was assessed at 9–13 days post-infection by dissecting midguts and counting oocysts at 400× magnification with a Leica DM2500 under DIC optics (Additional file 3d). Sporozoite prevalence was assessed at 17–21 days post-infection by dissecting salivary glands into 5 µl of PBS. Glands were ruptured by overlaying a 22 mm² coverslip and presence/absence of sporozoites was recorded for each mosquito at either 100× or 400× magnification with the same microscope (Additional file 3e).

All data analyses were performed in RStudio (version 1.1.423) [40] running the statistical software package R (version 3.5) [41]. Choice of GLMM family and corresponding link function were based on five dependent variables modelled—(1) rates of mature gametocytaemia in vitro, (2) oocyst prevalence (proportion of mosquitoes with oocysts on the midgut), (3) oocyst abundance (total number of oocysts per midgut regardless of infection status), (4) oocyst intensity (number of oocysts per midgut from infected midguts only), and (5) sporozoite prevalence (proportion of mosquitoes with sporozoites in the salivary glands). For all the analyses, storage method (refrigerated vs. cryo-preserved RBCs) and duration (4, 8, and 12 weeks post cryo-preservation) were specified as independent variables, with storage method and duration classified as a categorical and continuous predictor, respectively. Storage duration was centred and scaled over the grand mean. For analysing temporal patterns of gametocytogenesis in vitro, GLMMs with a binomial distribution (“logit” link) were performed with the total count of RBCs infected with male and female stage V gametocytes (i.e., total mature gametocytes) expressed as a proportion of the sum of RBCs (infected and uninfected) counted from each sample as the dependent variable. Oocyst and sporozoite prevalence in the midguts and salivary glands of mosquitoes, respectively, were modelled as the probability of mosquitoes being infected and infectious as dependent variables in GLMMs specified with a binomial distribution (“logit” link function). For analysing oocyst abundance and intensity, GLMMs with negative binomial distributions (“log” link) were utilized. All statistical analyses were performed using the package “lme4” unless stated otherwise [42]. For the in vitro analysis, the intercepts of the relationships between the rates of mature gametocytaemia and the predictor variables was allowed to vary between RBC donors as well as over sampling period (days post-seeding) between experimental blocks and flasks nested within each block. The same random effect structure was specified for analysing parasite fitness in the mosquitoes except infections/SMFAs were nested within blocks in lieu of flasks (Additional files 1 and 2). Tabulation of model outputs, along with tests for overdispersion were performed with the “sjstats” package. Where applicable, post hoc comparisons of storage method and duration were performed using Tukey contrasts of the estimated marginal means derived from the models above, as suggested in the “emmeans” package [43].

Finally, cross-validation of the models were performed by (1) testing fit after truncating the number of mosquitoes sampled across each infection and (2) removing experimental block(s) and/or flasks/infections within each block. Where applicable, random effects were checked for normal distribution patterns and the final output tables of statistical modelling prepared with the “sjPlot” package and “sjstats” packages [44, 45]. All figures were prepared using the “ggplot2” package [46].

The influence of storage method (cryo-preserved RBCs vs. the reference standard of refrigerated RBCs ≤ 1-week post-collection) was assessed for the rates of gametocytogenesis in vitro (Fig. 1, Additional file 3a and b for images of mature gametocytes and Additional file 4 for developmental rates of preceding immature stages II–IV). Mature gametocytaemia in vitro generally increased over time with detectable levels achieved by 12 days post-seeding (Fig. 1). Overall trends were independent of storage method (Fig. 1a and Table 1, Odds ratios (OR) = 1.31, standard error (se) = 0.41, p = 0.4) or duration (Fig. 1b and Table 1, OR = 1.01, se = 0.04, p = 0.72) with most of the random variation explained by differences between experimental block (Table 1 and Additional file 5a) and negligible contribution by RBC donor (Fig. 1b, Table 1 and Additional file 5a).

To determine if the sexually mature gametocytes generated in vitro are infectious, sporogony in An. stephensi mosquitoes was assessed, starting with oocyst prevalence in the midguts (infected/total midguts) at time points corresponding to peak stages of infection (Additional file 3d). The mean prevalence of mosquitoes infected with P. falciparum oocysts (NF54) was 53 ± 6.4% (mean ± se, n = 11 SMFAs, mean mosquito midguts sampled/SMFA = 26 (range = 17–44), total number of mosquito midguts sampled = 283) with 59 ± 4.6% (n = 2) and 52 ± 7.9% (n = 9) in mosquitoes infected with parasites cultured in refrigerated and cryo-preserved RBCs, respectively (Fig. 2a, left panel). While prevalence was independent of storage method (Fig. 2a, left panel and Table 1, OR = 1.05, se = 0.38, p = 0.89), the ability of cryo-preserved RBCs to culture infectious gametocytes showed a marginally insignificant decline with duration of storage (Fig. 2b, left panel and Table 1, OR = 0.89, se = 0.06, p = 0.07) with much of the unexplained variation represented by differences across experimental blocks (Table 1 and Additional file 5b), although some could also be attributed to RBC donor characteristics (Fig. 2b, left panel, Table 1 and Additional file 5b).

To verify sporozoite egress from the oocyst and completion of the sporogonic cycle, salivary glands from the same cohort of mosquitoes were dissected and scored for the presence or absence of sporozoites (Additional file 3e). The mean (± se) prevalence of mosquitoes carrying sporozoites of P. falciparum NF54 in the salivary glands was 54 ± 5.3% (n = 10 SMFAs, mean salivary glands sampled/SMFA = 24 (range = 20–30), total number of salivary glands sampled = 235) with 54 ± 4.2% (n = 2) and 54% ± 6.8% (n = 8) in mosquitoes infected with parasites cultured in refrigerated and cryo-preserved RBC, respectively (Fig. 2a, right panel). Overall trends in sporozoite prevalence were independent of storage method (Fig. 2a, right panel and Table 1, OR = 1.35, se = 0.84, p = 0.63) or duration (Fig. 2b, right panel and Table 1, OR = 0.93, se = 0.08, p = 0.36). Unlike the midguts however, most of the random variation was caused by differences among infections with negligible contributions from experimental block or RBC donor (Fig. 2b, right panel and Table 1).

Oocyst abundance and intensity were used to determine if parasite burdens in the midguts of mosquitoes differed due to storage method or duration. Mean (± se) oocyst abundance (oocyst counts from all sampled midgut) was 5.4 ± 2.53 (n = 11 SMFAs, mean mosquito midguts sampled/SMFA = 26 (range = 17–44), total number of mosquito midguts sampled = 283) while oocyst intensity (oocyst counts from infected midguts only) was 7.75 ± 2.85 (n = 11 SMFAs, range of infected midguts/SMFA = 1–28, total number of mosquito midguts sampled = 157) (Fig. 3). Overall, oocyst abundance (mean ± se) and intensities were 2.73 ± 0.434 (n = 2) and 4.5 ± 0.559 (n = 2) respectively in mosquitoes infected with parasites cultured in refrigerated RBCs and 5.72 ± 0.754 (n = 2) and 10.6 ± 1.23 (n = 9) respectively for parasites in cryo-preserved RBCs.

While both abundance and intensity were independent of storage method (Incidence rates ratio for abundance (IRR) = 1.20, se = 0.71, p = 0. 80, IRR for intensity = 1.12, se = 0.36, p = 0.75) or duration (IRR for abundance = 0.86, se = 0.1, p = 0.13, IRR for intensity = 0.95, se = 0.07, p = 0.48) (Table 1), most of the random variation was generated by the differences between experimental blocks with some contribution from RBC donor-specific characteristics and the least from the variation among infections (Table 1).

To determine if cryo-preservation can, in principle, support infectious cultures of other strains of P. falciparum, RBC donor number four used for the NF54 cultures above (Additional file 2) was used as a substrate for sexual stage cultures of a clinical isolate of Cambodian origin (P. falciparum CB132 [30]) following storage periods of six (n = 35 midguts and 23 salivary glands) and eight weeks (n = 49 midguts) (Additional file 6). Statistical analyses were not performed due to the lack of replication in the study design.