## Company Of Biologists

Band 3 modifications in Plasmodium falciparum-infected AA and CC erythrocytes assayed by autocorrelation analysis using quantum dots
Fuyuki Tokumasu, Rick M. Fairhurst, Graciela R. Ostera, Nathaniel J. Brittain, Jeeseong Hwang, Thomas E. Wellems, James A. Dvorak

## Summary

The molecular stability of hemoglobin is critical for normal erythrocyte functions, including oxygen transport. Hemoglobin C (HbC) is a mutant hemoglobin that has increased oxidative susceptibility due to an amino acid substitution (β6: Glu to Lys). The growth of Plasmodium falciparum is abnormal in homozygous CC erythrocytes in vitro, and CC individuals show innate protection against severe P. falciparum malaria. We investigated one possible mechanism of innate protection using a quantum dot technique to compare the distribution of host membrane band 3 molecules in genotypically normal (AA) to CC erythrocytes. The high photostability of quantum dots facilitated the construction of 3D cell images and the quantification of fluorescent signal intensity. Power spectra and 1D autocorrelation analyses showed band 3 clusters on the surface of infected AA and CC erythrocytes. These clusters became larger as the parasites matured and were more abundant in CC erythrocytes. Further, average cluster size (500 nm) in uninfected (native) CC erythrocytes was comparable with that of parasitized AA erythrocytes but was significantly larger (1 μm) in parasitized CC erythrocytes. Increased band 3 clustering may enhance recognition sites for autoantibodies, which could contribute to the protective effect of hemoglobin C against malaria.

## Introduction

Individuals with mutant hemoglobins (Hb) are protected against life-threatening manifestations of Plasmodium falciparum malaria. For example, hemoglobin C (β6: Glu to Lys) markedly reduces the risk of severe malaria in West Africa (up to 90% reduction with homozygous HbCC) (Agarwal et al., 2000; Modiano et al., 2001). It has been shown that frequent intra-erythrocytic parasite death may underlie the lower multiplication rates of P. falciparum in CC erythrocytes in vitro (Fairhurst et al., 2003). In addition, the formation of aberrant knobs on the surface of parasitized CC erythrocytes suggests an abnormal membrane remodeling process. It is possible that alterations in the quaternary structure of HbC may expose the heme moiety to oxidative damage, resulting in marked HbC denaturation, decreased solubility compared to HbA and a propensity to form HbC crystals in vivo (Fabry et al., 1981; Hirsch et al., 1985). The structural abnormalities of CC erythrocytes may make them more susceptible to host immune responses such as occur in hemoglobinopathic and normally aged erythrocytes (Ando et al., 1997; Schluter and Drenckhahn, 1986; Turrini et al., 1991). However, details of these membrane modifications in CC cells due to abnormal Hb and parasite infection have only recently been quantified (Arie et al., 2005).

Intracellular denaturation of Hb can cause significant changes in the erythrocyte membrane. For example, in Heinz body hemolytic anemia denatured hemoglobin precipitates to form insoluble hemichromes that bind to the inner surface of erythrocytes (Jacob and Winterhalter, 1970; Rachmilewitz, 1969; Rachmilewitz, 1974; Rifkind et al., 1994; Schneider et al., 1972; Winterbourn and Carrell, 1974). The consequent release of reactive oxygen intermediates increases membrane permeability and induces premature hemolysis (Hebbel, 1990). Hemichrome formation is also associated with the aggregation of membrane proteins during normal erythrocyte senescence (Low et al., 1985; Waugh and Low, 1985; Waugh et al., 1986). In sickle cell disease, the instability of HbS results in premature senescence of erythrocytes, as suggested by increased amounts of membrane-bound autologous IgG (Bosman, 2004). Aggregation of band 3 is thought to form a neoantigen that binds autologous immunoglobulin (IgG) and complement, thereby targeting aged erythrocytes for destruction (Ando et al., 1997; Liu et al., 1991; Lutz et al., 1984; Schluter and Drenckhahn, 1986; Turrini et al., 1991). Whether band 3 aggregates contribute to the malaria-protective effect of CC erythrocytes has not been determined.

To investigate this possibility, we monitored band 3 clustering in P. falciparum-infected AA and CC erythrocytes using a newly developed quantum-dot-based immunochemical technique (Tokumasu and Dvorak, 2003). The technique permits 3D reconstruction and quantification from a z-stack series of undistorted erythrocytes without fluorescence intensity loss caused by photobleaching. Here, we present the results of single-cell studies of total fluorescence intensity and distribution of band 3 in uninfected and infected erythrocytes using power spectra and 1D autocorrelation analyses. We found that the accessibility of an intracellular band 3 epitope to antibody was markedly reduced in CC compared to AA erythrocytes, indicating a high degree of band 3 clustering in CC erythrocytes in vivo and during intracellular parasite development in vitro.

## Materials and Methods

### Erythrocyte preparation and parasite culture

The preparation, hemoglobin phenotype determination, and P. falciparum (7G8 line) infection of AA and CC erythrocytes have been described previously (Fairhurst et al., 2003). Parasite cultures at 5% parasitemia were used.

### Membrane extracts and immunoblotting

Uninfected (native) and parasitized erythrocytes were washed in HEPES buffer, pH 7.4 (10 mM HEPES, 140 mM NaCl, 10 mM glucose (the conventional notation, rather than the recommendation in NIST-SP811, is employed for this publication with M representing mol/l). Erythrocytes were lysed and the membrane pellets were washed extensively with ice-cold Tris-HCl/EDTA pH 8.0, in the presence of protease inhibitors (Complete™, Roche Diagnostics, Germany). Membrane proteins were extracted from the pellets using mild detergent conditions as described (Giribaldi et al., 2001). Band 3 protein in the membrane extracts was analyzed using a 4-12% Bis-Tris acrylamide gel under non-reducing conditions (Invitrogen Corporation, Carlsbad, CA). Each well was loaded with 0.1 μg of total protein, except for CC erythrocytes where 0.3 μg of protein was used. Following electrophoresis, proteins were transferred to a PVDF membrane and probed with a monoclonal antibody (mAb) against the cytoplasmic domain of band 3 (cdb3) (Sigma-Aldrich, St Louis, MO) diluted to 1:5000 in casein-based blocking buffer (Sigma-Aldrich). Following incubation for 1 hour at room temperature, the membrane was washed in PBS-0.2% Tween-20, incubated with 1:50,000 HRP-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) and developed with Super Signal® West Pico chemical luminescence solutions (Pierce, Rockford, IL).

### Size exclusion HPLC and ELISA of uninfeced AA and CC membrane extracts

The size exclusion HPLC (SE-HPLC) of solubilized AA and CC membrane extracts was performed using a 7.5×300 mm TSK-gel G4000SW column (Toso-Haas, Montgomeryville, PA) fitted to an AKTA purifier-10 system (Amersham-Pharmacia, Piscataway, NJ). The running buffer was 5 mM sodium phosphate, 100 mM NaCl, and 0.1% C12E8 (octaethyleneglycol mono-n-dodecyl ether) detergent (Sigma), pH 7.0 as described (Rettig et al., 2001). The gel filtration standard used contained a mixture of molecular weight markers ranging from 670 to 1.35 kDa (Bio-Rad Laboratories, Hercules, CA). Protein elution was recorded at 220 and 280 nm. ELISA assays were performed on Reacti-Bind EIA microplates (Pierce) using anti-band 3 mAb at 1/5000 or 1/2000 dilutions (Sigma-Aldrich) followed by HRP-conjugated goat anti-mouse IgG at 1/20,000 dilution (Jackson ImmunoResearch Laboratories, Inc.). HRP substrate was incubated for 25 minutes before stopping the reaction with 1 N H2SO4. Reaction absorbance of the plate was read at 450 nm.

### Immunolabelling band 3 protein with quantum dots

Coverglass treatment, erythrocyte fixation and quantum dot immunolabelling were performed as described (Tokumasu and Dvorak, 2003), except that QD-585 quantum dots (Quantum Dot Corporation, Hayward, CA) were used instead of QD-605 quantum dots and 1% HEPES was added to the RPMI-1640 washing medium. In brief, erythrocytes immobilized on cover glasses using Alcian Blue (Sigma-Aldrich) were crosslinked with 50 mM dimethyl suberimidate (DMS) (Sigma-Aldrich) in 100 mM sodium borate buffer, pH 9.5, containing 1 mM MgCl2 for 1 hour, followed by 2% paraformaldehyde (Electron Microscopy Science, Fort Washington, PA). The crosslinking reaction was quenched using 0.1 M glycine in PBS, pH 7.4 for 1 hour. Erythrocytes were blocked for 1 hour in PBS-0.2% Tween 20, 3% BSA and incubated with anti-band 3 mAb (1/3000) in the same blocking solution. Erythrocytes were washed with PBS-0.2% Tween 20 three times and then incubated with biotin-conjugated goat anti-mouse IgG (Sigma-Aldrich) and further reacted with 10-15 nM of streptavidin-conjugated quantum dots (QD585) for 30 minutes. Hoechst 33258 (Molecular Probes, Eugene, OR) was added to stain parasite nuclei during the final washing step.

### Single cell fluorescence analysis Fluorescence microscopy

Wide-field fluorescence images were obtained with a Leica DM IRE2 inverted microscope (Leica Microsystems, Bannockburn, IL) using a 100× NA 1.4 PlanApochromat objective, 1.5× intermediate magnification, a custom-made QD-585 fluorescence filter set (ex. 430DF60/20, em. 580DF20, Omega Optical, Brattleboro, VT), and an XF136-2 filter module (Omega) for Hoechst 33258. Z-stack images (70 frames) were acquired with a Hamamatsu Orca digital camera (Hamamatsu Photonics System, Bridgewater, NJ) using Image Pro and Scope Pro ver. 4.5 (Media Cybernetics, Silver Spring, MD).

### Image deconvolution, fluorescence measurements and cluster analyses

Three-dimensional image reconstruction from 70-plane image stacks, total fluorescence measurements and multi-channel, blind deconvolution were performed using Auto Deblur ver. 9.3 (AutoQuant Imaging, Watervliet, NY). The deconvolved images were stored in an uncompressed AVI format. Total erythrocyte fluorescence was obtained as a sum projection of the image stack. The fluorescence intensity from each parasitized erythrocyte was normalized against the mean fluorescence intensity of multiple uninfected erythrocytes present in the same image field. Power spectra, custom-built 1D autocorrelation analyses, and statistical analyses including the fitting of autocorrelation data were performed in Origin ver. 7.5 (OriginLab, Northampton, MA).

## Results

### Immunoblotting and size-exclusion HPLC elucidates marked differences in band 3 oligomerization between AA and CC erythrocytes

In uninfected (native) erythrocyte membranes, band 3 exists predominantly as dimers; alterations of these higher order structures due to the presence of mutant hemoglobins or P. falciparum infection indicate changes in membrane structure. To determine whether hemoglobin C influences the membrane structure of both uninfected and parasitized erythrocytes, we used an anti-band 3 monoclonal antibody (mAb) to elucidate the degree of band 3 oligomerization under non-reducing conditions (Fig. 1). Immunoblot analyses of membrane protein extracts showed that the majority of band 3 in uninfected AA erythrocytes is in the dimeric form with a molecular weight (MW) of 190 kDa; small amounts of higher MW species were also present but the bands were not distinct. The oligomer to dimer ratio increased twofold in parasitized AA erythrocytes (Fig. 1A, arrow). Moreover, higher levels of the monomeric form (90 kDa) were also found suggesting that breakdown of oligomers occurs during parasite infection. Parasitized AA erythrocytes presented a slightly lower MW for both the monomer and dimer, possibly due to differential glycosylation of band 3.

Fig. 1.

Analysis of band 3 clustering in uninfected and parasitized erythrocytes. (A) Immunoblot of membrane extracts from both uninfected and parasitized erythrocytes using a mAb against the band 3 cytoplasmic domain (cdb3). All lanes, except for parasitized CC erythrocytes, were loaded with 0.1 μg of total protein. 0.3 μg of total protein from parasitized CC erythrocytes was used to obtain approximately the same amount of signal from band 3 protein. The arrows point out the signal from band 3 oligomers found in parasitized AA and CC erythrocytes. Much greater oligomerization is apparent in the lane of parasitized CC erythrocytes. (B) Quantification of band 3 in SE-HPLC fractions of AA and CC erythrocyte membrane extracts. The amount of protein used in CC extracts (0.63 μg/μl) was approximately double that in AA extracts (0.33 μg/μl). Band 3 in the SE-HPLC fractions was semi-quantified by ELISA using mAb against cdb3 at 1/5000 and 1/2000 dilutions. The anti-band 3 signal obtained for AA extracts is markedly higher than for CC extracts. In a competition ELISA, increasing the amount of mAb caused only a minor increase in anti-band 3 signals.

A higher MW band (300 kDa) that appeared in parasitized AA erythrocytes was also observed in uninfected CC erythrocytes (Fig. 1A), indicating that CC erythrocytes have altered band 3 intermolecular interactions in vivo. The oligomer to dimer ratio increased 2-fold in parasitized CC erythrocytes (Fig. 1A). In addition, a continuous distribution of higher MW band 3 species was found in parasitized CC erythrocytes, implying that either a variety of oligomers were formed or the proteins were markedly aggregated. However, the monomer to dimer ratio remained unchanged (data not shown). We also determined that a higher amount of total CC erythrocyte protein compared to AA erythrocyte protein was required to obtain an equivalent amount of total band 3 signal, implying that mAb accessibility and/or the amount of available epitope is lower in CC compared to AA erythrocytes.

To explore this possibility, we fractionated uninfected AA and CC erythrocyte membrane extracts by SE-HPLC and used an ELISA competition assay to measure their reactivity with anti-band 3 mAb (Fig. 1B). We observed reduced mAb reactivity in CC compared with AA membrane extracts even though the amount of total protein in the CC fractions (0.63 μg/μl) was twice that in the AA fractions (0.33 μg/μl). Increasing the amount of anti-band 3 mAb for ELISA assays of CC fractions resulted in a minor increase in signal; however, the signal in the CC fractions remained markedly lower compared with AA fractions. This finding may be explained by relatively high levels of hemichromes deposited on the cytoplasmic aspect of CC erythrocyte membranes (R.M.F. et al., unpublished data), where they could interfere with the ability of our anti-band 3 mAb to bind cdb3. In addition, the amount of higher MW oligomeric forms in CC fractions was slightly increased, confirming our immunoblot data. Dimers and higher MW species of band 3 proteins were not completely dissociated under reducing conditions (data not shown), as reported previously (Ando et al., 1997).

### AA and CC erythrocytes display marked morphological differences

As immunoblot data showed marked alterations in band 3 oligomerization between AA and CC erythrocytes, we compared their band 3 distribution patterns using a quantum dot-based immunofluorescence assay. Reconstructed 3D images of uninfected AA erythrocytes revealed smooth, biconcave shapes (Fig. 2A). By contrast, uninfected CC erythrocytes appeared wrinkled and withered when prepared under identical conditions (Fig. 2B). As shown previously (Smith and Krevans, 1959), some CC erythrocytes were microspherocytic. At the early (ring) stage of parasite development, neither AA nor CC erythrocytes showed significant morphological changes (Fig. 2C,F). At the trophozoite stage, however, both AA and CC erythrocytes began rounding up and losing their biconcave shape (Fig. 2D,G). This change was more pronounced in CC erythrocytes. At the schizont stage, both AA and CC erythrocytes were completely rounded up and their projected areas decreased (Fig. 2E,H).

Fig. 2.

Immunocytochemistry of infected erythrocytes. Membranes were immunolabeled with quantum dots for anti-band 3 mAb (red) and parasite nuclei were labeled with Hoechst 33258 (blue). These images are the result of 3D reconstruction and deconvolution. (A) Parasitized AA erythrocytes surrounded by uninfected erythrocytes. (B) Parasitized CC erythrocytes surrounded by uninfected erythrocytes. Note the abnormal membrane appearance of uninfected CC erythrocytes. Comparisons of erythrocyte shapes at early, mid, and late stages of parasitized AA (C-E) and CC erythrocytes (F-H). Bars, 5 μm (A,B); bars, 1 μm (C-H).

### Quantum dot-based fluorescence analyses show reduced band 3 accessibility in parasitized and hemoglobin C-containing erythrocytes

To quantify the effects of parasitization on erythrocyte membranes, we measured the total fluorescence intensity of individual erythrocytes by a quantum dot-based immunofluorescence assay using anti-band 3 mAb. In contrast to routine immunochemical methods, our preparative procedure does not lyse erythrocytes and thus allows us to measure the total fluorescence intensity of individual erythrocytes. The high photostability of quantum dots (Tokumasu and Dvorak, 2003; Wu et al., 2003) eliminated the photobleaching problem normally encountered in this type of analysis. For each erythrocyte, we collected a total of 70 z-section serial images and combined them into a single 3D image. To confirm that hemoglobin absorption did not affect the fluorescence intensity of our mAb signal, we first quantified the fluorescence intensity of each z-stack (Fig. 3A). We found that, in uninfected erythrocytes oriented parallel to the plane of the glass substrate, the fluorescence intensities of the top and bottom frames were statistically indistinguishable resulting in an overall symmetric intensity profile (skewness=0.36<0.53 at the 95% confidence level, n=51 frames). These results demonstrate that the quenching of band 3 fluorescence intensity due to hemoglobin absorption is negligible.

Fig. 3.

Fluorescence intensity measurements from uninfected and parasitized erythrocytes. (A) Representation example of fluorescence symmetry through a z-stack series of a horizontally oriented, healthy AA erythrocyte. Although the z-stack consisted of 70 frames, only those frames containing the erythrocyte are shown. Each fluorescence intensity profile was centered along the vertical axis for representational purposes. Owing to the high photostability of quantum dots, no measurable reduction in fluorescence intensity occurred throughout the z-stack sequence. (B) Cell-by-cell comparison of mean total fluorescence intensity occurring during the development of malaria parasites in AA and CC erythrocytes. The total fluorescence was measured using 3D-reconstructed, sum projection images. Data were normalized against uninfected erythrocytes (U) for each genotype. Error bars represent the standard error of the mean (s.e.m.). n=3-11 (see Table 1).

We then compared the total band 3 fluorescence intensities of uninfected and parasitized AA and CC erythrocytes (Fig. 3B). To correct for any possible irregularities in field illumination, the fluorescence intensity of each parasitized erythrocyte was normalized against the mean fluorescence intensity of multiple uninfected erythrocytes present in the same microscopic field. During the early and mid stages of infection, the fluorescence intensity of AA erythrocytes decreased slightly (10%). There was no significant difference between early (ring) and mid (early trophozoite) stages. However, a 40% decrease in fluorescence intensity occurred in late (late trophozoite and schizont) stages (Fig. 3B, Table 1). The reduction in fluorescence intensity was not apparent qualitatively. In fact, the reduction in projected area of late stage parasitized erythrocytes resulted in quantum dots being confined to a smaller area, giving the false impression that fluorescence intensity increased. By contrast, reductions in the fluorescence intensity of parasitized CC erythrocytes were readily apparent. Although their reduction of fluorescence intensity at early stages was similar to that found in parasitized AA erythrocytes, their reductions of fluorescence intensity by the mid and late stages were 25% and 60%, respectively (Fig. 3B, Table 1).

View this table:
Table 1.

Differences in band 3 fluorescence intensity from AA and CC erythrocytes at different parasite developmental stages

### Band 3 cluster sizes can be estimated mathematically

To evaluate the degree of band 3 clustering, we plotted the intensity profile around the periphery of a single image at the center of the deconvolved z-stack (Fig. 4A). The intensity profile (Fig. 4B) shows the presence of both high and low intensity fluctuation frequencies that are inversely related to cluster sizes. We extracted these data values and displayed them as a power spectrum plot (Fig. 4C). The presence of high power values at low frequencies implies the presence of high concentrations of larger clusters. To compare the effects of hemoglobin type and parasite stage on individual power spectra, we normalized power spectrum values against the highest power value present in the data set and re-plotted the resulting data on a double logarithmic scale (Fig. 4D). Linear regression analyses of the data are estimates of intra-cluster size distributions; steeper slopes imply that a higher proportion of large clusters exists in the population. To evaluate the effects of hemoglobin type and parasite stage on the degree of band 3 clustering, we compared the slopes of power spectrum regression lines. Both AA and CC erythrocytes showed a tendency to form larger clusters (steeper slopes) as parasites matured (Fig. 4E). The slope of late-stage-parasitized AA erythrocytes (Fig. 4E, single arrow) is similar to the slope of uninfected CC erythrocytes (Fig. 4E, double arrow), consistent with our immunoblot results (Fig. 1A). Moreover, the overall slope of CC erythrocytes (–0.072) is nearly twofold higher than AA erythrocytes (–0.047), indicating that the degree of band 3 clustering in CC erythrocytes is significantly higher than in AA erythrocytes.

Fig. 4.

Analysis of band 3 clustering. (A) Example of a line profile of fluorescence intensity drawn at the periphery of the center section of a deconvolved erythrocyte (arrow). Bar, 10 μm. (B) A graphical representation of the profile shown in A. 1 pixel=45 nm. (C) Power spectrum of the frequency profile of measured fluorescence fluctuations shown in B plotted as the inverse of pixel distance (f). (D) Double logarithmic plots of power spectra show linear relationship between frequency and normalized power. Multiple data sets for a developmental stage in AA erythrocytes and their linear regression lines are shown. (E) A comparison of regression slope values between AA and CC erythrocytes as a function of parasite developmental stage. These slopes demonstrate that CC erythrocytes have higher degrees and rates of clustering than AA erythrocytes throughout the intracellular parasite cycle. (F) Autocorrelation analyses of measured fluorescence intensity profiles of representative uninfected and parasitized AA and CC erythrocytes. Only the first 1 μm of the intensity profile is shown for clarity. Autocorrelation values were normalized against the value at C0. The first decay function of the data was fitted (solid lines) using a first-order exponential function to estimate characteristic decay lengths. The slower decay values shown by both uninfected and parasitized CC cells and parasitized AA erythrocytes represent higher degrees of band 3 clustering.

Although power spectrum analysis can be used to elucidate a frequency distribution, the actual clustering sizes within the distributions cannot be accurately estimated. Therefore, we performed 1D autocorrelation analyses of intensity profiles to estimate the actual cluster sizes. The autocorrelation function (ACF) has been used effectively to characterize the spatial domain size of lipid molecules in both uninfected and artificial lipid bilayers (Hwang et al., 1998; Tokumasu et al., 2004). A one-dimensional ACF, C(ξ), can be expressed as: $Math$ where I(x) is pixel intensity at pixel position x, N is the total number of pixels measured, and ξ is the spatial lag length. In our case, spatial lag increment was set to 1 pixel because this was the signal sampling interval. To ensure spatial consistency between erythrocytes having different periphery lengths, we set the total number of peripheral pixels to 230, which corresponds to 10.3 μm. The data of the primary autocorrelation decay calculation can be fitted with a first-order exponential equation. The autocorrelation value at 1/e is a half-width characteristic decay length (ξ′) that estimates the average half domain (cluster) size at the erythrocyte periphery. The analytical results for ξ′ are shown in Table 2. The autocorrelation plot contains non-periodic fluctuations, precluding the possibility of fitting with a simple decay function. Therefore, we considered only the first decay values for cluster size analysis. In uninfected AA and CC erythrocytes, the estimated ξ′ was 159 nm and 243 nm, respectively, indicating a relatively higher degree of band 3 clustering in CC erythrocytes. In parasitized (late stage) AA and CC erythrocytes, these values increased to 257 nm and 570 nm, indicating that parasite development induces band 3 clustering to a greater degree in CC compared with AA erythrocytes. Long-range spatial correlations, both autocorrelation values and decay patterns for parasitized AA and uninfected CC, are similar up to 5 μm (data not shown). These results are consistent with our power spectrum data and confirm that the degree of band 3 clustering in parasitized AA erythrocytes is similar to that already present in CC erythrocytes in vivo.

View this table:
Table 2.

Comparison of band 3 cluster sizes in AA and CC erythrocytes

## Discussion

Band 3 proteins exist predominantly as dimers in the erythrocyte membrane. The oligomerization (clustering) of band 3 to tetramers or higher molecular weight forms is promoted by a variety of processes, including erythrocyte senescence and P. falciparum infection. Compared with uninfected AA erythrocytes, we found that uninfected CC erythrocytes exhibit higher band 3 oligomer to dimer ratios as well as larger amounts of higher MW band 3 species, suggesting that this hemoglobin type influences the clustering of band 3. Quantitative quantum-dot-based fluorescence, power spectra and autocorrelation analyses independently confirmed that CC erythrocytes have increased band 3 clustering. In addition, the size of band 3 clusters was mathematically estimated to be significantly higher in CC erythrocytes. High autocorrelation values occurred in CC erythrocytes throughout the cell periphery and the rapid decrease in the autocorrelation function in uninfected (native) AA erythrocytes implied that the total length of the cell periphery (or the domain density) contributes partially to the autocorrelation results.

ELISA experiments using SE-HPLC fractions and anti-band 3 mAb resulted in unexpectedly lower band 3 signals in CC compared to AA samples. One possible explanation for this finding is that band 3 proteolysis is enhanced in CC erythrocytes, destroying the epitope of the mAb. During normal erythrocyte senescence, band 3 proteolysis is known to generate a faint 60 kDa band (Kay, 1981), which we detected in both band 3 immunoblots and silver stained gels of CC erythrocytes. This faint band, which may be related to senescent antigen' (Kay, 1984), was not detected in AA erythrocytes. These data show that band 3 molecules undergo enhanced degradation in CC erythrocytes and suggest that these cells undergo accelerated senescence in vivo. A second possibility is that band 3 clustering in CC erythrocytes sterically hinders the ability of the mAb to react with its epitope on cdb3. Although we have no direct evidence regarding this possibility, it has been shown that clustered band 3 molecules themselves sterically restrict the access of another mAb to its cdb3 epitope (Rouger and Anstee, 1988). A third possibility is that hemichromes attached to cdb3 compete directly with the mAb for shared or partially overlapping cdb3 epitopes. Our identification of reduced ELISA signals (Fig. 1B) and increased hemichrome levels in CC erythrocytes (R.M.F. et al., unpublished) support this possibility.

P. falciparum induces band 3 oligomerization in both AA and CC erythrocytes. However, this effect is more pronounced in parasitized CC erythrocytes as these show increased levels of higher MW species relative to parasitized AA erythrocytes. We also noticed that a larger number of parasitized CC erythrocytes were required on immunoblot analyses to produce a signal level equivalent to that of uninfected CC or AA erythrocytes (parasitized or not). The marked differences in immunofluorescence intensity between uninfected and parasitized erythrocytes raised the possibility that parasite-encoded proteases are involved in the degradation of band 3 and the consequent loss of anti-band 3 mAb signal. There is precedence for this observation in that band 3 cleavage during merozoite invasion (Roggwiller et al., 1996) and digestion of cytoskeleton-associated proteins by parasite-derived proteases during schizont rupture (Dua et al., 2001; Hanspal et al., 2002) have been described. However, we could not detect marked differences in density for the 60 kDa band (related to senescent antigen') between uninfected and parasitized CC erythrocytes, perhaps due to an enhanced protection of digestion sites by hemichromes. Therefore, the reduced accessibility of the anti-band 3 mAb to cdb3 provides evidence that P. falciparum induces a greater degree of band 3 clustering in CC than AA erythrocytes. These data support the power spectra and autocorrelation results showing the rate of band 3 clustering relative to parasite developmental stage is even higher in CC erythrocytes.

Normal erythrocyte senescence is associated with the oxidation of hemoglobin molecules to hemichromes, which cluster band 3 and generate senescent antigen (Low et al., 1985; Waugh and Low, 1985). The clustering of band 3 may reveal cryptic epitopes for autologous IgG or result in the de novo formation of conformational epitopes (Guthrie et al., 1995; Lutz et al., 1984). Although the precise molecular structure of senescent antigen is unknown, its formation on the erythrocyte surface results in autologous IgG binding and complement fixation. Senescent erythrocytes can be opsonized for efficient removal from the circulation by splenic macrophages (Giger et al., 1995; Lutz et al., 1988).

P. falciparum also has been shown to induce hemichrome deposition, band 3 clustering, autologous IgG binding, complement fixation, and phagocytosis by macrophages (Fig. 5) (Giribaldi et al., 2001). Although the mechanisms involved in band 3 clustering might vary between senescent and parasitized erythrocytes, the downstream events in the opsonization pathway would likely be the same.

Fig. 5.

Proposed model for the mechanism of fluorescence reduction due to band 3 aggregation in AA and CC erythrocytes. Anti-band 3 mAb against cdb3 has easy access to band 3 epitopes in uninfected AA erythrocytes. During the course of P. falciparum infection, however, hemichromes form and bind with high affinity to the cdb3, hindering the anti-band 3 mAb binding. Band 3 also clusters to form oligomeric forms which may provide better recognition site for autologous antibody. Uninfected CC erythrocytes have increased band 3 oligomerization compared to uninfected AA erythrocytes, and a considerable amount of hemichromes already bound to the cdb3 that make those epitopes less accessible to anti-band 3 mAb, as shown in Fig. 1B. When these CC erythrocytes become infected, hemichromes form even larger deposits that further reduce antibody accessibility to band 3 epitopes, and band 3 forms higher levels of clustering, as shown Fig. 1A, Fig. 3B, Fig. 4E and 4F.

P. falciparum-induced changes in band 3 clustering have been proposed to enhance the splenic clearance of parasites in vivo during the course of a clinical malaria episode (Hogh et al., 1994; Kennedy, 2001). The potential importance of this mechanism in malaria protection is highlighted by several reports showing that this opsonization scheme operates with higher efficiency in erythrocytes carrying malaria protective mutations. For example, enhanced phagocytosis of parasitized erythrocytes has been proposed as a mechanism for the malaria protective effects of glucose-6-phosphate dehydrogenase deficiency, sickle-trait, and β-thalassemia trait (Ayi et al., 2004; Cappadoro et al., 1998).

Whether enhanced band 3 clustering of parasitized erythrocytes contributes to malaria protection by HbC remains to be determined. By undergoing accelerated oxidation to hemichromes and binding cdb3 with higher affinity, HbC erythrocytes would have a greater propensity than HbA erythrocytes to cluster band 3. Our recent identification of elevated hemichromes and autologous IgG in freshly drawn HbC erythrocytes (R.M.F. et al., unpublished) is consistent with this hypothesis and raises the possibility that band 3-directed opsonization could enhances the clearance of parasitized HbC erythrocytes and contribute to malaria protection in vivo.

## Acknowledgements

We thank Hongjian Liu, Quantum Dot Corporation, for helpful suggestions in the use of quantum dots, Jonathan Girroir and Steve Carano, AutoQuant Imaging, for assistance in using AutoDeblur, Takayuki Arie, LMVR, NIAID, NIH, for helpful discussions on mathematical analyses, and John Andersen, LMVR, NIAID, NIH, for technical assistance with SE-HPLC. Although certain commercial equipment, instruments and materials are identified in this paper, such identification does not imply recommendations or endorsements by NIST, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

• Accepted November 24, 2004.

View Abstract

## Summary

The molecular stability of hemoglobin is critical for normal erythrocyte functions, including oxygen transport. Hemoglobin C (HbC) is a mutant hemoglobin that has increased oxidative susceptibility due to an amino acid substitution (β6: Glu to Lys). The growth of Plasmodium falciparum is abnormal in homozygous CC erythrocytes in vitro, and CC individuals show innate protection against severe P. falciparum malaria. We investigated one possible mechanism of innate protection using a quantum dot technique to compare the distribution of host membrane band 3 molecules in genotypically normal (AA) to CC erythrocytes. The high photostability of quantum dots facilitated the construction of 3D cell images and the quantification of fluorescent signal intensity. Power spectra and 1D autocorrelation analyses showed band 3 clusters on the surface of infected AA and CC erythrocytes. These clusters became larger as the parasites matured and were more abundant in CC erythrocytes. Further, average cluster size (500 nm) in uninfected (native) CC erythrocytes was comparable with that of parasitized AA erythrocytes but was significantly larger (1 μm) in parasitized CC erythrocytes. Increased band 3 clustering may enhance recognition sites for autoantibodies, which could contribute to the protective effect of hemoglobin C against malaria.

Key words
• Hemoglobin C
• Malaria
• Band 3 clustering
• Quantum dots
• Power spectrum
• Autocorrelation
• Membrane modification

## Introduction

Individuals with mutant hemoglobins (Hb) are protected against life-threatening manifestations of Plasmodium falciparum malaria. For example, hemoglobin C (β6: Glu to Lys) markedly reduces the risk of severe malaria in West Africa (up to 90% reduction with homozygous HbCC) (Agarwal et al., 2000; Modiano et al., 2001). It has been shown that frequent intra-erythrocytic parasite death may underlie the lower multiplication rates of P. falciparum in CC erythrocytes in vitro (Fairhurst et al., 2003). In addition, the formation of aberrant knobs on the surface of parasitized CC erythrocytes suggests an abnormal membrane remodeling process. It is possible that alterations in the quaternary structure of HbC may expose the heme moiety to oxidative damage, resulting in marked HbC denaturation, decreased solubility compared to HbA and a propensity to form HbC crystals in vivo (Fabry et al., 1981; Hirsch et al., 1985). The structural abnormalities of CC erythrocytes may make them more susceptible to host immune responses such as occur in hemoglobinopathic and normally aged erythrocytes (Ando et al., 1997; Schluter and Drenckhahn, 1986; Turrini et al., 1991). However, details of these membrane modifications in CC cells due to abnormal Hb and parasite infection have only recently been quantified (Arie et al., 2005).

Intracellular denaturation of Hb can cause significant changes in the erythrocyte membrane. For example, in Heinz body hemolytic anemia denatured hemoglobin precipitates to form insoluble hemichromes that bind to the inner surface of erythrocytes (Jacob and Winterhalter, 1970; Rachmilewitz, 1969; Rachmilewitz, 1974; Rifkind et al., 1994; Schneider et al., 1972; Winterbourn and Carrell, 1974). The consequent release of reactive oxygen intermediates increases membrane permeability and induces premature hemolysis (Hebbel, 1990). Hemichrome formation is also associated with the aggregation of membrane proteins during normal erythrocyte senescence (Low et al., 1985; Waugh and Low, 1985; Waugh et al., 1986). In sickle cell disease, the instability of HbS results in premature senescence of erythrocytes, as suggested by increased amounts of membrane-bound autologous IgG (Bosman, 2004). Aggregation of band 3 is thought to form a neoantigen that binds autologous immunoglobulin (IgG) and complement, thereby targeting aged erythrocytes for destruction (Ando et al., 1997; Liu et al., 1991; Lutz et al., 1984; Schluter and Drenckhahn, 1986; Turrini et al., 1991). Whether band 3 aggregates contribute to the malaria-protective effect of CC erythrocytes has not been determined.

To investigate this possibility, we monitored band 3 clustering in P. falciparum-infected AA and CC erythrocytes using a newly developed quantum-dot-based immunochemical technique (Tokumasu and Dvorak, 2003). The technique permits 3D reconstruction and quantification from a z-stack series of undistorted erythrocytes without fluorescence intensity loss caused by photobleaching. Here, we present the results of single-cell studies of total fluorescence intensity and distribution of band 3 in uninfected and infected erythrocytes using power spectra and 1D autocorrelation analyses. We found that the accessibility of an intracellular band 3 epitope to antibody was markedly reduced in CC compared to AA erythrocytes, indicating a high degree of band 3 clustering in CC erythrocytes in vivo and during intracellular parasite development in vitro.

## Materials and Methods

### Erythrocyte preparation and parasite culture

The preparation, hemoglobin phenotype determination, and P. falciparum (7G8 line) infection of AA and CC erythrocytes have been described previously (Fairhurst et al., 2003). Parasite cultures at 5% parasitemia were used.

### Membrane extracts and immunoblotting

Uninfected (native) and parasitized erythrocytes were washed in HEPES buffer, pH 7.4 (10 mM HEPES, 140 mM NaCl, 10 mM glucose (the conventional notation, rather than the recommendation in NIST-SP811, is employed for this publication with M representing mol/l). Erythrocytes were lysed and the membrane pellets were washed extensively with ice-cold Tris-HCl/EDTA pH 8.0, in the presence of protease inhibitors (Complete™, Roche Diagnostics, Germany). Membrane proteins were extracted from the pellets using mild detergent conditions as described (Giribaldi et al., 2001). Band 3 protein in the membrane extracts was analyzed using a 4-12% Bis-Tris acrylamide gel under non-reducing conditions (Invitrogen Corporation, Carlsbad, CA). Each well was loaded with 0.1 μg of total protein, except for CC erythrocytes where 0.3 μg of protein was used. Following electrophoresis, proteins were transferred to a PVDF membrane and probed with a monoclonal antibody (mAb) against the cytoplasmic domain of band 3 (cdb3) (Sigma-Aldrich, St Louis, MO) diluted to 1:5000 in casein-based blocking buffer (Sigma-Aldrich). Following incubation for 1 hour at room temperature, the membrane was washed in PBS-0.2% Tween-20, incubated with 1:50,000 HRP-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) and developed with Super Signal® West Pico chemical luminescence solutions (Pierce, Rockford, IL).

### Size exclusion HPLC and ELISA of uninfeced AA and CC membrane extracts

The size exclusion HPLC (SE-HPLC) of solubilized AA and CC membrane extracts was performed using a 7.5×300 mm TSK-gel G4000SW column (Toso-Haas, Montgomeryville, PA) fitted to an AKTA purifier-10 system (Amersham-Pharmacia, Piscataway, NJ). The running buffer was 5 mM sodium phosphate, 100 mM NaCl, and 0.1% C12E8 (octaethyleneglycol mono-n-dodecyl ether) detergent (Sigma), pH 7.0 as described (Rettig et al., 2001). The gel filtration standard used contained a mixture of molecular weight markers ranging from 670 to 1.35 kDa (Bio-Rad Laboratories, Hercules, CA). Protein elution was recorded at 220 and 280 nm. ELISA assays were performed on Reacti-Bind EIA microplates (Pierce) using anti-band 3 mAb at 1/5000 or 1/2000 dilutions (Sigma-Aldrich) followed by HRP-conjugated goat anti-mouse IgG at 1/20,000 dilution (Jackson ImmunoResearch Laboratories, Inc.). HRP substrate was incubated for 25 minutes before stopping the reaction with 1 N H2SO4. Reaction absorbance of the plate was read at 450 nm.

### Immunolabelling band 3 protein with quantum dots

Coverglass treatment, erythrocyte fixation and quantum dot immunolabelling were performed as described (Tokumasu and Dvorak, 2003), except that QD-585 quantum dots (Quantum Dot Corporation, Hayward, CA) were used instead of QD-605 quantum dots and 1% HEPES was added to the RPMI-1640 washing medium. In brief, erythrocytes immobilized on cover glasses using Alcian Blue (Sigma-Aldrich) were crosslinked with 50 mM dimethyl suberimidate (DMS) (Sigma-Aldrich) in 100 mM sodium borate buffer, pH 9.5, containing 1 mM MgCl2 for 1 hour, followed by 2% paraformaldehyde (Electron Microscopy Science, Fort Washington, PA). The crosslinking reaction was quenched using 0.1 M glycine in PBS, pH 7.4 for 1 hour. Erythrocytes were blocked for 1 hour in PBS-0.2% Tween 20, 3% BSA and incubated with anti-band 3 mAb (1/3000) in the same blocking solution. Erythrocytes were washed with PBS-0.2% Tween 20 three times and then incubated with biotin-conjugated goat anti-mouse IgG (Sigma-Aldrich) and further reacted with 10-15 nM of streptavidin-conjugated quantum dots (QD585) for 30 minutes. Hoechst 33258 (Molecular Probes, Eugene, OR) was added to stain parasite nuclei during the final washing step.

### Single cell fluorescence analysis Fluorescence microscopy

Wide-field fluorescence images were obtained with a Leica DM IRE2 inverted microscope (Leica Microsystems, Bannockburn, IL) using a 100× NA 1.4 PlanApochromat objective, 1.5× intermediate magnification, a custom-made QD-585 fluorescence filter set (ex. 430DF60/20, em. 580DF20, Omega Optical, Brattleboro, VT), and an XF136-2 filter module (Omega) for Hoechst 33258. Z-stack images (70 frames) were acquired with a Hamamatsu Orca digital camera (Hamamatsu Photonics System, Bridgewater, NJ) using Image Pro and Scope Pro ver. 4.5 (Media Cybernetics, Silver Spring, MD).

### Image deconvolution, fluorescence measurements and cluster analyses

Three-dimensional image reconstruction from 70-plane image stacks, total fluorescence measurements and multi-channel, blind deconvolution were performed using Auto Deblur ver. 9.3 (AutoQuant Imaging, Watervliet, NY). The deconvolved images were stored in an uncompressed AVI format. Total erythrocyte fluorescence was obtained as a sum projection of the image stack. The fluorescence intensity from each parasitized erythrocyte was normalized against the mean fluorescence intensity of multiple uninfected erythrocytes present in the same image field. Power spectra, custom-built 1D autocorrelation analyses, and statistical analyses including the fitting of autocorrelation data were performed in Origin ver. 7.5 (OriginLab, Northampton, MA).

## Results

### Immunoblotting and size-exclusion HPLC elucidates marked differences in band 3 oligomerization between AA and CC erythrocytes

In uninfected (native) erythrocyte membranes, band 3 exists predominantly as dimers; alterations of these higher order structures due to the presence of mutant hemoglobins or P. falciparum infection indicate changes in membrane structure. To determine whether hemoglobin C influences the membrane structure of both uninfected and parasitized erythrocytes, we used an anti-band 3 monoclonal antibody (mAb) to elucidate the degree of band 3 oligomerization under non-reducing conditions (Fig. 1). Immunoblot analyses of membrane protein extracts showed that the majority of band 3 in uninfected AA erythrocytes is in the dimeric form with a molecular weight (MW) of 190 kDa; small amounts of higher MW species were also present but the bands were not distinct. The oligomer to dimer ratio increased twofold in parasitized AA erythrocytes (Fig. 1A, arrow). Moreover, higher levels of the monomeric form (90 kDa) were also found suggesting that breakdown of oligomers occurs during parasite infection. Parasitized AA erythrocytes presented a slightly lower MW for both the monomer and dimer, possibly due to differential glycosylation of band 3.

Fig. 1.

Analysis of band 3 clustering in uninfected and parasitized erythrocytes. (A) Immunoblot of membrane extracts from both uninfected and parasitized erythrocytes using a mAb against the band 3 cytoplasmic domain (cdb3). All lanes, except for parasitized CC erythrocytes, were loaded with 0.1 μg of total protein. 0.3 μg of total protein from parasitized CC erythrocytes was used to obtain approximately the same amount of signal from band 3 protein. The arrows point out the signal from band 3 oligomers found in parasitized AA and CC erythrocytes. Much greater oligomerization is apparent in the lane of parasitized CC erythrocytes. (B) Quantification of band 3 in SE-HPLC fractions of AA and CC erythrocyte membrane extracts. The amount of protein used in CC extracts (0.63 μg/μl) was approximately double that in AA extracts (0.33 μg/μl). Band 3 in the SE-HPLC fractions was semi-quantified by ELISA using mAb against cdb3 at 1/5000 and 1/2000 dilutions. The anti-band 3 signal obtained for AA extracts is markedly higher than for CC extracts. In a competition ELISA, increasing the amount of mAb caused only a minor increase in anti-band 3 signals.

A higher MW band (300 kDa) that appeared in parasitized AA erythrocytes was also observed in uninfected CC erythrocytes (Fig. 1A), indicating that CC erythrocytes have altered band 3 intermolecular interactions in vivo. The oligomer to dimer ratio increased 2-fold in parasitized CC erythrocytes (Fig. 1A). In addition, a continuous distribution of higher MW band 3 species was found in parasitized CC erythrocytes, implying that either a variety of oligomers were formed or the proteins were markedly aggregated. However, the monomer to dimer ratio remained unchanged (data not shown). We also determined that a higher amount of total CC erythrocyte protein compared to AA erythrocyte protein was required to obtain an equivalent amount of total band 3 signal, implying that mAb accessibility and/or the amount of available epitope is lower in CC compared to AA erythrocytes.

To explore this possibility, we fractionated uninfected AA and CC erythrocyte membrane extracts by SE-HPLC and used an ELISA competition assay to measure their reactivity with anti-band 3 mAb (Fig. 1B). We observed reduced mAb reactivity in CC compared with AA membrane extracts even though the amount of total protein in the CC fractions (0.63 μg/μl) was twice that in the AA fractions (0.33 μg/μl). Increasing the amount of anti-band 3 mAb for ELISA assays of CC fractions resulted in a minor increase in signal; however, the signal in the CC fractions remained markedly lower compared with AA fractions. This finding may be explained by relatively high levels of hemichromes deposited on the cytoplasmic aspect of CC erythrocyte membranes (R.M.F. et al., unpublished data), where they could interfere with the ability of our anti-band 3 mAb to bind cdb3. In addition, the amount of higher MW oligomeric forms in CC fractions was slightly increased, confirming our immunoblot data. Dimers and higher MW species of band 3 proteins were not completely dissociated under reducing conditions (data not shown), as reported previously (Ando et al., 1997).

### AA and CC erythrocytes display marked morphological differences

As immunoblot data showed marked alterations in band 3 oligomerization between AA and CC erythrocytes, we compared their band 3 distribution patterns using a quantum dot-based immunofluorescence assay. Reconstructed 3D images of uninfected AA erythrocytes revealed smooth, biconcave shapes (Fig. 2A). By contrast, uninfected CC erythrocytes appeared wrinkled and withered when prepared under identical conditions (Fig. 2B). As shown previously (Smith and Krevans, 1959), some CC erythrocytes were microspherocytic. At the early (ring) stage of parasite development, neither AA nor CC erythrocytes showed significant morphological changes (Fig. 2C,F). At the trophozoite stage, however, both AA and CC erythrocytes began rounding up and losing their biconcave shape (Fig. 2D,G). This change was more pronounced in CC erythrocytes. At the schizont stage, both AA and CC erythrocytes were completely rounded up and their projected areas decreased (Fig. 2E,H).

Fig. 2.

Immunocytochemistry of infected erythrocytes. Membranes were immunolabeled with quantum dots for anti-band 3 mAb (red) and parasite nuclei were labeled with Hoechst 33258 (blue). These images are the result of 3D reconstruction and deconvolution. (A) Parasitized AA erythrocytes surrounded by uninfected erythrocytes. (B) Parasitized CC erythrocytes surrounded by uninfected erythrocytes. Note the abnormal membrane appearance of uninfected CC erythrocytes. Comparisons of erythrocyte shapes at early, mid, and late stages of parasitized AA (C-E) and CC erythrocytes (F-H). Bars, 5 μm (A,B); bars, 1 μm (C-H).

### Quantum dot-based fluorescence analyses show reduced band 3 accessibility in parasitized and hemoglobin C-containing erythrocytes

To quantify the effects of parasitization on erythrocyte membranes, we measured the total fluorescence intensity of individual erythrocytes by a quantum dot-based immunofluorescence assay using anti-band 3 mAb. In contrast to routine immunochemical methods, our preparative procedure does not lyse erythrocytes and thus allows us to measure the total fluorescence intensity of individual erythrocytes. The high photostability of quantum dots (Tokumasu and Dvorak, 2003; Wu et al., 2003) eliminated the photobleaching problem normally encountered in this type of analysis. For each erythrocyte, we collected a total of 70 z-section serial images and combined them into a single 3D image. To confirm that hemoglobin absorption did not affect the fluorescence intensity of our mAb signal, we first quantified the fluorescence intensity of each z-stack (Fig. 3A). We found that, in uninfected erythrocytes oriented parallel to the plane of the glass substrate, the fluorescence intensities of the top and bottom frames were statistically indistinguishable resulting in an overall symmetric intensity profile (skewness=0.36<0.53 at the 95% confidence level, n=51 frames). These results demonstrate that the quenching of band 3 fluorescence intensity due to hemoglobin absorption is negligible.

Fig. 3.

Fluorescence intensity measurements from uninfected and parasitized erythrocytes. (A) Representation example of fluorescence symmetry through a z-stack series of a horizontally oriented, healthy AA erythrocyte. Although the z-stack consisted of 70 frames, only those frames containing the erythrocyte are shown. Each fluorescence intensity profile was centered along the vertical axis for representational purposes. Owing to the high photostability of quantum dots, no measurable reduction in fluorescence intensity occurred throughout the z-stack sequence. (B) Cell-by-cell comparison of mean total fluorescence intensity occurring during the development of malaria parasites in AA and CC erythrocytes. The total fluorescence was measured using 3D-reconstructed, sum projection images. Data were normalized against uninfected erythrocytes (U) for each genotype. Error bars represent the standard error of the mean (s.e.m.). n=3-11 (see Table 1).

We then compared the total band 3 fluorescence intensities of uninfected and parasitized AA and CC erythrocytes (Fig. 3B). To correct for any possible irregularities in field illumination, the fluorescence intensity of each parasitized erythrocyte was normalized against the mean fluorescence intensity of multiple uninfected erythrocytes present in the same microscopic field. During the early and mid stages of infection, the fluorescence intensity of AA erythrocytes decreased slightly (10%). There was no significant difference between early (ring) and mid (early trophozoite) stages. However, a 40% decrease in fluorescence intensity occurred in late (late trophozoite and schizont) stages (Fig. 3B, Table 1). The reduction in fluorescence intensity was not apparent qualitatively. In fact, the reduction in projected area of late stage parasitized erythrocytes resulted in quantum dots being confined to a smaller area, giving the false impression that fluorescence intensity increased. By contrast, reductions in the fluorescence intensity of parasitized CC erythrocytes were readily apparent. Although their reduction of fluorescence intensity at early stages was similar to that found in parasitized AA erythrocytes, their reductions of fluorescence intensity by the mid and late stages were 25% and 60%, respectively (Fig. 3B, Table 1).

View this table:
Table 1.

Differences in band 3 fluorescence intensity from AA and CC erythrocytes at different parasite developmental stages

### Band 3 cluster sizes can be estimated mathematically

To evaluate the degree of band 3 clustering, we plotted the intensity profile around the periphery of a single image at the center of the deconvolved z-stack (Fig. 4A). The intensity profile (Fig. 4B) shows the presence of both high and low intensity fluctuation frequencies that are inversely related to cluster sizes. We extracted these data values and displayed them as a power spectrum plot (Fig. 4C). The presence of high power values at low frequencies implies the presence of high concentrations of larger clusters. To compare the effects of hemoglobin type and parasite stage on individual power spectra, we normalized power spectrum values against the highest power value present in the data set and re-plotted the resulting data on a double logarithmic scale (Fig. 4D). Linear regression analyses of the data are estimates of intra-cluster size distributions; steeper slopes imply that a higher proportion of large clusters exists in the population. To evaluate the effects of hemoglobin type and parasite stage on the degree of band 3 clustering, we compared the slopes of power spectrum regression lines. Both AA and CC erythrocytes showed a tendency to form larger clusters (steeper slopes) as parasites matured (Fig. 4E). The slope of late-stage-parasitized AA erythrocytes (Fig. 4E, single arrow) is similar to the slope of uninfected CC erythrocytes (Fig. 4E, double arrow), consistent with our immunoblot results (Fig. 1A). Moreover, the overall slope of CC erythrocytes (–0.072) is nearly twofold higher than AA erythrocytes (–0.047), indicating that the degree of band 3 clustering in CC erythrocytes is significantly higher than in AA erythrocytes.

Fig. 4.

Analysis of band 3 clustering. (A) Example of a line profile of fluorescence intensity drawn at the periphery of the center section of a deconvolved erythrocyte (arrow). Bar, 10 μm. (B) A graphical representation of the profile shown in A. 1 pixel=45 nm. (C) Power spectrum of the frequency profile of measured fluorescence fluctuations shown in B plotted as the inverse of pixel distance (f). (D) Double logarithmic plots of power spectra show linear relationship between frequency and normalized power. Multiple data sets for a developmental stage in AA erythrocytes and their linear regression lines are shown. (E) A comparison of regression slope values between AA and CC erythrocytes as a function of parasite developmental stage. These slopes demonstrate that CC erythrocytes have higher degrees and rates of clustering than AA erythrocytes throughout the intracellular parasite cycle. (F) Autocorrelation analyses of measured fluorescence intensity profiles of representative uninfected and parasitized AA and CC erythrocytes. Only the first 1 μm of the intensity profile is shown for clarity. Autocorrelation values were normalized against the value at C0. The first decay function of the data was fitted (solid lines) using a first-order exponential function to estimate characteristic decay lengths. The slower decay values shown by both uninfected and parasitized CC cells and parasitized AA erythrocytes represent higher degrees of band 3 clustering.

Although power spectrum analysis can be used to elucidate a frequency distribution, the actual clustering sizes within the distributions cannot be accurately estimated. Therefore, we performed 1D autocorrelation analyses of intensity profiles to estimate the actual cluster sizes. The autocorrelation function (ACF) has been used effectively to characterize the spatial domain size of lipid molecules in both uninfected and artificial lipid bilayers (Hwang et al., 1998; Tokumasu et al., 2004). A one-dimensional ACF, C(ξ), can be expressed as: \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} $\ C({\xi})=\frac{{{\sum}_{n=-N{/}2}^{N{/}2}}I(x_{n}){\cdot}I(x_{n}+{\xi})}{{{\sum}_{n=-N{/}2}^{N{/}2}}I^{2}(x_{n})},\$ \end{document} where I(x) is pixel intensity at pixel position x, N is the total number of pixels measured, and ξ is the spatial lag length. In our case, spatial lag increment was set to 1 pixel because this was the signal sampling interval. To ensure spatial consistency between erythrocytes having different periphery lengths, we set the total number of peripheral pixels to 230, which corresponds to 10.3 μm. The data of the primary autocorrelation decay calculation can be fitted with a first-order exponential equation. The autocorrelation value at 1/e is a half-width characteristic decay length (ξ′) that estimates the average half domain (cluster) size at the erythrocyte periphery. The analytical results for ξ′ are shown in Table 2. The autocorrelation plot contains non-periodic fluctuations, precluding the possibility of fitting with a simple decay function. Therefore, we considered only the first decay values for cluster size analysis. In uninfected AA and CC erythrocytes, the estimated ξ′ was 159 nm and 243 nm, respectively, indicating a relatively higher degree of band 3 clustering in CC erythrocytes. In parasitized (late stage) AA and CC erythrocytes, these values increased to 257 nm and 570 nm, indicating that parasite development induces band 3 clustering to a greater degree in CC compared with AA erythrocytes. Long-range spatial correlations, both autocorrelation values and decay patterns for parasitized AA and uninfected CC, are similar up to 5 μm (data not shown). These results are consistent with our power spectrum data and confirm that the degree of band 3 clustering in parasitized AA erythrocytes is similar to that already present in CC erythrocytes in vivo.

View this table:
Table 2.

Comparison of band 3 cluster sizes in AA and CC erythrocytes

## Discussion

Band 3 proteins exist predominantly as dimers in the erythrocyte membrane. The oligomerization (clustering) of band 3 to tetramers or higher molecular weight forms is promoted by a variety of processes, including erythrocyte senescence and P. falciparum infection. Compared with uninfected AA erythrocytes, we found that uninfected CC erythrocytes exhibit higher band 3 oligomer to dimer ratios as well as larger amounts of higher MW band 3 species, suggesting that this hemoglobin type influences the clustering of band 3. Quantitative quantum-dot-based fluorescence, power spectra and autocorrelation analyses independently confirmed that CC erythrocytes have increased band 3 clustering. In addition, the size of band 3 clusters was mathematically estimated to be significantly higher in CC erythrocytes. High autocorrelation values occurred in CC erythrocytes throughout the cell periphery and the rapid decrease in the autocorrelation function in uninfected (native) AA erythrocytes implied that the total length of the cell periphery (or the domain density) contributes partially to the autocorrelation results.

ELISA experiments using SE-HPLC fractions and anti-band 3 mAb resulted in unexpectedly lower band 3 signals in CC compared to AA samples. One possible explanation for this finding is that band 3 proteolysis is enhanced in CC erythrocytes, destroying the epitope of the mAb. During normal erythrocyte senescence, band 3 proteolysis is known to generate a faint 60 kDa band (Kay, 1981), which we detected in both band 3 immunoblots and silver stained gels of CC erythrocytes. This faint band, which may be related to senescent antigen' (Kay, 1984), was not detected in AA erythrocytes. These data show that band 3 molecules undergo enhanced degradation in CC erythrocytes and suggest that these cells undergo accelerated senescence in vivo. A second possibility is that band 3 clustering in CC erythrocytes sterically hinders the ability of the mAb to react with its epitope on cdb3. Although we have no direct evidence regarding this possibility, it has been shown that clustered band 3 molecules themselves sterically restrict the access of another mAb to its cdb3 epitope (Rouger and Anstee, 1988). A third possibility is that hemichromes attached to cdb3 compete directly with the mAb for shared or partially overlapping cdb3 epitopes. Our identification of reduced ELISA signals (Fig. 1B) and increased hemichrome levels in CC erythrocytes (R.M.F. et al., unpublished) support this possibility.

P. falciparum induces band 3 oligomerization in both AA and CC erythrocytes. However, this effect is more pronounced in parasitized CC erythrocytes as these show increased levels of higher MW species relative to parasitized AA erythrocytes. We also noticed that a larger number of parasitized CC erythrocytes were required on immunoblot analyses to produce a signal level equivalent to that of uninfected CC or AA erythrocytes (parasitized or not). The marked differences in immunofluorescence intensity between uninfected and parasitized erythrocytes raised the possibility that parasite-encoded proteases are involved in the degradation of band 3 and the consequent loss of anti-band 3 mAb signal. There is precedence for this observation in that band 3 cleavage during merozoite invasion (Roggwiller et al., 1996) and digestion of cytoskeleton-associated proteins by parasite-derived proteases during schizont rupture (Dua et al., 2001; Hanspal et al., 2002) have been described. However, we could not detect marked differences in density for the 60 kDa band (related to senescent antigen') between uninfected and parasitized CC erythrocytes, perhaps due to an enhanced protection of digestion sites by hemichromes. Therefore, the reduced accessibility of the anti-band 3 mAb to cdb3 provides evidence that P. falciparum induces a greater degree of band 3 clustering in CC than AA erythrocytes. These data support the power spectra and autocorrelation results showing the rate of band 3 clustering relative to parasite developmental stage is even higher in CC erythrocytes.

Normal erythrocyte senescence is associated with the oxidation of hemoglobin molecules to hemichromes, which cluster band 3 and generate senescent antigen (Low et al., 1985; Waugh and Low, 1985). The clustering of band 3 may reveal cryptic epitopes for autologous IgG or result in the de novo formation of conformational epitopes (Guthrie et al., 1995; Lutz et al., 1984). Although the precise molecular structure of senescent antigen is unknown, its formation on the erythrocyte surface results in autologous IgG binding and complement fixation. Senescent erythrocytes can be opsonized for efficient removal from the circulation by splenic macrophages (Giger et al., 1995; Lutz et al., 1988).

P. falciparum also has been shown to induce hemichrome deposition, band 3 clustering, autologous IgG binding, complement fixation, and phagocytosis by macrophages (Fig. 5) (Giribaldi et al., 2001). Although the mechanisms involved in band 3 clustering might vary between senescent and parasitized erythrocytes, the downstream events in the opsonization pathway would likely be the same.

Fig. 5.

Proposed model for the mechanism of fluorescence reduction due to band 3 aggregation in AA and CC erythrocytes. Anti-band 3 mAb against cdb3 has easy access to band 3 epitopes in uninfected AA erythrocytes. During the course of P. falciparum infection, however, hemichromes form and bind with high affinity to the cdb3, hindering the anti-band 3 mAb binding. Band 3 also clusters to form oligomeric forms which may provide better recognition site for autologous antibody. Uninfected CC erythrocytes have increased band 3 oligomerization compared to uninfected AA erythrocytes, and a considerable amount of hemichromes already bound to the cdb3 that make those epitopes less accessible to anti-band 3 mAb, as shown in Fig. 1B. When these CC erythrocytes become infected, hemichromes form even larger deposits that further reduce antibody accessibility to band 3 epitopes, and band 3 forms higher levels of clustering, as shown Fig. 1A, Fig. 3B, Fig. 4E and 4F.

P. falciparum-induced changes in band 3 clustering have been proposed to enhance the splenic clearance of parasites in vivo during the course of a clinical malaria episode (Hogh et al., 1994; Kennedy, 2001). The potential importance of this mechanism in malaria protection is highlighted by several reports showing that this opsonization scheme operates with higher efficiency in erythrocytes carrying malaria protective mutations. For example, enhanced phagocytosis of parasitized erythrocytes has been proposed as a mechanism for the malaria protective effects of glucose-6-phosphate dehydrogenase deficiency, sickle-trait, and β-thalassemia trait (Ayi et al., 2004; Cappadoro et al., 1998).

Whether enhanced band 3 clustering of parasitized erythrocytes contributes to malaria protection by HbC remains to be determined. By undergoing accelerated oxidation to hemichromes and binding cdb3 with higher affinity, HbC erythrocytes would have a greater propensity than HbA erythrocytes to cluster band 3. Our recent identification of elevated hemichromes and autologous IgG in freshly drawn HbC erythrocytes (R.M.F. et al., unpublished) is consistent with this hypothesis and raises the possibility that band 3-directed opsonization could enhances the clearance of parasitized HbC erythrocytes and contribute to malaria protection in vivo.

## Acknowledgements

We thank Hongjian Liu, Quantum Dot Corporation, for helpful suggestions in the use of quantum dots, Jonathan Girroir and Steve Carano, AutoQuant Imaging, for assistance in using AutoDeblur, Takayuki Arie, LMVR, NIAID, NIH, for helpful discussions on mathematical analyses, and John Andersen, LMVR, NIAID, NIH, for technical assistance with SE-HPLC. Although certain commercial equipment, instruments and materials are identified in this paper, such identification does not imply recommendations or endorsements by NIST, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

• Accepted November 24, 2004.

View Abstract