Magnetic Bead-Based Separation (MBS) of Pneumococcal Serotypes

The separation of pneumococcal serotypes from a complex polymicrobial mixture may be required for different applications. For instance, a minority strain could be present at a low frequency in a clinical sample, making it difficult to identify and isolate by traditional culture-based methods. We therefore developed an assay to separate mixed pneumococcal samples using serotype-specific antiserum and a magnetic bead-based separation method. Using qPCR and colony counting methods, we first show that serotypes (12F, 23F, 3, 14, 19A and 15A) present at ∼0.1% of a dual serotype mixture can be enriched to between 10% and 90% of the final sample. We demonstrate two applications for this method: extraction of a known pneumococcal serotype from saliva samples and efficient purification of capsule switch variants from experimental transformation experiments. Moreover, this method may have further laboratory or clinical applications when the selection of specific serotypes is required.


INTRODUCTION 34
Streptococcus pneumoniae (pneumococcus) is an opportunistic pathogen that resides 35 asymptomatically in the upper respiratory tract of many healthy adults and children worldwide. 36 This asymptomatic colonization is a pre-requisite for the development of pneumococcal 37 disease, including upper respiratory tract infections (such as otitis media), lower respiratory 38 tract infections (such as pneumonia), and invasive pneumococcal disease (IPD) (such as 39 meningitis and bacteremia). Pneumococcal disease often occurs in the very young, elderly, or 40 immunocompromised 1 . Pneumococcus is a leading cause of lower respiratory disease, and 41 contributed to 1,189,937 deaths globally in 2016. 2 42 The capsular polysaccharide (CPS) is the outermost layer of encapsulated strains of S. 43 pneumoniae, and more than 100 antigenically distinct serotypes have been identified. 3 44 Pneumococcal conjugate vaccines (PCV) are highly effective against pneumococcal disease 45 but only cover up to 20 of these serotypes. While pneumococcal disease declined following the 46 introduction of PCVs, a concomitant increase in disease caused by non-vaccine serotypes 47 occurred. This emergence of non-vaccine serotypes in carriage and invasive disease is called 48 serotype replacement. 4 Serotype replacement occurs for two reasons, first the opening of a new 49 niche in which existing strains expressing capsules not targeted by the vaccine can thrive. 50 Second, vaccine-targeted strains can acquire the capsule biosynthesis cassette from a different 51 serotype, allowing them to evade vaccine-induced immunity. Serotype switching occurs when 52 the cps locus from one S. pneumoniae serotype (or related species) is transferred into the 53 genetic backbone of another S. pneumoniae serotype by transformation. 5 Genetic exchange 54 between two S. pneumoniae serotypes requires co-colonization of two or more serotypes. 55 56 In addition to naturally occurring serotype switches, 6-8 researchers have been generating cps 57 switch mutants in the lab for nearly 100 years. The first capsule switch experiments conducted 58 by Griffith in 1928, were accomplished by mixing avirulent, unencapsulated pneumococci with 59 virulent, but killed, encapsulated strains, and injecting this mixture into a mouse. The capsule-60 switched strains could then be isolated from the mouse. 9 More recently, generating cps switch 61 mutants in the lab has been accomplished using various genetic cassettes. 10,11 These types of 62 studies have permitted the generation of a number of capsule switch mutants, and this allows 63 for detailed experimental evaluation of the relative importance of capsule and genetic 64 background for different phenotypes. 8,12,13 Current methods for generating capsule-switched 65 variants require the use of selectable markers, are labor intensive, and not easily scalable. 66 Methods that allow for separation of multiple serotypes could eliminate the need for selection 67 pressure altogether or could be used in combination to conduct higher throughput 68 transformations. 69 70 There is also a need to isolate individual pneumococcal strains from clinical samples. 71 Nasopharyngeal swabs have long been considered the gold standard sample type for the 72 detection of carriage of S. pneumoniae, 14 but recent studies have demonstrated utility for saliva 73 to improve the detection of carriage in adults. 15,16 Whilst testing saliva improves the detection 74 of pneumococci when using molecular methods (such as qPCR), it can be challenging for the 75 isolation of live pneumococcal colonies due to the density and diversity of bacteria present in 76 saliva. A method that enables the separation of pneumococci, in a serotype-specific manner, 77 from other species present in saliva would be useful for clinical and laboratory studies alike. 78 79 We developed a magnetic bead-based separation (MBS) method which requires no selection 80 markers and can be used to extract live pneumococci, of a known serotype, from a mixture of 81 pneumococci or from clinical samples containing other bacteria (such as saliva). 82 83 METHODS 84 85 Figure 1 summarizes the MBS method; briefly, a mixture of serotypes is incubated with 86 antisera pool(s) unique to the desired serotype, then following wash steps is incubated with 87 secondary antibody conjugated to a magnetic bead. for serotyping are outlined in Table S1. 122 123 124  To demonstrate proof of concept for the MBS method we used three pairs of six different  125  serotypes where one serotype in each pair was penicillin resistant and the other penicillin  126 sensitive. It is important to note that different penicillin sensitivity is not necessary for 127 separation but was instead used to make the quantification of the efficiency of this method 128 easier. The three pairs were 12F and 23F (Pair 1), 3 and 14 (Pair 2) and 19A and 15A (Pair 3). 129

Proof of concept and primary analysis
Serotype 3 exists as two distinct morphologies; small non-mucoid colony variant (SCV) and 130 mucoid variant. 17 We therefore isolated SCV and mucoid variants and chose to work primarily 131 with the SCV for three reasons; SCVs are easier to count, easier to isolate as single colonies 132 (for serotyping) and less easy to distinguish from other serotypes based on morphology, thus 133 reducing selection bias during the colony selection for serotyping. The MIC of each serotype 134 was determined using penicillin E-strips, and then the exact concentration of penicillin for 135 blood agar plates was determined experimentally by varying the penicillin concentration and 136 plating out cells at known CFU/mL. The concentration of penicillin used in the blood agar 137 plates was the concentration at which the resistant serotype grew equally well on a penicillin 138 containing plate, as it did on a plain plate, whilst the susceptible serotype showed no growth 139 on the penicillin containing plate but normal growth on a plain plate. For Pairs 1, 2 and 3, BAPs 140 containing 0.018 µg/mL, 0.036 µg/mL and 0.18 µg/mL penicillin were used, respectively. 141 142 For all three pairs, Sample R is when the penicillin resistant serotype is the minority species, 143 and Sample S is when the penicillin sensitive serotype is the minority species. Samples were 144 plated out onto BAPs with and without penicillin, at two stages in the protocol; immediately 145 prior to the first incubation (PRE), and after extraction (POST). In all cases 5 µl of sample was 146 serially diluted in 45 µl PBS, in triplicate. For samples where the minority strain was penicillin 147 resistant, 20 µl of sample at a 10 -1 dilution was plated on penicillin plates, while 20 µl of sample 148 at a 10 -4 dilution was plated on plain blood agar plates. In samples where the majority serotype 149 was penicillin resistant, 20 µl of sample at a 10 -4 dilution was plated on both BAPs with and 150 without penicillin. In addition to the diluted samples, 10 µl of undiluted sample at the PRE and 151 POST stage, and the remaining volume (~40 µl) after elution was plated on BAPs, to provide 152 DNA for qPCR experiments conducted to establish separation efficiency. In all cases 10 µl or 153 20 µl samples were pipetted onto the BAP and the plate was then tilted to allow the sample to 154 run down the length of the plate. The BAPs were incubated overnight. 155 156 Secondary analyses 157 To establish if separation efficiency was similar for both mucoid (Muc) and single colony 158 variants (SCV) of Serotype 3, two additional pairs; 23F and 3SCV (Pair 4), and 23F and 3Muc 159 (Pair 5) were investigated. These experiments were conducted in duplicate, and efficiency 160 assessed by colony counting and qPCR methods. Pair 4 and 5 used BAPs containing 0.072 161 µg/mL penicillin. 162 163 To investigate the effect of initial proportion of minority serotype on the efficiency of 164 separation, 23F and 12F (Pair 1) were again used. The initial amount of majority serotype (12F) 165 was kept constant at 1x10 7 CFUs, while the minority serotype (23F) was varied (5x10 4 , 1x10 4 , 166 5x10 3 and 1x10 3 ). These experiments were conducted once for each dilution, and efficiency 167 was assessed by colony counting and qPCR methods. 168 169 The experiments above were conducted using two pooled antisera that were specific for the 170 minority serotype. We investigated whether a single pool of antisera could also be used. This 171 is important because certain pairs of serotypes can only be distinguished by one pool. Serotype 172 pairs which could not be distinguished based on penicillin sensitivity (and therefore could not 173 be assessed by colony counting methods), were used for this analysis, and for pairs which 174 shared a common antisera pool, only the unique antisera was used. These experiments were 175 conducted once for each condition, and efficiency was assessed by qPCR alone. 176 177 Colony counting to quantify separation efficiency 178 Colonies were counted and the mean colony number was determined, which was then used for 179 downstream analysis. For the mixed transformation sample, all colonies were harvested using a cotton swab and 245 resuspended in 1.5 mL Brain Heart Infusion (BHI) media + 10% (v/v) glycerol. As a control, 246 100 µl of the mixed sample was serially diluted to 10 -6 , then 100 µl of 10 -4 , 10 -5 and 10 -6 247 dilutions were plated on BAP, and incubated overnight. Following, 100 µl of the mixed sample 248 was aliquoted into four 1.5 mL Eppendorf tubes, centrifuged at 13,000 rpm resuspended in 500 249 µl Buffer 1 and processed through MBS using the appropriate antisera pool(s) for targeting the 250 appropriate serotype. The elution was plated on BAP and incubated overnight. Thirty-two 251 colonies were selected from the mixed sample that did not undergo MBS, and eight colonies 252 were selected from each of the four samples that had undergone MBS. were plotted against CFU/mL of 19A in the raw saliva sample ( Figure S1). 272 273 Using data from Figure S1 in combination with data from previous studies 19 we were able to 274 determine suitable concentrations for spiked-saliva, that reflect levels commonly found in 275 saliva obtained from the healthy individuals during carriage studies. Pneumococcus-negative 276 saliva was spiked with pneumococci (serotype 19A) at varying concentrations (5x10 4 , 5x10 3 , 277 5x10 2 and 5x10 1 CFU/mL) and left at room temperature for 2 hours. Following, 100 µl of each 278 sample was plated onto Gent plates and incubated overnight. The lawn of the culture-enriched 279 saliva was harvested into 2100 µl BHI + 10% (v/v) glycerol. 280 281 From each culture-enriched saliva sample, 10 µl was added to 490 µl Buffer 1, and cell 282 separated using the MBS protocol, with the following modifications. The primary incubation 283 step was conducted using SSI antisera (~16.8 µg total protein) and SunFire Bio monoclonal 284 antibody (mAb) (~16.8 µg total protein) combined. The secondary incubation was conducted 285 using ~48 µg total protein of anti-mouse IgG MicroBeads (Miltenyi Biotech) to target the mAb 286 only. As a negative control, culture-enriched saliva samples did not undergo MBS and were 287 instead serially diluted to 10 -6 , the 10 -4 , 10 -5 and 10 -6 dilutions were plated on BAPs and 288 incubated overnight. 289 290 Colonies that looked like pneumococci (small, grey, moist colonies with a green zone of alpha-291 hemolysis), were isolated and expanded onto new BAP: sixteen colonies from each sample 292 (with and without cell separation) that contained 5x10 4 , 5x10 3 , 5x10 2 CFU/mL of 19A in raw 293 saliva samples, and 24 colonies from the sample (with and without cell separation) that 294 contained 5x10 1 CFU/mL of 19A in raw saliva. Each expanded colony was optochin tested to 295 confirm whether it was pneumococcus (optochin sensitive) or another oral bacteria (optochin 296 resistant). Where a ring of optochin sensitivity was observed but a second (contaminating) 297 bacteria with optochin resistance was also present or, where satellite colonies of pneumococcus 298 were present within the zone of inhibition, samples were considered 'pneumococcal colonies' 299 since pure pneumococci can be isolated from the contaminant. 300 301 RESULTS 302 303 The MBS proof of concept experiments showed that for all six serotypes, the minority serotype 304 was successfully enriched from ~0.1% starting percentage to between 13% (serotype 14) and 305 90% (serotype 3) post MBS, corresponding to a 100-to-900-fold enrichment (Figure 2a). The 306 final percentage of the minority varied between serotypes but was relatively consistent between 307 the three replicates. There was generally good concordance in the estimated MBS efficiency as 308 determined by the qPCR and colony counting (Figure 2b), however efficiency determined by 309 colony counting seemed to be higher and lower than with qPCR for serotype 14 and 3, 310 respectively. Eight colonies from each elution plate were selected at random and in every single 311 case, minority serotype colonies were identified by serotyping (Table 1). This demonstrates 312 that this technique can be used to recover a desired serotype from a dual mixture.

321
A secondary analysis was conducted to identify whether serotype 3Muc was also enriched with 322 a similar efficiency as serotype 3SCV, and to gain insight into how separation efficiency varies 323 when the majority serotype of the pair is altered. MBS was conducted on Pair 4 (23F and 3SCV 324 and Pair 5 (23F and 3Muc). The results were compared to MBS results obtained previously for 325 enrichment of minority serotypes 23F or 3SCV when paired with another majority serotype 326 (namely serotype 12F and serotype 14 from Pair 1 and Pair 2, respectively). The percentage 327 enrichment for both 23F and 3 remained similar even when the majority serotype of the pair 328 was altered (Figure 3). Furthermore, it demonstrates that the MBS method permits successful 329 enrichment of both SCV and mucoid variants of serotype 3, and that the efficiency is similar 330 regardless of the morphology. In all cases minority serotype single colonies were isolated from 331 the elution plate by selection of single colonies and confirmed to be the desired serotype using 332 SSI latex agglutination (Table 1)

354
Results shown are singlicate data points only.

355
Additional analysis aimed to determine whether enrichment was constant at different % 356 minorities. The 23F and 12F pair were used with the majority serotype (12F) remaining 357 constant at 1x10 7 CFUs and the minority serotype (23F) at four different concentrations in the 358 initial sample. Enrichment of the minority serotype can be achieved even when the starting 359 percentage of a minority serotype is as low as 1x10 3 CFUs. However, as the initial % minority 360 decreases the percentage minority recovered following MBS also decreases. For initial samples 361 containing 5x10 4 , 1x10 4 , 5x10 3 and 1x10 3 CFU's of minority serotype 23F, the corresponding 362 percentage of 23F present in the final samples were 27%, 14%, 8% and 6% respectively as 363 determined by qPCR, or 49%, 23%, 13% and 9% respectively as determined by colony 364 counting (Figure 4b). 365

366
In order to separate serogroups that share reactivity to one antiserum pool, the MBS method 367 should be used with only a single antiserum pool. We therefore investigated outcomes when 368 using one or two antisera Pools and compared the efficiency of antisera pools in the presence 369 of different majority serotypes. MBS of serotype 14 from a majority serotype 3, using both 370 antisera Pool H and Pool P, resulted in the final sample containing ~13% of serotype 14. 371 However, use of only Pool H or Pool P, at an equal final volume to the combined pools, resulted 372 in serotype 14 being 10% and 45% of the final samples respectively. Therefore, in this example, 373 Pool P alone achieves the greatest efficiency of MBS, but in the absence of knowing which 374 antisera is more efficient, and if the serotype pairs permit dual use, it would be prudent to 375 combine both antisera pools. Furthermore, we confirm that the overall efficiency of enrichment 376 achieved by any antisera pool, is not only dependent upon the minority serotype alone, but also 377 the majority serotype. and compared the success of identifying pneumococcal colonies in the presence and absence 420 of MBS (Table 4). For both saliva A and saliva B, at all concentrations of 19A, the MBS 421 method resulted in equal or improved isolation of pneumococcal colonies. In saliva A, the MBS 422 method was still able to enrich for pneumococcus when the concentration of 19A was 5x10 1 423 CFU/mL in raw saliva, however for Saliva B the MBS method was only successful at a 19A 424 concentration of 5x10 3 CFU/mL in raw saliva. The sensitivity of this assay is therefore 425 dependent upon not only the concentration of pneumococci in the sample but also the 426 composition of saliva itself, and may vary from sample to sample. The MBS method was then 427 tested on a saliva sample that had tested qPCR-positive for serotype 15B/C but from which we 428 had been unable to isolate pneumococcus using the standard culture-based dilution method. 20 429 Here, the MBS method successfully enriched for pneumococcus in the sample, and of the 32 430 colonies selected, 29 were confirmed to be pneumococcus. 431 432 presence in the sample using the colony counting method, but qPCR would provide a more 458 accurate estimation. Despite some differences in efficiency estimates between colony counting 459 and qPCR methods, we were able to successfully isolate minority serotype colonies post MBS 460 in all cases. This demonstrates a tangible utility for this method in the laboratory setting. When 461 separating a mixture of cells only a small number of colonies must be isolated to identify the 462 desired serotype. This method therefore allows for the easy recovery of serotype-specific S. 463 pnuemoniae isolates. 464 465 In the secondary supporting analysis, we compared how enrichment of a minority serotype 466 varied when in the presence of different majority serotypes. A minority serotype 23F was 467 paired with a majority serotype of either 12F or 3, and minority serotype 3 was paired with a 468 majority serotype of either 14 or 23F. With minority 23F, some variation in efficiency of MBS 469 was noted when the majority serotype was changed, however for minority serotype 3, the 470 enrichment efficiency remained very similar despite the change in majority serotype pair. This 471 suggests that the serotype with which the minority is mixed may have some impact on the 472 efficiency of MBS, but it is likely primarily determined by the avidity of the antisera for the 473 desired serotype. Unlike the majority of pneumococcal serotypes, serotype 3 utilizes the 474 synthase-dependent pathway for CPS production, resulting in non-covalently bound CPS 475 which can be released from the glycolipids or synthase. 21 The CPS of serotype 3 is not 476 covalently linked to the peptidoglycan and can be released, 22 which leads to a reduction in the 477 protective effect of anti-Type 3 CPS antibodies induced by the PCV13, 23 we were therefore 478 surprised to find that the MBS method can successfully extract serotype 3 from a mixed sample. 479 This success may be explained by the fact that the cells are not actively growing and likely 480 therefore not releasing CPS into the environment. Furthermore, it is intriguing but reassuring 481 that the efficiency of enrichment between mucoid and SCV serotype 3 is very similar; the MBS 482 method can be successfully used on serotype 3 samples, which are of particular interest due to 483 the reduced effectiveness of PCV13 on serotype 3 IPD. 24-26 484 485 We demonstrate that good separation can be achieved with only one unique antiserum, meaning 486 that serotypes with cross-reactivity to one antiserum can still be separated using this method. 487 As expected, we demonstrate that the efficiency of enrichment achieved by each of the two 488 antisera pools is not equal and therefore, depending on the desired serotype one antisera may 489 be preferred over another. Furthermore, enrichment of a serotype can occur even when a 490 serotype is present at only 0.01% of the total sample (1x10 3 minority serotype with 1x10 7 491 majority serotype). 492 493 A key limitation of the MBS method in general, is that due to cross-reactivity within 494 serogroups, SSI antisera Pools can only be used to separate S. pneumoniae serotypes belonging 495 to different serogroups. However, use of type-specific antisera or a mAb instead of pooled 496 antisera would circumvent this limitation. Another limitation is the total proportion of minority 497 cells that can be recovered. While enrichment from 0.1% up to >10% has been demonstrated, 498 it is worth noting that only a small proportion (~1%) of the total minority cells present in the 499 initial mixture are successfully extracted. This may be overcome by increasing antibody 500 incubation periods or antibody concentration to increase binding capacity. 501 502 Having optimized the MBS method, we evaluated its potential for laboratory applications. was not isolated using the other methods. This suggests that the MBS method may be 518 particularly useful to enrich for serotypes which transform with low efficiency. The MBS 519 technique was not 100% specific, and a small amount of cross-reactivity was observed, 520 however, since each sample is enriched for the desired serotype, and the serotype of each 521 colony is confirmed by latex agglutination, these contaminants are of little concern are for this 522 particular application. 523 524 We also show that the MBS method can be modified to successfully enrich for pneumococci 525 of a known serotype from saliva samples. Enrichment is possible even in saliva samples where 526 pneumococci is present at very low concentrations (5x10 1 CFU/mL), for which isolation of 527 pneumococci using standard methods is typically very challenging. This permits easy 528 identification and isolation of pneumococci present in saliva at concentrations too low to detect 529 using standard dilution and plating methods. The use of SSI antisera alone on a polymicrobial 530 sample such as saliva was problematic due to antisera reactivity with non-pneumococcal 531 bacteria present in saliva. In general, we found that the SSI antisera outperformed mAbs in 532 terms of total number of pneumococcal colonies isolated, we hypothesize that this is due to the 533 increased avidity of antisera (presence of IgA, IgM) which agglutinates pneumococci 534 increasing the overall yield during MBS. Therefore, to take advantage of the increased avidity 535 of antisera and simultaneously the high specificity of mAbs, we combined both in the primary 536 incubation step, but only targeted the mAb in the secondary antibody step. While the elution 537 from saliva samples was not 100% pure pneumococci, contaminating non-pneumococcal 538 bacteria was reduced, and identification and selection of single pneumococcal colonies was 539 improved when compared to the standard dilution and plating method. The enrichment 540 observed varies depending on concentration of pneumococci present in the sample, but also on 541 the saliva composition itself. The composition of bacterial community in saliva varies between 542 different age groups 27 and so the success of the MBS method will likely vary accordingly. 543 Since