Direct PCR offers a fast and reliable alternative to conventional DNA isolation methods for animal gut microbiomes

The gut microbiome of animals is emerging as an important factor influencing ecological and evolutionary processes. A major bottleneck in obtaining microbiome data from large numbers of samples is the time-consuming laboratory procedures, specifically the isolation of DNA and generation of amplicon libraries. Recently, direct PCR kits have been developed that circumvent conventional DNA extraction steps, thereby streamlining the laboratory process by reducing preparation time and costs. However, the reliability and efficacy of the direct PCR method for measuring host microbiomes has not yet been investigated other than in humans with 454-sequencing. Here, we conduct a comprehensive evaluation of the microbial communities obtained with direct PCR and the widely used MoBio PowerSoil DNA extraction kit in five distinct gut sample types (ileum – caecum – colon – faeces – cloaca) from 20 juvenile ostriches, using 16S rRNA Illumina MiSeq sequencing. We found that direct PCR was highly comparable over a range of measures to the DNA extraction method in caecal, colon, and faecal samples. However, the two methods recovered significantly different microbiomes in cloacal, and especially ileal samples. We also sequenced 100 replicate sample pairs to evaluate repeatability during both extraction and PCR stages, and found that both methods were highly consistent for caecal, colon, and faecal samples (rs > 0.7), but had low repeatability for cloacal (rs = 0.39) and ileal (rs = −0.24) samples. This study indicates that direct PCR provides a fast, cheap, and reliable alternative to conventional DNA extraction methods for retrieving 16S data, which will aid future gut microbiome studies of animals.

microbiomes relies on large sample sizes, and so it is important to find fast, cost effective, 64 and reliable ways of processing microbiome samples.

66
The conventional way of generating amplicon libraries for microbiome studies is to first 67 extract and purify DNA, for example, using kits such as the MoBio PowerSoil DNA isolation 68 kit. This procedure is recommended by the Earth Microbiome Project (Caporaso et al. 2012; 69 Gilbert et al. 2014), and is widely used in human and non-human animal microbiota studies. 70 The DNA extraction protocol involves mechanical and chemical lysis of cells, and a DNA 71 purification procedure, which adds up to 32 separate steps (Table 1)  control and 2 blank samples, 40 extraction replicates, and 10 PCR replicates, that were all 220 prepared both with the DNA extraction and the direct PCR method (Table S1). An additional 221 11 control and 2 blank samples from a subsequent run were also evaluated to increase the 222 number of controls (Table S1). Sequencing of blank (negative) samples resulted in extremely few sequence reads: three 267 blank samples had < 320 reads and the other three had < 3,000 reads, compared to an average 268 of 10,689 reads for the other sample types ( Figure S1). Control swabs showed highly 269 dissimilar microbial composition to all other samples (see Videvall et al. 2017) and therefore 270 we did not include these in any further analyses. 271 To evaluate bacterial abundances, we first filtered out all OTUs with less than 10 272 sequence reads and then, using DESeq2 (v. 1.14.1), counts were modelled with a local 273 dispersion model and normalised per sample using the geometric mean (Love et al. 2014). and Hochberg false discovery rate for multiple testing (Benjamini & Hochberg 1995). OTUs 279 were labelled significant if they had a corrected p-value (q-value) < 0.01. 280 We examined the repeatability of the two methods by evaluating the strength of the 281 correlation in normalised OTU abundance between paired sample replicates. This was done 282 separately for the two methods, and for the two replicate sets (extraction replicates and PCR 283 replicates). Correlation coefficients were calculated using Spearman's rank correlations on all 284 OTUs with non-zero abundances. 285 286

Practical aspects of direct PCR and DNA extraction 289
The total time spent extracting DNA using the direct PCR method was considerably shorter 290 (45 minutes) compared to the conventional DNA extraction method (8 hours) ( Table 1). The 291 cost of using the direct PCR method was also lower, as were the number of steps in each 292 protocol (Table 1). Nevertheless, the number of sequence reads obtained per sample (mean = 293 10,689) did not differ between the DNA extraction method and the direct PCR method (two 294 sample t-test: t = 1.25, df = 290.7, p = 0.21) (Figure S1), as expected since equimolar PCR 295 products from the samples were combined before sequencing. 296 297 298

Description of the microbiomes obtained with direct PCR and DNA extraction 299
Samples clustered strongly according to sample type both in the Principal Coordinates 300 Analysis (PCoA) ( Figure 1A) and the network analysis ( Figure 1B), although some minor 301 separation between the direct PCR and DNA extraction methods was evident. The two library 302 preparation methods yielded fairly consistent patterns for the total number of OTUs per 303 bacterial class and sample type, but also reflected some notable differences in taxa 304 composition ( Figure 1C). Specifically, Bacilli were slightly more abundant in all sample 305 types with the DNA extraction method relative the direct PCR method, and the ileum in 306 particular showed the largest class differences, with a higher abundance of Mollicutes, 307 Gammaproteobacteria, and Bacteroidia with the direct PCR method ( Figure 1C). 308 To further investigate if the microbiota of samples from the direct PCR and DNA 309 extraction methods differed depending on gut site, we performed separate PCoAs for each of 310 the five sample types. The caecum, colon, and faeces showed very high correspondence in 311 beta diversity for identical samples prepared using direct PCR and DNA extraction, as they 312 clustered by individual and not method, whereas differences were much greater for cloacal 313 and ileal samples (Figure 2). 314 315 316

Differences in microbiomes obtained with direct PCR and DNA extraction 317
We next evaluated differences in alpha diversity (OTU richness) between direct PCR and 318 DNA extraction methods for the different sample types ( Figure 3A). There were no 319 differences in alpha diversity between the two methods in the colon, faeces, and cloaca 320 (paired Wilcoxon signed rank test: npairs = 20 per type, p > 0.4) ( Figure 3A). However, alpha 321 PCR caecal samples (npairs = 20, V = 199, p = 0.0001) ( Figure 3A). Correlation analyses of 324 alpha diversity between the two methods also showed higher diversity for ileal direct PCR 325 samples and slightly lower diversity for caecal direct PCR samples, but the strength of the 326 correlations between methods were generally high for all sample types (r = 0.56-0.91) 327 ( Figure S2). 328

Dissimilarities in microbiome composition between samples, as calculated by the 329
Bray-Curtis distance measure, showed significant effects of method, sample type, individual, 330 and the interaction between method and sample type (PERMANOVA: all effects: p < 0.001). 331 The overall variance explained by method and method*sample type was, however, extremely 332 small (R2 = 0.014, and R2 = 0.019, respectively), whereas the variance explained by host 333 individual (R2 = 0.283) and sample type (R2 = 0.201) were substantially larger. 334 Examining the Bray-Curtis distances between the two methods within sample types 335 revealed that while the caecum, colon, and faeces all showed relatively low distances (mean: 336 0.23, 0.25, and 0.29, respectively), the cloaca (mean = 0.39) and most notably, the ileum 337 (mean = 0.56), displayed much greater distances and much higher variances ( Figure 3B). 338 Specifically, the distances between identical samples prepared with each method from the 339 ileum were significantly higher than the corresponding distances in the caecum, colon, and Analyses of differences in the abundance of specific OTUs between the two methods 354 resulted in very few significantly different OTUs in the caecum (n = 9), colon (n = 13), and 355 faeces (n = 24) (Figure 4). However, there were many more in the cloaca (n = 67), and the 356 ileum demonstrated a staggering 324 significant OTUs between the DNA extraction and the 357 direct PCR method (Figure 4). Notably, the vast majority of significant OTUs across all 358

extraction. 360
Comparing the exact OTUs that had significantly different abundances in the five 361 sample types, showed that they were unique to each sample type (i.e. OTUs were only 362 significantly different within one type) (Table S2). We found one genus, however, 363 Mycoplasma, with significant OTUs present in all sample types. All significant Mycoplasma 364 OTUs (family: Mycoplasmataceae) had higher relative abundances in the samples from the 365 direct PCR compared to the DNA extraction method ( Figure 5). Other genera with 366 significantly different abundances in multiple sample types were e.g. Anaerofustis (higher 367 abundance with the direct PCR method in colon, faecal, and cloacal samples) and Klebsiella 368 (more numerous in the direct PCR method of ileum, caecum, and colon). The genus 369 Prevotella (class: Bacteroidia) was the most prevalent in the list of significant genera, 370 representing 43 unique OTUs in the cloaca and ileum, all of which (100%) had higher 371 abundance in the direct PCR samples (Table S2). 372 The phylum with the highest number of significantly differentially abundant OTUs 373 was the Firmicutes (n = 210 in total), and in particular the class Clostridia (n = 171 in total) 374 ( Figure 5; Table S2), which was in majority in most sample types ( Figure 1C). The genera 375 which comprised the most significant differentially abundant OTUs between the two methods 376 were the Oscillospira (class: Clostridia), Mycoplasma (class: Mollicutes), and Coprococcus 377 (class: Clostridia) (Table S2). 378

Repeatability of replicate samples with direct PCR and DNA extraction 381
Next, we evaluated the repeatability of the DNA extraction and direct PCR methods by 382 calculating correlations of OTU abundances and diversity between pairs of replicate samples. 383 For the "extraction replicates", the correlation coefficient of OTU abundance was almost 384 identical for the DNA extraction method (rs = 0.73; Figure 6A) and the direct PCR method (rs 385 = 0.70; Fig 6B). For the "PCR replicates", the strength of the correlation was slightly higher 386 but again similar for the two methods (DNA extraction: rs = 0.82, Figure 6C; and direct PCR: 387 rs = 0.80, Figure 6D). 388 When we partitioned the OTU abundance data according to sample type, large 389 differences in repeatability were observed ( Figure 6E). The caecal, colon, and faecal samples 390 had the strongest correlations between replicates, with both methods having an average rs = 391 0.70-0.74 for the extraction replicates and an average rs = 0.76-0.83 for the PCR replicates 392 ( Figure 6E). In contrast, the extraction replicates from the cloaca were characterised by a 393 much weaker correlation (rs = 0.36), as did the cloacal PCR replicates (rs = 0.49), and the 394 correlations between ileal replicates were even negative (extraction replicates: rs = -0.27, PCR 395 replicates: rs = -0.05) ( Figure 6E). 396 Finally, we examined the correlation between alpha diversity estimates in the replicate 397 samples to evaluate the repeatability of the community between methods. Relative to the 398 OTU abundance data, there was higher repeatability in alpha diversity for cloacal and ileal 399 samples (r = 0.79-0.96), while the caecal, colon, and faecal samples again had high 400 repeatability for both methods (r = 0.70-0.97) (Figures S4-S5). 401

403
This study shows that direct PCR provides highly comparable results to the widely used and 404 recommended DNA extraction method in analyses of gut microbiomes of animals. Both 405 techniques give qualitatively and quantitatively similar estimates of microbial diversity and 406 abundance for caecal, colon, and faecal samples, and were highly repeatable for these sample 407 types. However, the two methods present dissimilar microbiomes for cloacal, and in 408 particular, ileal samples, recovering large differences and poor repeatability in OTU 409 abundances across replicates. We discuss hypotheses that may explain why these methods 410 perform well with some sample types, but not others.  Figures S4-S5) and between the methods ( Figure S2). This suggests that although it is 424 difficult to measure relative abundances of specific bacterial taxa when DNA concentrations 425 are low, it may still be possible to gain an accurate measure of the community composition 426 using both of these methods.

428
The direct PCR samples from the ileum had significantly higher alpha diversity ( Figure 3A) 429 and higher relative abundances in the majority of differentially abundant OTUs (Figure 4), 430 compared to the DNA extraction method. One potential reason for this difference is that more 431 DNA is lost during the DNA extraction procedure, which is column-based with several wash 432 and transfer steps. This is known to be associated with high DNA loss, whereas with direct 433 PCR, the individual samples are contained in just one plate well during the full extraction 434 procedure. In samples with low starting DNA concentrations, direct PCR may therefore be 435 superior to conventional extraction methods at recovering rare bacterial taxa. The higher 436 diversity in the ileal direct PCR samples could also be a consequence of the regular Taq 437 polymerase included in the direct PCR reaction mix, which is slightly more error-prone than 438 OTUs, raise the alpha diversity and the number of differentially abundant OTUs. However, 441 Taq polymerase does not explain the consistent changes observed in the PCoA (Figure 2), the 442 correlation in diversity between the methods ( Figure S2), or the changes in relative 443 abundances of specific taxonomic groups ( Figure 1C and Figure 5). 444 445 Despite high correspondence in abundance estimates across most OTUs, there were some 446 consistent differences in the abundance of specific taxa between direct PCR and the DNA 447 extraction method. It is possible that morphological differences between bacterial groups may 448 influence the efficiency with which the two methods recover DNA. For example, compared 449 to the DNA extraction method, we found that direct PCR had higher relative abundances of     fold changes indicate higher relative OTU abundance in the DNA extraction method, and 640 negative log2 fold changes signify higher abundance in the direct PCR method. 641