Improved sensitivity, accuracy and prediction provided by a high‐performance liquid chromatography screen for the isolation of phytase‐harbouring organisms from environmental samples

HPLC‐based screening of soil identifies Multiple Inositol Polyphosphate Phosphatase as a contributor to aggregate soil phytase activity. HPLC also reveals the position of attack on phytate by different histidine phosphatases and affords opportunity for isolation of phytases for biotechnological use from other environments.


Introduction
There are four forms of phytic acid (inositol hexakisphosphate, InsP 6 ), which have been identified in nature, myo, neo-, scyllo-and D-chiro-, that differ in their stereochemical conformation (Fig. S1) and association with metal ions as phytates in different soils (Turner et al., 2002). Among these, myo-inositol hexakisphosphate (InsP 6 ) garners the most attention from plant scientists. It is the principal storage form of phosphorus in plants, seeds and grains representing between 50% and 85% of the total phosphate in plants and forming as much as 1-5% of the dry weight in many seeds, grains and fruits (Raboy and Dickinson, 1993).
Monogastric animals such as swine and poultry are fed diets that are largely cereal-and/or grain-based, but they lack sufficient levels of endogenous phytase, a mixed group of phosphatases that dephosphorylate phytate (Pandey et al., 2001). The undigested phytate and other 'higher' inositol phosphates are potent antinutrients by virtue of their ability to interfere with protein digestion and to chelate metal ions such as calcium, iron, magnesium, manganese and zinc, reducing their bioavailability. The first commercially produced phytase, Natuphosâ, was released to the market in 1991 to improve the digestibility of grain phytate in the gastrointestinal tract of non-ruminants (Lei and Porres, 2003). Since then, phytases have become a major sector of a global enzyme market of estimated value ca. $5 billion in 2015, with annual growth estimated at 6-8% from 2016 to 2020 (Guerrand, 2018).
The small fraction of environmental organisms amenable to culture has the consequence that the biodiversity of phytase producers is grossly underestimated. Consequently, metagenomic and metaproteomic approaches have supplanted culture-based approaches for study of the relationship of microbiological diversity and soil phosphorus (Neal et al., 2017;Yao et al., 2018;Chen et al., 2019). Alternatively, others have employed amplicon sequencing of functional phosphatases using phoD alkaline phosphatase-specific primers (Ragot et al., 2015). When allied with heterologous expression, metagenomic methods have revealed novel catalytic diversity among phytate-degraders extending classification beyond the four canonical classes (Castillo Villamizar et al., 2019a,b) as have more conventional functional genomic methods (Sarikhani et al., 2019).
Irrespective of the method of identification of candidate phytases, whether as commercial product leads or as contributors to environmental processes, both culture-independent approaches and their culture-dependent counterparts commonly rely on informative enzyme assays for characterization of the reactions catalysed. One issue with the assay most commonly used, phosphate detection with reagents such as molybdenum blue/malachite green, is the purity of the phytate substrate. Commercially available phytate is impure (Fig. S2A) and often contains substantial mole fractions of lower inositol phosphate and inorganic phosphate impurities (Nagul et al., 2015). Consequently, unless assays follow disappearance of phytate they risk measurement of pre-existing inorganic phosphate or risk misidentification of enzymic activity towards 'lower' inositol phosphates. The literature offers historic precedent: isolates capable of degrading InsP 5 but not InsP 6 were identified in a seminal study of phytase isolation (Cosgrove et al., 1970).
The issue of substrate quality is relatively solvable; purification of InsP 6 is well described (Cosgrove, 1980;Dorsch et al., 2003;Madsen et al., 2019), but rarely discussed in screening for phytases. 'Phytase-specific media' (PSM) (Howson and Davis, 1983;Kerovuo et al., 1998) are used widely and rely on formation of clearing (of phytate precipitate) zones around bacterial colonies. The method suffers a high rate of false positives, arising from bacterial secretion of low molecular weight organic acids capable of solubilizing the phytate precipitates (Iyer et al., 2017). This itself highlights another issue with the approachthat it is not suitable for screening at low pHa condition for which many commercial enzymes have been optimized. While solubilization may be overcome by a two-step counterstaining test to re-precipitate acid-solubilized phytate (Bae et al., 1999), re-precipitation does not indicate to what extent the available phytate has been degraded, since other 'higher' inositol phosphates can also be re-precipitated. Autoclaving of the medium can also result in phytate degradation (see Fig. S2B,C) and resultant change to the pH of the media. Overall, clearing zones may not exclusively indicate enzymatic hydrolysis of phytate in the plate (Fredrikson et al., 2002), while pH limitations of the method will necessarily be selective of the organisms cultured. There is therefore opportunity for sensitive methodologies that allow characterization of the substrate and its utilization. Here, we adopt the PSM methodology and supplement it with high-performance liquid chromatography (HPLC) to demonstrate a more accurate and quantitative method allowing screening and isolation of phytase-producing organisms from environmental samples. We also show how different isolates produce different inositol phosphate profiles from phytate and extend the analysis to soil samples supplemented with phytate to follow the activity of aggregate cohorts of microbes. A schematic diagram of the range of analyses enabled is shown (Fig. S3).

Results and discussion
Acid extraction of phytate from PSM plates The PSM plate approach is one of the most commonly used methods for isolation of phytase-positive organisms from soil, but it is not without the substantial drawbacks discussed above. Control strains of Escherichia coli-pDES17-Btminpp harbouring a plasmid-borne MINPP from Bacteroides thetaiotaomicron (Stentz et al., 2014), Bacillus subtilis strain ESKAPE (predicted to contain bPPhy) and Pseudomonas putida J450 (predicted to contain bPPhy) were each streaked onto fresh PSM plates and allowed to grow over three days at 30°C. All isolates generated clearing zones around their biomass on these PSM plates. Cores of agar from 'cleared' and 'non-cleared, cloudy' zones were extracted with HCl and the inositol phosphate profile thereof examined by HPLC (Fig. 1). While there can be slight differences in the efficiency of extraction between the cleared and cloudy zones, comparison of individual peaks within the respective profiles makes evident the different extents and pathways of phytate degradation by the strains.
All profiles from the 'non-cleared' zones show the predominant peak of InsP 6 with a retention time of c. 37 min and a smaller peak of InsP 5 [1/3-OH] contaminant with a retention time of c. 28 min, representing approximately 5% of total inositol phosphate in this 'clean' InsP 6 substrate. Inorganic phosphate (Pi) elutes with the solvent front at c. 2.8 min. In respect of the 'cleared' zones, the B. subtilis strain ESKAPE (Fig. 1A) showed a small amount of InsP 6 degradation with a small increase in InsP 5 [1/3-OH] and a concomitant increase in Pi. In this experiment, much of the InsP 6 remained. The InsP 5 [1/3-OH] peak is the expected product of the known InsP 6 D-3-phosphatase activity of the bPPhy (Kerovuo et al., 1998) originally characterized (Powar and Jagannathan, 1982). The E. coli-pDES17-Btminpp strain (Fig. 1B) showed considerably more activity, producing multiple peaks of InsP 5 , InsP 4 and InsP 3 intermediates, characteristic of MINPP enzyme (Haros et al., 2009;Tamayo-Ramos et al., 2012;Stentz et al., 2014). There is also a larger Pi peak. Finally, the P. putida strain ( Fig. 1C) showed little difference in the profile of 'cleared' vs. 'non-cleared' agar despite the known InsP 6 D-3-phosphatase activity of other Pseudomonas sp. (Cosgrove et al., 1970;Irving and Cosgrove, 1972).
Collectively, these comparisons demonstrate that zone clearing without careful normalization is a poor assay for phytate degradation even of well-characterized organisms. It does illustrate however that HPLC can be combined with media-based culture and extraction of agar for testing of phytate degradation to provide high sensitivity and diagnostic analysis of the likely enzyme activity, by the simple expedient of observation of the occurrence of InsP peaks not present in 'non-cleared' regions of agar plates.

Assay of phytate degradation by mixed population soil cultures
Phytate degradation may also be demonstrated with mixed cultures that might ordinarily be subjected to standard dilution and culture techniques for discrimination of individual isolates. In Fig. 2, we show the result of mixing soil with minimal medium containing InsP 6 as the sole phosphate source. The soil was untilled (for the season) agricultural soil from Fakenham, Norfolk, UK, which we used to first test the technique. In this experiment, this agricultural soil was incubated with shaking at 30°C. Degradation of InsP 6 was observed initially on day 3; by day 5 <5% of starting InsP 6 remained, consistent with the accumulation of Pi, which co-elutes with InsP 1 on this column-gradient method. The generation of multiple inositol phosphate peaks at all stages of dephosphorylation (InsP 5 , InsP 4 , InsP 3 and InsP 2 ) probably arises as a consequence of the action of several phytase enzymes, since the classification of phytases reflects predominant attack in discrete sequences and predominant accumulation of single InsP 5 and InsP 4 species. This experiment was repeated on five well-characterized soil or plant growth matrices, all of which showed evidence of phytase activity. The first (Fig. 3A) is Levington compost F2, obtained from the John Innes Centre, Norwich, UK. The second ( Fig. 3B) is soil sampled from Church Farm, the field study site of the John Innes Centre in Bawburgh, Norwich UK. The next three soils were sampled from two long-term field experiments from Rothamsted Research, Harpenden, UK (Supporting information). The first sample (Fig. 3C) was obtained from continuous arable plots growing winter wheat (Triticum aestivum L.) of the Highfield Ley-Arable Experiment. Also, from this site, soil was sampled from permanent bare fallow plots (Fig. 3E) that have been maintained crop-and weed-free by regular tilling for over 50 years. The bPPhy genes in both these soils have been characterized by shotgun metagenomics (Neal et al., 2017). The gene sequences identified show homology to genes identified in Bacillus, Paenibacillus, Alteromonas and Cyanothece species. Soil was also collected from a plot of the Broadbalk Winter Wheat Experiment (Fig. 3D). Shotgun metagenome analysis of DNA extracted from this soil similarly showed the bPPhy gene sequences to be homologous to those in Bacillus, Paenibacillus, Alteromonas and Cyanothece (Neal and Glendining, 2019).
All the soil plant growth matrix types degraded phytate when added to liquid medium, generating distinct phytate degradation profiles and, but for one concomitant accumulation of inorganic phosphate (Fig. 3).
While degradation of phytate by some matrices, Levington's compost (Fig. 3A), Church Farm (Fig. 3B) and Bare Fallow (Fig. 3D) proceeded to completion or close to it, indicated by predominant accumulations of Pi, other soils, which presumably had less abundant or active cohorts of microbes, yielded diagnostic InsP 5 peaks in the timescale of the experiment. Of the Rothamsted soils, the Broadbalk soil removed phytate from liquid media such that neither phytate nor lower inositol phosphates were recovered, at day 0 (not shown). We attribute this to sorption of phytate to soil particles as this has up to 35% clay content. Nevertheless, by supplementing the soil/liquid mixture with 1 mM phytate we were subsequently able to show that the soil and associated microorganisms were capable of processing added phytate over 8 days (Fig. 3E). For this soil, Pi did not accumulate in the mediumsuggesting that the microflora were efficiently scavenging the released phosphate.
To interrogate further the phylogenetic separation of histidine (acid) phosphatase between Acinetobacter spp. and Buttiauxella spp., revealed in Table 1, a diverse selection of accessions (reference genomes) of each sp. was searched by tblastn in NCBI with the different phytase reference sequences of Table 1 as query. The results are shown in Table S1, in which crosses indicate the presence of the different phytase proteins in selected genome-sequenced Acinetobacter and Buttiauxella strains yielding E value < 0.00005. Only a single histidine (acid) phosphatase was present in the Buttiauxella genomes analysed. These were either AppA phytases or bifunctional glucose-1-phosphatase/inositol phosphatases. The phytase complements of Acinetobacter genomes were more varied, revealing the presence of all different classes of phytase with the exception of protein tyrosine phosphatase. Additionally, while predominantly only containing a single phytase, there were some cases of Acinetobacter sp. containing two different classes of phytase: either MINPP and bPPhy, or histidine (acid) phosphatase and bPPhy.
Phytate degradation profiles of isolated Acinetobacter and Buttiauxella strains reveal distinct histidine phosphatase activities To confirm the ability of identified isolates bearing defined cohort(s) of phytase(s) to degrade phytate and to characterize those enzyme activities, the isolates Acinetobacter sp. AC1-2 (AC1-2) and Buttiauxella sp. isolate CH-10-6-4 were incubated with phytate and subjected to HPLC analysis (Fig. 4A,B). This demonstrated that enzymes associated with AC1-2 are promiscuous in their site of initial attack on phytate substrate, yielding among InsP 5 isomers a dominant 4/6-OH peak, a smaller 5-OH peak and little to no detectable degradation at the 1/3-position (Fig. 4A). Interestingly, strain CH-10-6-4 did not show any phytase activity in minimal medium, but it did however degrade 1 mM phytate when incubated in a 20 mM Tris-HCl and 0.1% NaCl solution (Fig. 4B).
While both the Acinetobacter and Buttiauxella strains showed preferential 1D-4/6 selectivity of attack on phytate, they differ in terms of the resulting InsP 4 intermediates: the Acinetobacter strain produced four InsP 4 intermediates, while the Buttiauxella strain produced two, a predominant peak with the chromatographic properties of d/l-Ins(2,3,4,5)P 4 and a minor peak with that of d/l-Ins(1,2,3,4)P 4 . Again, HPLC can be shown to distinguish between classes of phytase without assistance of 16S rRNA gene. The phytate degradation profile of the Buttiauxella isolate is characteristic of 1D-6-directed histidine (acid) phosphatase, that of the Acinetobacter strain was indicative of the MINPP subclass of the histidine (acid) phosphatases (Tamayo-Ramos et al., 2012;Stentz et al., 2014). Congruent with these predictions, strain CH-10-6-4 was shown by PCR to contain a histidine (acid) phosphatase, 100% identical at the amino acid level to that in Buttiauxella ferragutiae. Furthermore, the genome sequence of AC1-2 was shown to encode a MINPP 98.28% identical at amino acid level to that in Acinetobacter calcoaceticus. With this additional information, we undertook an alignment of phytase protein sequences for thirty-one histidine (acid) phosphatases and twenty-seven MINPPs using the online multisequence alignment tool MAFFT (Katoh et al., 2019), reporting the output as an Interactive Tree of Life, iTOL (Letunic and Bork, 2019) (Fig. 5). The results of this analysis split MINPP sequences into two clades, those whose origins are from animals and plants (Cho et al., 2006;Dionisio et al., 2007), and those from bacteria (Haros et al., 2009;Tamayo-Ramos et al., 2012;Stentz et al., 2014). Both are distinct from bacterial histidine (acid) phosphatases, with bacterial MINPPs more closely related to eukaryotic MINPPs than bacterial histidine (acid) phosphatases. Of the bacterial MINPPs, the Acinetobacter enzyme was more deeply rooted than the MINPPs of previously characterized gut commensals Bifidobacter and Bacteroides spp.

Improved, predictive HPLC-based screening for phytases
The foregoing analyses highlight considerations that apply to culture-dependent isolation of phytases, here from environmental samples. The methods described overcome problems associated with the purity of phytate substrate (Madsen et al., 2019) and 'zone-clearing' assays (Fredrikson et al., 2002). Nevertheless, PSM can be a useful media for obtaining a diverse set of bacteria (Greiner et al., 1997;Richardson & Hadobas, 1997;Kerovuo et al., 1998) or for the screening of engineered bacteria and plants (Shulse et al., 2019).
Here, the opportunity to characterize enzyme activity of isolates before functional cloning, expression, purification, subsequent verification of catalytic activity is a considerable shortcut that focuses attention among isolates on those with bona fide phytase activity. Moreover, sequencing of the Acinetobacter and Buttiauxella strains revealed the power of this HPLC-based screening strategy to illuminate phytase diversity. The two different histidine phosphatases, MINPP and histidine (acid) phosphatase, are typical of the families of enzymes identified in sequenced genera. The assembled sequenced genome (JABFFO000000000) of the Acinetobacter strain AC1-2 harbours a single histidine (acid) phosphatase of the MINPP class, rather than a canonical histidine (acid) phosphatase.
The enzyme bears a hepta-peptide catalytic site sequence motif of RHGSRGL: RHG is characteristic of the histidine phosphatase superfamily (Rigden, 2008), and the proton donor motif is HAE, with glutamate replacing aspartate of the HD motif of histidine (acid) phosphatases. AC1-2 MINPP is more closely related to eukaryotic, plant and animal MINPP than it is to bacterial histidine (acid) phosphatases. Significantly, the only prior functional identification of a bacterial MINPP is that of the human gut commensals Bifidobacterium pseudocatenulatum and longum subsp. infantis (Haros et al., 2009;Tamayo-Ramos et al., 2012) and Bacteroides thetaiotaomicron (Stentz et al., 2014) that share the HAE motif. Other homologues can be found among the Actinobacteria, Betaproteobacteria and Gammaproteobacteria (Tamayo-Ramos et al., 2012;Stentz et al., 2014). Our identification of significant contribution of MINPP to aggregate environmental phytase activity and to Acinetobacter, particularly, serves to highlight novel biotechnological opportunity of exploitation of environmental samples. Acinetobacter spp. are commonly cited in context, but in no means as the principal agent, of enhanced biological phosphorus removal (Seviour et al., 2003). They harbour a polyphosphate kinase ppk that is induced by Pi starvation (Trelstad et al, 1999). It seems likely therefore that the function of MINPP may be related to Poly P accumulation in soil Acinetobacter.
The second isolate was identified as a Buttiauxella strain, and comparison with published genomes of similar strains revealed, in contrast, a single canonical histidine (acid) phosphatase. BLAST searches of Buttiauxella accessions for all phytase classes yielded only histidine (acid) phosphatase with E values less than 10 -68 . These were of the E. coli AppA family histidine acid phosphatase (Lim et al., 2000) with RHGVRAP and HDTN motifs, or bifunctional glucose 1-phosphatase/phytase Lee et al., 2003) class with RHNLRAP (similar to RANLRAP ) and HDSN (similar to HDQN ) motifs. The Buttiauxella sp. AppA and its engineered variants (Cervin et al., 2008) are already a commercial product used widely to improve pig and poultry performance (e.g. Adedokun et al., 2015). Other bacterial AppA enzymes, e.g., from E. coli and Citrobacter spp., are used similarly (Sommerfeld et al., 2018;da Silva et al., 2019). Our unbiased, for phytase class, screening approach is clearly capable of identifying candidate phytases with potential as commercial leads.

Media
Agar was obtained from Sigma (Merck Life Science UK Limited, Dorset, UK). Tryptone and yeast extract for preparation of Lysogeny broth were obtained from Formedium (UK).

Preparation of Soil Cultures for HPLC Analysis
Five hundred µl of a well-mixed soil culture in media was centrifuged at 13 000 g for 5 min. The supernatant was filtered through a 13 mm diameter 0.45 µm pore PTFE syringe filter (Kinesis) and centrifuged again, and an aliquot (200 µl) was dispensed into an HPLC vial.

HPLC Analysis of Inositol Phosphates
Inositol phosphates were analysed according to Whitfield et al. (2018). Chromatography data were exported as x,y data and redrawn in GraphPad Prism v.6.0 (GraphPad Software, San Diego, CA, USA).

Sequencing of strains
The Acinetobacter sp. strain AC1-2 genome was sequenced by MicrobesNG (University of Birmingham, UK) using Illumina technology. This Whole Genome Shotgun Project has been deposited at DDBJ/ENA/Gen-Bank under the accession JABFFO000000000. The version described in this paper is version JABFFO010000000. Genomic completeness was analysed using BUSCO v3 (Simao et al., 2015), an opensource software that provides quantitative measures for genomic completeness based on evolutionarily informed expectations of gene content from near-universal singlecopy orthologs selected. The Acinetobacter sp. strain AC1-2 completeness was measured at 98 and 98.9% from both BUSCO's bacterial and Gammaproteobacteria databases, respectively.

Conflict of interest
None declared.

Author contributions
GDR performed experiments, curated data and provided an original draft. JDT supervised experiments and edited the manuscript. ALN supervised experiments, curated data and wrote the manuscript. CAB secured funding, supervised experiments and wrote the manuscript.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Naturally occurring forms of inositol hexakisphosphate (phytate). Fig. S2. Purity of commercial phytate. Fig. S3. Schematic of workflow. Table S1. Canonical phytase complements of referenced Acinetobacter and Buttiauxella genomes summarized in Table 1.