Alexandrium strain ACHK-NT was kindly provided by the Collection Center of Marine Bacteria and Algae (CCMBA), Xiamen University, China. This strain was originally isolated from Hong Kong and identified as “A. catenella” morphologically. Based on the small subunit ribosomal RNA gene (SSU rDNA) isolated from this strain (unpublished) and phylogenetic analysis guided by information on the intragenomic SSU rDNA variation of “Alexandrium tamarense complex” (Miranda et al., 2012), this strain belongs to Clade IIC, which is exclusively composed of Asian “A. catenella”. The culture was grown in K medium (Keller and Guillard, 1985; Keller et al., 1987) with 0.22 μm-filtered and autoclaved seawater (28 PSU) at 20°C under a L:D = 14 h:10 h photocycle with a photon flux of 100 ± 10 μE s⁻¹ m⁻². One million cells were collected during exponential stage by centrifugation, resuspended in Trizol reagent (Invitrogen, Carlsbad, CA, USA) and stored at −80°C for subsequent RNA extraction.

The cells of three different dinoflagellate species (“A. catenella,” A. carterae, and K. brevis) under P limitation in the late stationary growth stage (DIP below detection limit of 0.2 μM in the medium) were collected by centrifugation and the cell pellets were then incubated in 1 mL 70% ethanol for 30 min. The samples were centrifuged and the supernatant was aspirated off. The pellets were incubated in 100 μL 0.22 μm-filtered sterile seawater with ELF^®-97 phosphatase substrate (Invitrogen, Carlsbad, CA, USA) in a final concentration of 0.25 mM for 30 min in the dark. The samples were washed three times using sterile seawater and stored in 100 μL sterile seawater in the dark (González-Gil et al., 1998; Ou et al., 2010). The stable green fluorescent precipitates produced in situ by the AP activity and red autofluorescence of the live cells produced by chloroplast were both observed under Nikon 90i epifluorescence microscope (green fluorescence excited by UV, 340–380 nm; Nikon Corporation, Tokyo, Japan).

Total RNA was extracted as previously reported (Lin et al., 2011) and stored at −80°C. The 1st strand cDNA was synthesized from ∼400 ng of the RNA using M-MLV reverse transcriptase (Promega, Madison, WI, USA) with a modified oligo-dT primer then purified using Zymo DNA Clean and Concentrator (Zymo Research, Orange, CA, USA).

Based on the nucleotide (nt) and amino acid (aa) sequence alignments (see Figure 1) of A. carterae and K. brevis AP genes (Lin et al., 2011, 2012), degenerate primers were designed to obtain AP gene fragments from “A. catenella” under the following PCR conditions: 94°C for 1 min followed by 5 cycles of 95°C for 15 s, 52°C for 30 s, and 72°C for 1 min, and 30 cycles of 95°C for 15 s, 56°C for 30 s, and 72°C for 1 min. Two PCRs were carried out with the 1st PCR performed using forward primer DinoAPNF4a (5′-TTTGGWATGGGWGCMGCRGARCA-3′) paired with reverse primer DinoAPNR6 (5′-CTTRTCRTCRCCTTCRTYMGCKGT-3′). The PCR product was diluted 100 or 1000 fold and used as template for the 2nd (nested) PCR, in which DinoAPNR6 was paired with DinoAPNF4b (5′-GGTATGGGWGCMGCRGARCAGATT-3′) as the primer set. The PCR product with expected size was purified by Zymo DNA Clean and Concentrator (Zymo Research, Orange, CA, USA) and directly sequenced (Invitrogen, Guangzhou, China).

Based on this partial sequence, specific primers were designed for both 5′- and 3′-rapid amplification of cDNA ends (RACE) to obtain the full-length cDNA by nested PCR. In the 5′-RACE, reverse primers AtaapR2 (5′-TGATGGACAAGACGACGATGG-3′, outside) and AtaapR3 (5′-AGTTCACCTTGCCGTTGCTTAC-3′, inside) coupled with forward primer DinoSL (Zhang et al., 2007b) were used in nested PCRs. In the 3′-RACE, forward primers AtaapF3 (5′-GACGACTCCAAGGTGTATGTGAAC-3′, outside) and AtaapF4 (5′-CTGGACATTGCCAGTGGCACGAT-3′, inside) were paired separately with modified oligo-dT primer in nested PCRs. The PCR products obtained from RACE were either directly sequenced, or cloned into a T-vector and several resultant plasmid clones were sequenced as reported (Lin et al., 2012). The sequences obtained were analyzed using Blastx against GenBank non-redundant protein sequences to confirm that the sequences are of the AP gene.

To recruit as many eukaryotic putative AP genes as possible for phylogenetic analysis, AP sequences of A. carterae and K. brevis were used as queries using basic local alignment search tool (program tblastn) against GenBank database. We also used the keyword search with “alkaline phosphatase” against GenBank database to recruit any reported sequences which were not hit by tblastn search. Furthermore, we searched the completed eukaryotic algal genome datasets at the Joint Genome Institute (JGI) databank using dinoflagellate putative AP sequences as the queries with E-value cutoff set at 1e-3, an e-value commonly used to search for protein homologs (e.g., Rensing et al., 2005; Ueno et al., 2010). In addition, to find AP sequences from the natural plankton assemblages, AP aa sequences of the two dinoflagellates were used as query to tblastn against the Bermuda Atlantic Time-Series Study (BATS) and Hawaii Ocean Time-Series (HOT) 454-sequencing reads as well as Sargasso Sea open reading frame datasets in the Community Cyber infrastructure for Advanced Marine Microbial Ecology Research and Analysis (CAMERA) databank. E-value cutoff was set at 1e-3, and 100 hits were set as the target for retrieval. A previous study showed that BATS water is DIP-poor whereas HOT water contains DIP concentrations which stoichiometrically fit the Redfield ratio (Wu et al., 2000).

To examine the relationship between “A. catenella” AP and counterparts from other organisms, deduced aa sequence of the “A. catenella” full-length cDNA was aligned with AP sequences from the representative organisms using ClustalX (Thompson et al., 1997). The environmental sequences retrieved from CAMERA were too short and therefore were not included in the analysis. Sequences of different forms of APs (phoA, phoD, phoX, phoV) from cyanobacteria, a group of organism whose APs have been relatively well studied, were collected and compared with putative dinoflagellate APs (Dino-APs) separately. The phylogenetic analyses were done only to sequences belonging to the same type (phoA) because no reliable alignment could be achieved for a mixed-type dataset. Alignment was performed for each of these datasets using ClustalW¹ with default setting (Gapopen 15, Gapext 6.66, Gapdist 8, Maxdiv 40). The alignments were then inspected and corrected manually. Neighbor-joining (NJ) analysis (Saitou and Nei, 1987) was run with 1000 bootstraps, and Bayesian analysis (Hueslenbeck and Ronquist, 2001) was run with model jumping setting for 725,000 generations (two independent runs were performed and the final standard deviation of split frequencies was 0.01), with trees sampled every 100 cycles and the first 25,000 cycles discarded (burn-in).

A set of programs were used to spatially characterize the deduced protein. The signal peptide was determined by TargetP V1.1 (Emanuelsson et al., 2000) and SignalP V4.0 (Petersen et al., 2011). GPI-SOM (Fankhauser and Mäser, 2005) and PredGPI (Pierlenoi et al., 2008) were used to locate the glycosylphosphatidylinositol (GPI) anchor in the deduced protein sequence. TMpred² was used to predict transmembrane domains.

Four existing bioinformatics models were adopted to predict subcellular localizations of AP in eukaryotic phytoplankton. Interpretable Subcellular Localization Prediction (YLoc; Briesemeister et al., 2010) predicts localization of a protein among nine possible subcompartments of the cell: extracellular space, plasma membrane, cytoplasm, mitochondrion, chloroplast, nucleus, Golgi apparatus, endoplasmic reticulum, and peroxisome. The YLoc+ algorithm in the package was used with plant selected as the target organism. The second program was Euk-mPLoc2.0 (Chou and Shen, 2010), which was designed specifically for predicting subcellular localization of eukaryotic proteins in 22 different compartments (Chou and Shen, 2010), including the nine in YLoc. The other two programs were WoLF PSORT (Horton et al., 2007) and CELLO (Yu et al., 2006). These programs have been used in various studies, such as functional annotation in C. reinhardtii and the subcellular localization prediction of the secretome for fungi (Soanes et al., 2007; Ghamsari et al., 2011; Rai et al., 2012). Because it has not been verified that these models suit APs or other proteins in eukaryotic phytoplankton. We first ran a set of proteins whose locations of functions are known to test the accuracy of the predictions. For AP, we tested a set of APs from the chlorophyte algae V. carteri and C. reinhardtii that are known to be related to phoX, an extracellular AP in bacteria and a homolog from E. huxleyi (EHAP1, GenBank accession number ABI51308) that has been shown to be extracellular (Xu et al., 2006). Included in the test was also AP from humans, whose cell surface distribution has been shown (Cutler et al., 1974). For dinoflagellates, a total of 98 protein sequences were tested, including representative proteins associated with cytoskeleton (actin), light harvesting in the chloroplast (peridinin-chlorophyll a-binding protein or PCP) and cell membrane (rhodopsin), carbon fixation in the chloroplast (ribolose-1, 5-bisphosphate carboxylase/oxygenase or Rubisco), and the cell cycle related protein proliferating cell nuclear antigen (PCNA), a well-characterized primarily nuclear protein. No model gave 100% correct prediction and no one failed 100% either, and in most cases predictions shared by two or more of these four models were correct (Table S1 in Supplementary Material). Therefore, we decided to use all four models so that the reliability of the predictions could be assessed, with the likelihood of the predicted AP localization ranked according to the number of the models that supported the prediction.

After assembling the initial partial AP sequence with the 5′ and 3′ RACE sequences, a full-length cDNA sequence was obtained from “A. catenella,” with an open reading frame of 2,181 bp (GenBank Accession No. JQ806383). Its deduced aa sequence (named as AlecaAP) was used to search against the GenBank protein database and the top two hits were A. carterae (with Expect value [E-value] to be 0.0) and K. brevis (E-value 2e-155), verifying that this is an putative AP cDNA. The AP sequences are moderately conserved among three dinoflagellates (48–52% identitical or 63–66% chemically similar at the aa level) and the overall characteristics are similar (Table 1, Figure 1A). AlecaAP cDNA has a ∼61% G+C content, highest among the three Dino-APs, predicted to encode a protein of 726-aa residues with a molecular mass of 76.6 kDa. Similar to other Dino-APs, AlecaAP also possesses a low pI (4.46), contributed by higher percentage of aspartic acid and glutamic acid residues (6.7 and 5.4%, respectively). A signal peptide was detected with a probability of 0.770 by TargetP 1.1 and confirmed by SignalP V4.0. The GPI anchor located at C-terminus (residues 721–726), which ensures AP to attach to cell membrane, was predicted by GPI-SOM and further confirmed by PredGPI (Table 1).

Alkaline phosphatase aa sequences are highly variable among the eukaryotic phytoplankton. No remarkable similarity was observed between the Dino-APs and other well-characterized eukaryotic APs such as EHAP1 from E. huxleyi (Landry et al., 2006; Xu et al., 2006, 2010) or phoX from V. carteri and C. reinhardtii (Hallmann, 1999; Moseley et al., 2006). Nevertheless, tblastn of Dino-APs against E. huxleyi CCMP1516 genome data (JGI) produced several hits with E-value lower than 1e-31. As predicted using TMpred, both the putative Dino-APs and other phytoplankton APs examined all contained 3–7 transmembrane domains. Interestingly, Dino-APs showed 15–30% of aa similarity with the APs from cyanobacteria, including a characterized atypical phoA gene (C-terminal half of the gene) from Synechococcus sp. PCC7942 (Ray et al., 1991; the phoA gene with GenBank accession number AAA27331). Further sequence comparison showed that V. carteri, C. reinhardtii also possessed phoA-like AP genes in addition to the phoX type mentioned above. The putative phoA-like AP sequences from P. tricornutum, T. pseudonana, V. carteri, C. reinhardtii, Aureococcus anophagefferens, cyanobacteria and bacteria share 10–16 conserved domains with Dino-APs (Figure 1B). The phylogenetic tree of phoA-like AP from different organisms showed that all of the eukaryotic phytoplankton APs clustered together, in sister relation to APs from marine bacteria (Figure 2).

No putative AP sequences were found in the Sargasso Sea ORF, likely due to its relatively small database size. Five significant hits were obtained from HOT and 15 from BATS sharing several conserved regions with Dino-APs. When these retrieved sequences were blasted against GenBank database using blastp, several of them matched bacterial AP as their top hit, giving two HOTS and five BATS sequences as eukaryotic AP as they best matched APs of dinoflagellates or diatoms (Table 2). Given that the metagenomic data were obtained from microbial plankton, the AP genes detected were probably restricted to pico- or nano-plankton. This might explain why so few sequences of AP genes were found and partially why none were found from the Sargasso Sea ORF dataset.

The test runs using protein sequences from various organisms showed that most of the proteins were predicted correctly when the prediction was supported by two or more of the four models used. APs in humans were predicted to be extracellular by YLoc+, WoLF PSORT, and CELLO and cell membrane by Euk-mPLoc, consistent to observed extracellular space distribution of this protein (Cutler et al., 1974). Like phoX in bacteria, the phoX-like APs in chlorophyte C. reinhardtii and V. carteri have been shown to be extracellular (Hallmann, 1999; Moseley et al., 2006); both of them were predicted to be extracellular by two of the models (Table S1 in Supplementary Material). Among dinoflagellate proteins, actin was uniformly predicted to be cytoskeleton by Euk-mPLoc, and to be cytoplasmic by YLoc+ and CELLO, the two models that do not have the cytoskeleton category. PCP was correctly and uniformly predicted to be chloroplastic by all four models. Rhodopsin, a known transmembrane protein, was predicted as a plasma membrane protein in all the 19 cases by Euk-mPLoc and CELLO and 11 of these are supported by three or more of the models. All 16 Rubisco sequences, from various species of dinoflagellates, were predicted as chloroplast proteins by three or more models. All 16 PCNA genes examined were predicted to be nuclear proteins, and 5 of these were supported by three or more models. All these results confirmed that the joint-multiple prediction models could be applied to dinoflagellates and other phytoplankton proteins by the majority rule (≥2 models). Predicting results based on this “rule” are shown in Table 3. Putative AP in “A. catenella” would be localized in the extracellular space (2 models) and chloroplast (2 models), with equal probability. Putative AP in A. carterae would most likely be in the extracellular space (plus endoplasmic reticulum, which can be considered as transporting and secreting out of the cell; 3 models) but cytoplasm (1 model) and chloroplast (1 model) are possible as well. In K. brevis, putative AP was predicted to be most likely extracellular (also plus Endoplasmic Reticulum; 4 models) or cell surface (cell wall plus membrane; 2 models) with a small probability to be found in other intracellular compartments (1 model each).

These computational predictions of putative Dino-AP localization were verified by enzyme labeled fluorescence (ELF) labeling. In the three dinoflagellate species we have analyzed so far, AP activity as indicated by ELF showed different localization patterns (Figure 3). In “A. catenella,” AP appeared to be located both intracellularly and on the cell surface. In A. carterae, AP activity was more on the cell surface than intracellular but both locations had substantial activities. In K. brevis, AP activity was mostly distributed on the cell surface. The ELF labeling patterns were consistent with model predictions.

Among other groups of eukaryotic phytoplankton putative AP was predicted to be in the extracellular space and plasma membrane of P. tricornutum and most likely in the cytoplasm of T. pseudonana and A. anophagefferens. In E. huxleyi, the phoA-like AP (jgi|Emihu1|444279|estExtDG_fgenesh_newKGs_kg.C_3060016:21-652) was more likely extracellular but could be cytoplasmic as well whereas EHAP1 (ABI51308) was predicted to be extracellular and cell surface (plasma membrane; Table 3). EHAP1 has been experimentally shown to be extracellular (Xu et al., 2006).

With different forms and unique characteristics found in this study and documented in literature, AP is clearly a highly variable gene in marine microorganisms. Three AP families, phoA, phoX, and phoD have been found in marine bacteria, and their gene sequences, subcellular localizations, metal cofactor requirements, and substrates have been analyzed experimentally and computationally (Luo et al., 2009, 2011; Sebastian and Ammerman, 2009). Similar cases (but more complicated) were also observed in cyanobacteria, which are one of the most common phytoplankton lineages. So far, the only well-characterized AP genes in marine cyanobacterium were phoA (145 kDa) and phoV (61.3 kDa) from Synechococcus strain PCC7942 (Ray et al., 1991; Wagner et al., 1995). The previously identified atypical phoA, with an unusually large molecular weight, showed no sequence homology with other AP genes and was believed to exhibit extracellular phosphatase activity without a cleavable signal sequence (Ray et al., 1991). The phoD type of AP was characterized in halotolerant cyanobacterium Aphanothece halophytica (Kageyama et al., 2011). The heterogeneous sequences of AP made it very difficult to identify the AP gene homology in different strains of Synechococcus and Prochlorococcus (Moore et al., 2005). In addition, a fourth type, PhoV, was identified as a membrane-associated protein localized in the periplasmic space requiring Zn²⁺ for activity (as required by phoA in bacteria); the sequence of this gene was 34% identical to phoA in the α-proteobacterium Zymomonas mobilis (Wagner et al., 1995). A recent study on the genome sequence of Trichodesmium spp. confirmed that there are two putative APs, one sharing a conserved domain with the characterized phoX from marine γ-proteobacteria, and the other sharing 52% identity to 50% of the translated phoA gene from Synechococcus PCC7942 (Orchard et al., 2009).

Among eukaryotic microalgae, only APs from chlorophyte C. reinhardtii and V. carteri have been categorized with respect the types recognized in bacteria; phoX like sequences occur in both species (Hallmann, 1999; Moseley et al., 2006). The transcriptomic and proteomic study indicated that there are several putative APs in T. pseudonana (Dyhrman et al., 2012). Before the analysis of the AP genes from dinoflagellates reported here, the only characterized AP coding gene in marine eukaryotic phytoplankton is ehap1, identified in E. huxleyi (Xu et al., 2006). The gene product EHAP1, with its signal peptide identified at the N-terminus and without a GPI anchor at the C-terminus, can be released from the cell surface into the medium. But it does not show any similarity to the known types of APs. Because of the very limited knowledge of eukaryotic phytoplankton AP genes, it is difficult to categorize the putative APs from dinoflagellates and other eukaryotic phytoplankton into those recognized types of marine APs. However, our sequence comparison results showed that Dino-APs and other putative AP sequences in other eukaryotic phytoplankton genomes retrieved from the public databases share highest (despite still very low) identifiable sequence similarity with the phoA AP from bacteria and cyanobacteria among all the recognized types of AP. These AP amino acid sequences clustered together in the phylogenetic tree (Figure 2). Regardless if these are indeed phoA type APs, all our analyses suggest that they are undoubtedly a subset of putative APs.

It is interesting to observe that while cyanobacteria, chlorophytes C. reinhardtii and V. carteri, and the haptophyte E. huxleyi have multiple types of AP, essentially only one form of putative AP sequence has been identified in dinoflagellates. There is little sequence variation in the coding region, most of which involves synonymous nt substitutions (Lin et al., 2011, 2012). In the 2,181-bp long AP cDNA coding region in “A. catenella,” only two sites show changes. This is striking because most of the genes in dinoflagellates that have been studied exist in numerous variations. For instance, pcna, which usually occurs in one copy in other eukaryotes, exists in tens of copies in the dinoflagellate Pfiesteria piscicida, which contain many synonymous and substantial numbers of non-synonymous nucleotide substitutions (Zhang et al., 2006). Similarly, Rubisco genes in P. minimum occurs in over 100 copies, which differ from each other by as much as 5%, with both synonymous and non-synonymous changes as well as insertions/deletions (Zhang and Lin, 2003). If the duplication of genes in dinoflagellates has resulted from whole genome duplication, as recently postulated (Hou and Lin, 2009), the negligible level of variation in AP relative to that of other genes may suggest that Dino-APs probably face strong evolutionary purifying selection (Kim et al., 2011). However, if dinoflagellate genes duplicated independently, then the low variability of the gene can simply indicate that the gene copies have only arisen from recent duplication events (Thornton, 2007 and the refs there in).

The sequence variability of putative AP between lineages is strikingly high. Although it has been shown that 25–30% of the aa residues, especially the active sites, in APs are well conserved among mammals, yeast, and E. coli (Murphy et al., 1995), this does not seem to be the case for dinoflagellates and other eukaryotic phytoplankton. Putative Dino-APs did not share obvious similarity with APs from other eukaryotes. There were also no obvious conserved domains observed between putative Dino-APs and phoV from S. elongatus PCC 7942. When comparing putative Dino-APs with phoA from Trichodesmium spp. (Orchard et al., 2009) and Synechococcus PCC7942 (Ray et al., 1991), more than 10 conserved domains were identified between putative Dino-APs and phoA; however, no conserved region was found between putative Dino-APs and phoX. The lack of long conserved sequence regions and the overall high variability rendered the alignment of nt sequences of counterparts from different lineages (e.g., dinoflagellates and diatoms) impossible; however, our careful alignment of aa sequences did reveal several conserved domains (Figure 1). The putative Dino-APs shared no similarity with EHAP1 from E. huxelyi (Xu et al., 2006, 2010), but hit several unidentified E. huxelyi genome contigs with significant E-values (e.g., jgi 444279, 3e−157 to 0.0). A recently identified protein with AP activity in A. lagunensis also shared similarity with the AP from proteobacteria Idiomarina baltica OS145 (Sun et al., 2012). In the phylogenetic tree, putative Dino-APs formed a monophyletic group with other eukaryotes whose chloroplasts belong to the red-type (i.e., chromalveolate), which as a whole is sister to APs from some bacteria. This is likely due to the existence of a number of chemically similar aa residues among the chromalveolate APs, which increased the similarities between the sequences. Whether the somewhat closer phylogenetic relationship between the red-type algae and bacteria has resulted from a horizontal gene transfer cannot be determined with the limited data currently available, but is worth exploring in the future.

The computational study of subcellular localization of marine bacterial APs implies different strategies of DOP utilization by different forms of AP (Luo et al., 2009, 2011). The high abundance of cytoplasmic APs (phoA and phoD) suggests that intracellular hydrolysis of transported small DOP molecules is also an important P acquisition mechanism (Luo et al., 2009). Presence of phoX (extracellular) and its relatively wide distribution is thought to provide a growth advantage to the bacteria (may be cyanobacteria too) because it has broad substrate specificity as well as using Ca²⁺ as co-factor instead of Zn²⁺, which is often limited in the oligotrophic ocean (Orchard etal.,2009; Sebastian and Ammerman, 2009). According to previous studies, APs are located on the cell surface in most of the species examined (e.g., Chaetoceros affinis, Skeletonema costatum, P. tricornutum, Chlorella sp., Isochrysis galbana, P. Donghaiense, and Peridiniumcinctum) whereas are intracellular in some other species (e.g., P. mican, “A. fundyense,” and “A. catenella”; Ou et al., 2010). Our ELF labeling patterns were suggestive of AP being localized both intracellularly and on the cell surface of “A. catenella,” which is consistent with the computational prediction of chloroplast and cell surface localization. Chloroplast in “A. catenella” was highly branched and abundant in the cytoplasm (Hansen and Moestrip, 1998), rendering chloroplast AP ELF labeling as scattered in the cytoplasm (Figure 3). Comparative studies have shown that the diatom S. costatum quickly released the AP into the medium while “A. catenella” and P. donghaiense harbor most of the AP inside the cells and on the cell surface, respectively, so that these algae could use DOP in the ambient water and survive longer under P-limited condition (Ou et al., 2008, 2010).

The computational prediction of some putative eukaryotic phytoplankton APs also showed that AP localization (i.e., cell surface or intracellular) varied with species. AP localized in the extracellular space may enable that species to utilize DOP from the external environment as cleaved phosphate. This would obviate the need to import bulky DOP into the cell. In contrast, APs located inside the cell would only be able to utilize DOP found inside the cell, either transported from the external environment or produced inside the cell (e.g., degraded or denatured proteins and other organic matter). As suggested in marine bacteria, cytoplasmic APs probably function in internal organophosphate hydrolysis (Luo et al., 2009). The differentiation of subcellular localization thus would confer adaptive strategies in utilizing different sources of DOP. This differentiation may be the reason that different phytoplankton species respond to DIP limitation in the environment in different manner. For instance, A. carterae responded to DIP limitation by expressing AP activities detectable on live cells (i.e., likely extracellular or cell surface) very quickly and is able to maintain growth rates comparable to that under DIP-replete conditions (Lin et al., 2011). In contrast, K. brevis growth rate declined shortly after DIP was depleted and detectable AP activity increased long after (Lin et al., 2012). Perhaps the more cytoplasmic presence of AP enables A. carterae to access DOP generated inside the cell quickly compared to K. brevis, in which the dominant extracellular/cell surface distribution of AP may force the cells to rely on external DOP generated from cell death and on the import process for the cleaved DIP to enter the cell. Yet, such linkage between AP sequence, subcellular localization, and P nutritional strategy requires much more dedicated studies to verify.