Proteomic Characterization of Cellular and Molecular Processes that Enable the Nanoarchaeum equitans-Ignicoccus hospitalis Relationship

Nanoarchaeum equitans, the only cultured representative of the Nanoarchaeota, is dependent on direct physical contact with its host, the hyperthermophile Ignicoccus hospitalis. The molecular mechanisms that enable this relationship are unknown. Using whole-cell proteomics, differences in the relative abundance of >75% of predicted protein-coding genes from both Archaea were measured to identify the specific response of I. hospitalis to the presence of N. equitans on its surface. A purified N. equitans sample was also analyzed for evidence of interspecies protein transfer. The depth of cellular proteome coverage achieved here is amongst the highest reported for any organism. Based on changes in the proteome under the specific conditions of this study, I. hospitalis reacts to N. equitans by curtailing genetic information processing (replication, transcription) in lieu of intensifying its energetic, protein processing and cellular membrane functions. We found no evidence of significant Ignicoccus biosynthetic enzymes being transported to N. equitans. These results suggest that, under laboratory conditions, N. equitans diverts some of its host's metabolism and cell cycle control to compensate for its own metabolic shortcomings, thus appearing to be entirely dependent on small, transferable metabolites and energetic precursors from I. hospitalis.


Introduction
The hyperthermophiles Nanoarchaeum equitans and Ignicoccus hospitalis engage in the only physical and specific cell-cell interaction documented thus far for two species of Archaea [1]. With a highly reduced genome and nearly absent biosynthetic and energetic functions, N. equitans requires physical contact with I. hospitalis [1,2]. This enables uptake of small molecules (amino acids, lipids) from the host by some yet unknown mechanisms [2,3]. Categorizing the interaction between the two organisms to a specific type of traditionally defined symbiotic relationship has been difficult due to the unclear impact that N. equitans has on its host which, in laboratory cultures, ranges from neutral to inhibitory [2]. Evidence of horizontal gene transfer between the two organisms and indication of genome streamlining in I. hospitalis nevertheless suggests that, at the genome level, the interaction has impacted not only N. equitans but its host as well [4]. As it represents the simplest ''symbiotic'' system known thus far, functional analysis of the roughly 1800 genes combined between the two organisms provides unique insight into specific genetic and physiological mechanisms of their relationship and can perhaps address more general principles of microbial interspecies interaction and co-evolution.
Among the approaches that enable global monitoring of genome function, whole-cell shotgun proteomics provides direct protein measurements and thus physical evidence for the expression of any individual protein-encoding gene from the genome [5]. When relative protein abundance levels are determined, quantitative variations of individual proteins can be linked to physiological activity or other measurable phenotypic states of the cell. Previous studies, which focused on identifying major proteins expressed by I. hospitalis grown as a pure culture [6] or those found in membrane fractions prepared from co-culture with N. equitans [7] detected a limited subset of the predicted proteome of the two organisms. For other archaeal systems, differential proteomics has been successfully applied to identify genes/proteins that are involved in adaptation and physiological response to various environmental challenges such as low temperature in Methanococcoides burtonii [8] or energy and nutrient starvation in Methanococcus maripaludis [9] and Halobacterium salinarum [10]. In this study, we applied an on-line, twodimensional liquid chromatography tandem mass spectrometry (2D-LC-MS/MS) proteomic approach, utilizing the high resolving power and accuracy of a hybrid LTQ/Orbitrap mass spectrometer to maximize the number of detected proteins in several conditional whole-cell lysates. Normalized spectral abundance factors were employed to quantitatively compare the proteomic differences between I. hospitalis grown by itself or when engaged with N. equitans. Proteins that exhibit changes in their abundance levels per condition were identified, leading to a better understanding of the impact that N. equitans has on its host, both in terms of changes in gene expression and/or protein abundance thus providing physiological insight into the nature of their relationship.

Proteome coverage overview
Deep whole-cell proteomic coverage for I. hospitalis and N. equitans was achieved using Multidimensional Protein Information Technology (MudPIT) 2D-LC-MS/MS [11]. More importantly, the relative abundance of detected proteins was determined and compared between both I. hospitalis grown independently and in conjunction with N. equitans. By comparing I. hospitalis protein abundance between states, potential genes, proteins or cellular processes that respond to and may be involved in the association of these two archaea could be identified.
To compare the protein abundance profiles of both sets of samples (I. hospitalis and I. hospitalis-N. equitans co-culture), separate large-scale fermentations were performed. As previously shown, the number of N. equitans cells that are associated with any given host cell increases during the course of cultivation, from an average of less than one in early growth stages to over ten towards stationary phase [2]. While it is understood that the potential physiological differences associated with co-culture progression would be better defined by replicate time-course samples, this study focused more on developing a whole cell proteomic approach suitable for studying these organisms and thus used a single, stationary culture time point as the basis for comparison. In an effort to be comprehensive, an N. equitans sample was purified from a co-culture and analyzed by 2D LC-MS/MS, though quantitative analysis was not performed as this sample was derived from a co-culture with I. hospitalis and is thus not biologically relevant with regard to differential proteomics.
Three replicate measurements were obtained for each of the samples totaling nine, independent 2D LC-MS/MS runs. Average false-discovery rates were ascertained to be 2.19%, 2.76%, and 4.54% for the co-culture, I. hospitalis pure culture, and purified N. equitans samples respectively. Global pair-wise reproducibility between technical replicates at individual protein level ranged from 0.90-0.99 with average correlation values of 95.4% and 94.7% for I. hospitalis proteins or 93.4% and 92.2% for N. equitans proteins, in the co-culture or isolates respectively. Correlation values between the pure and co-cultures calculated independently for each organism, i.e. I. hospitalis proteins in the single vs. coculture (86.3%) or N. equitans proteins in the purified sample vs. coculture (93.2%) indicate the extent to which each organism's proteome varies as a function of culture condition. As follows, I. hospitalis' proteome is more dynamic than that of N. equitans and shows a marked correlational difference between technical replicates and culture condition. In contrast, the variability between N. equitans technical replicates is very similar to that observed between samples, suggesting little difference between the proteomes of each state. As a ''pure'' N. equitans sample can only be derived from co-culture with I. hospitalis, the proteomes were expected to be similar between conditions. Thus, the ensuing differential proteomic analysis focused on I. hospitalis samples.
Further details about the data analysis strategy are provided online (Supporting information S1 and Figure S1).
Overall, a total of 1058 I. hospitalis proteins were identified out of 1444 open reading frames (ORFs), representing over 73% of the predicted proteome (Table S1). This remarkable level of proteome coverage suggests that a larger proportion of the proteome is constitutively expressed and supports the hypothesis that the I. hospitalis genome is streamlined, containing few redundant or nonfunctional genes [4]. The total number of SEQUEST-assigned spectra was similar between all sample conditions: 139,556 for I. hospitalis when grown as a pure culture, 145,894 assigned to the co-culture (68.5% to I. hospitalis and 31.5% to N. equitans), and 133,418 assigned to the purified N. equitans sample (90.0% to N. equitans and 10.0% to I. hospitalis). Across both N. equitans datasets (purified plus co-culture), we detected a total of 476 proteins out of the predicted 556 (85%). This is amongst the highest proteome coverage ever reported for an organism [12].
Individual, pre-normalized protein spectra for each organism covered over three orders of magnitude per MS run, from 2 spectra (our conservatively defined minimum) to over 8000. At the whole proteome level, no correlation was observed between the size of a protein and the number of spectra obtained for that protein, especially after NSAF-based normalization, which effectively corrects for any protein length-derived abundance bias ( Figure 1). When analyzing proteins containing predicted transmembrane domains (TMD), however, we observed a 2-3 fold lower average of the number of spectra per protein as compared to non-membrane proteins; these SpC values were independent of size as well ( Figure 1). The general protocol for whole-cell shotgun proteomics is known to be superior for soluble proteins relative to membrane proteins, and thus translated into a lower fraction of identified transmembrane proteins as compared to total proteins (48.2% for both Ignicoccus and Nanoarchaeum combined compared to 82.5% of cytosolic proteins).
Overview of detected I. hospitalis and N. equitans protein families Based on comparative genomic analyses, archaeal proteins have been assigned to orthologous groups (archaeal COGs, arCOGs) [13]. They reflect diversification of gene families and allow functional inferences to be made for individual organisms based on gene acquisition (duplication, horizontal transfer) or loss. In both N. equitans and I. hospitalis, arCOG-based analyses have revealed a low degree of functional redundancy (paralogs), gene loss and a high fraction of hypothetical genes with no homologs in other genomes (300 genes, ,20% of the genome in I. hospitalis, and 169 genes, 30% of the genome in N. equitans) [4]. Though denoted as ''hypothetical'' as gene calls, approximately half of their encoded proteins were detected in the analyzed samples, indicating that they are valid and expressed genes that should be reclassified as ''proteins with unknown function''. Among the arCOG categories, as expected, a large fraction (70-90%) of the proteins expressed in both organisms is involved in cell cycle control and genetic information processing (DNA replication and repair, transcription, translation and protein processing ( Figure 2). Notably, we detected over 90% of the I. hospitalis proteins predicted to be involved in energy generation, amino acid and nucleotide metabolism and transport, again supporting the notion that this Archaeon has a streamlined, non-redundant genome in which most metabolic genes are constitutively expressed, even in laboratory setting. Functional categories that were covered to a lesser degree include genes with no assigned orthology to other genomes (40-50% of the predicted genes in that group) and/or cellular functions that are represented by a relatively low number of genes which may or may not respond to environmental variables not reproduced in the laboratory (inorganic ion transport, defense, motility/pili). A large number of genes in those categories also encode membrane proteins, which are in general less effectively detected by trypsin-based bottom-up proteomic methodologies and are therefore under-represented with regards to protein identification and/or overall sequence coverage. Even in those categories, however, several functions were clearly active, with corresponding proteins detected at significant levels (RSpC values 100-500). For example, the protein encoded by Iho_0670, which has been shown to form a unique type of cell surface appendage/fiber [14], was present in appreciable amounts in both single culture and co-culture with N. equitans.
In N. equitans, arCOG categories that encode nearly all primary metabolic functions are severely underrepresented, containing ten or fewer genes per category [15] (Figure 2b). Although this small subset of enzymes cannot support independent metabolism, detection of a large portion of these proteins (between 70-100%) indicates that N. equitans has not entirely lost its physiological capabilities. Even though overall the contribution of the metabolic COG classes to the total N. equitans proteome is modest, when normalized to the actual number of genes they include the normalized abundance of those expressed proteins is significant (Figure 2c,d). Interestingly, several enzymes present at significant levels are likely involved in amino acid and nucleotide metabolism (deaminases, dehydrogenases, kinases and hydrolases) and may perform a limited but important range of molecular inter-conversions and/or metabolic transformations of precursors imported from I. hospitalis, or perhaps recycling of its own metabolic products. An abundant inorganic pyrophosphatase (Neq461) prevents accumulation of pyrophosphate from DNA and RNA synthesis. The ATP synthase subunit A (Neq103) and the very short subunit B (Neq263) are expressed at appreciable and relatively equivalent levels. Likewise, all the other predicted subunits were detected except for the transmembrane oligomeric subunit c (NEQ217). The ATP-biosynthetic capability of the complex has not yet been experimentally confirmed as it lacks several subunits [16]; therefore, the energetic independence of N. equitans remains questionable. Aside The average adjusted spectral counts for proteins containing TMDs or not were calculated for each of the four biological samples. There is no correlation between protein size and the number of detected spectra, however, membrane proteins generated on average approximately three times less spectra than non-TMD proteins. doi:10.1371/journal.pone.0022942.g001 from ATP synthase subunits, several redox proteins were also detected, including a predicted ferredoxin (Neq373), flavodoxin reductase (Neq051) and desulfoferredoxin (Neq011). The genetic information processing categories were covered extensively, with near 100% of the predicted proteins being detected. The full list of I. hospitalis and N. equitans expressed genes classified by functional arCOG categories and their relative protein abundance is presented in the Table S1.
The I. hospitalis proteome Since spectral abundance is a relative quantification measure, proteins for which a large number of spectra were detected indicate major cellular constituents. By identifying such highly abundant proteins, important inferences about cellular processes and/or structures may be drawn. However, as mentioned above, quantification of membrane and other smaller proteins is likely underestimated by virtue of the relative lack of tryptic cleavage sites leading to MS-incompatible peptides. Therefore, the proteome map of I. hospitalis-N. equitans is still somewhat incomplete, despite that over 75% of the joint, predicted proteome Table 1. The most abundant I. hospitalis proteins, based on the sum of RSpC values of three independent measurements on the pure culture (Igni) and the co-culture (Igni_Nano) samples as well as the ratio between the abundance in the pure culture versus the co-culture (I/I_N). of the two constituent Archaea was identified with relatively conservative criteria. This value ranks with some of the most thoroughly identified proteomes to date, most notably representatives of the genus Mycoplasma [12]. Inferences into cellular function are perhaps more easily understood by observing abundance trends in proteins that belong to similar or linked functional categories. One of the protein categories present in high abundance in I. hospitalis includes chaperones and proteins involved in the oxidative stress response ( Table 1). These include peroxiredoxin (Iho0459), an Hsp20 family protein (Iho_1363), and the thermosome subunits (Iho_0096 and 0897). The thermosome is a type of chaperone involved in protein folding and molecular adaptation to high temperature and has been previously identified as a major cytosolic protein in I. hospitalis [6]. An adjacent gene to thermosome subunit Iho0096 encodes an abundant small protein (Iho0095; 108 aa) that belongs to a ferritin-like domain family that includes a rubrerythrin domain. Rubrerythrin is an abundant protein present in other thermophiles that is also part of the constitutive oxidative stress response mechanism. Specifically, it has been shown to protect against peroxides in Pyrococcus [17]. Another abundant protein that could be categorized similarly is the universal stress family protein UspA (Iho0144), which, interestingly, was down-regulated in the co-culture with N. equitans. The gene is part of an operon that encodes two hypothetical proteins (Iho0142 and 0143) with no known homologs, which were also expressed at similar levels to UspA. Two FAD-dependent pyridine nucleotide-disulphide oxidoreductases (Iho0673 and 0899) were expressed at high levels as well. One of the two (Iho0673) was significantly up-regulated in the co-culture (,5fold). This family of enzymes may also be involved in protection against oxidative stress and regeneration of oxidized pyridine nucleotides [18]. Iho0673 belongs to a predicted three-gene operon that includes a conserved unknown protein (Iho0672) that is up-regulated ten-fold in co-culture and a sulfur oxidoreductase that is present at relatively unchanged levels. The significance of these genes' up-regulation in the presence of N. equitans is unclear but may indicate an increased level of cellular stress due to the association. Another protein potentially involved in cellular stress, an AAA-ATPase (Iho1431) is also present at high levels and upregulated (2-fold) in co-culture as well. The other oxidoreductase, Iho0899 has a signal sequence, suggesting secretion into the periplasm and is one of the previously identified candidates of horizontal gene transfer with N. equitans [4]. Based on RSpC, it appears to be the most abundant Ignicoccus protein measured in the co-culture with N. equitans (RSpC 2387). Its close homolog in N. equitans, Neq024 is also a highly abundant protein. While the presence of high levels of stress response proteins and their apparent up-regulation in co-culture may indicate that I. hospitalis reacts negatively to the presence of N. equitans, further studies in different culture stages will be necessary to test this.
Proteins and molecular complexes responsible for energy generation were also detected at significant levels. All the archaeal ATP synthase subunits were identified, except for the transmembrane subunit c (Iho0682), which forms the oligomeric ring. That protein is relatively small (113 aa) and upon sequence analysis we determined its predicted tryptic peptides have sizes outside the detection criteria, perhaps explaining its absence. The A and B subunits (Iho1305 and 0679), on the other hand, were found at high, equivalent levels (average RSpC 957 and 981), reflecting their equimolar multimeric involvement in the ATP synthase architecture. The other subunits were also detected at moderate to high abundance except for subunit a (Iho0609), which is an integral membrane protein and may have been inefficiently recovered during initial protein extraction. Subunit F (Iho1081), while very abundant (average RSpC 1836), appears down regulated two-fold in the presence of N. equitans. The I. hospitalis ATP synthase has recently been shown to be located on the outer cell membrane [19]. Another energy-related membrane complex detected at high levels, Ni-hydrogenase, was up-regulated two-fold in the presence of N. equitans. The four proteins predicted to form the complex (small and large hydrogenase subunits, a 4Fe-4S ferredoxin, and a transmembrane membrane protein) are encoded by genes that appear to be part of a single transcriptional unit (Iho1366-1369); all constituents were detected though the transmembrane protein was identified at very low levels, probably due the aforementioned tryptic incompatibility. Unlike ATP synthase and Ni-hydrogenase, the sulfur reductase appears to be less abundant. This complex has been previously shown by antibody staining to be located on the outer membrane like ATP synthase [19]. Though its exact architecture remains to be characterized, there appears to be two operons encoding potential protein subunits, Iho0801-0803 and Iho0528-0530, the latter possibly of bacterial origin [4]. In the I. hospitalis sample, the proteins encoded by Iho0801-0803 were present at low to moderate levels (RSpC 2-60) while their abundance was even lower in the co-culture, suggesting a more than five-fold downregulation by the presence of N. equitans. The proteins encoded by Iho0528-0530 were slightly more abundant (RSpC 13-174) and their level increased approximately two-fold in the presence of N. equitans. Another membrane complex that so far has not been associated with any cellular function is the outer membrane pore, assembled by the oligomerization of Iho1266 [20]. This abundant protein constituent of the cell [6] appears to be slightly downregulated (1.5 fold) in co-culture. While its involvement in the transfer of metabolites between I. hospitalis and N. equitans is uncertain, clearly the presence of N. equitans does not induce a significant increase in the number of pores on the host outer membrane.
Among the proteins involved in genetic information processing, the large subunit of ribosomal protein L7AE (Iho0230) was the most abundant protein identified in the I. hospitalis pure culture (RSpC 4160) but was down regulated two-fold in the presence of N. equitans. L7AE is an important, multifunctional RNA-binding protein that recognizes the K-turn motif in ribosomal box H/ACA and box C/D sRNAs, and is thus involved in RNA modification pathways. As will be discussed further, other proteins involved in RNA synthesis and control are down-regulated as well in the coculture. Additional proteins of high abundance involved in translation and transcription include the elongation factors 1A and 2 (Iho1150 and Iho1383) as well as an AsnC family transcription factor (Iho0226), all of which are present at similar levels in both pure and co-culture. Among the abundant proteins involved in chromosomal structure and cell division are the nucleoid protein Alba (Iho0174) and a putative membrane associated protein involved in DNA segregation, annotated as a multiheme protein (Iho1359); both proteins are down-regulated in the co-culture. A protein that is up-regulated three-fold in the presence of N. equitans is Iho0929, a member of the roadblock/ LC7 superfamily. The function of these proteins in prokaryotes is not known but it has been suggested to be involved in regulation of GTPase activity in the cell [21].
Abundant metabolic enzymes include acetate-CoA ligase (Iho0256, RSpC 574), which likely provide an additional source of acetyl-CoA using acetate. In the original genome, two adjacent genes (Iho0256 and 0257) encoded separate regions of the enzyme. We have re-sequenced that region of the genome and determined that in fact a sequencing error was responsible for the initial split of that enzyme in two separate ORFs. The genomic sequence deposited in GenBank and the gene annotation is being corrected. A recent independent analysis of that gene also reached the same conclusion and also conformed that Iho0256 can use acetate as a substrate [22]. Other abundant enzymes important to central metabolism were identified including phosphoglycerate kinase (Iho0274), fructose 1,6-bisphosphatase (Iho0363), 4-hydroxybutyryl-CoA dehydratase (Iho0595), enzymes involved in the biosynthesis of thiamin and riboflavin (Iho0560 and Iho936) as well as in amino acid metabolism (the two subunits of methionine synthase, Iho0747 and 0748). Phosphoenolpyruvate synthase (Iho1113), a key enzyme in carbon fixation, is up-regulated two fold in co-culture, potentially reflecting the increased metabolic demand imposed by the association with N. equitans. Similarly, the pyruvate ferredoxin oxidoreductase complex, which fixes CO 2 to pyruvate and is encoded by a four-gene operon (Iho1256-1259), is also up-regulated close to two-fold in co-culture. Phosphoenolpyruvate carboxylase (Iho0341) also follows this trend. Though perhaps indicative of increased metabolic and energetic demand put upon I. hospitalis by its association with N. equitans, it remains to be seen whether modest two-fold variations in the amount of individual proteins or complexes involved in critical steps of carbon fixation and energy generation are biologically significant for the relationship between the two organisms.
The N. equitans proteome As described above, differential proteomics based on the proteome derived from co-culture purified N. equitans may not be as biologically informative as compared to I. hospitalis. Nevertheless, the measured proteome provides a glimpse into the metabolic processes undertaken by this obligate ectoparasite. Structural proteins, regulators and enzymes involved in packaging and processing of the genetic material were found in high abundance both in the purified N. equitans sample and in the co-culture. These include the archaeal histone Neq348 (for which in the purified N. equitans, we observed the highest number of rebalanced normalized spectral counts, RSpC 9540) and the nucleoid protein Alba (Neq363), components of the replication, recombination and repair machineries (Neq537, 426, and possibly 368), as well as the cell division proteins FtsZ (Neq133) and MinD (Neq119). Several ribosomal proteins, translation elongation factor 1A, as well as two transcription factors (Neq098 and 534) were also among the top 30 most abundant proteins ( Table 2).
Among the most abundant enzymes identified is an inorganic phosphatase (Neq461; RSpC,2000), likely important for preventing a buildup of intracellular pyrophosphate resulting from nucleic acid biosynthesis, aminoacyl-tRNA synthetases (e.g. Neq535), glutamate dehydrogenase (Neq077) and nucleoside diphosphate kinase (Neq307), which can regenerate the NTP pool using ATP. These proteins appear to enable important metabolic transformations that cannot be provided by the host and led to the maintenance of endogenous N. equitans genes even though the majority of other biosynthetic functions were lost.
Protein processing and turnover are highly active as well, as evidenced by the abundance of chaperones (Neq141 and 344), the proteasome (Neq203), and several other peptidases. On the cell surface, the S-layer protein (Neq300) and its associated, smaller companion Neq236 dominate (RSpC.7000 and .1000, respectively). Several of these proteins have previously been identified in isolated membrane fractions that contained contact point between the two organisms [7].
Though the above mentioned abundant proteins are physiologically important to N. equitans, it is more likely that proteins involved in host recognition, cell surface interaction, transfer of metabolites, and/or other types of membrane proteins would provide more pertinent biological insights with regard to the association. As follows, many of the expressed N. equitans proteins, including predicted membrane proteins, have no recognizable homologs. Several of those identified at high levels include Neq035, 099, 222 and 492. Proteins that cannot be assigned a clear function in N. equitans but have been classified to existing microbial families include those involved in signaling and potential cell-cell interaction. Several members were detected at high or moderate levels, including a RecA-superfamily ATPase, implicated in signal transduction and containing the KaiC domain (Neq174, RSpC.3000), and two Flp pilus proteins (Neq267 and 268). Identifying the functions of all these proteins and uncovering the mechanism of interaction with the Ignicoccus host will clearly require more in-depth biochemical and ultrastructural inquiry.

I. hospitalis proteome changes induced by N. equitans
Among the major questions surrounding the I. hospitalis-N. equitans relationship is how the presence of N. equitans impacts its host's physiology. By identifying changes in I. hospitalis protein expression occurring when N. equitans is present on its surface, relative to its solitary growth, mechanisms that enable this association should become apparent. In fact, defining the very nature of the association requires such observations. Here we provide a first view of the proteomic changes that occur in I. hospitalis as N. equitans populates its surface.
To evaluate the impact that N. equitans has on its host proteome at broad cellular activities level we have employed a Gene Set Enrichment Analysis (GSEA) [23,24] to measure the variation of I. hospitalis arCOG categories between the two culture conditions (presence or absence of N. equitans). Figure S2 summarizes the result of the GSEA, where positive normalized enrichment scores (NES) reflect the degree to which an arCOG category is overrepresented (enriched) in the I. hospitalis-N. equitans co-culture while negative scores reflect over-represented categories in the pure culture relative to the co-culture. Overall, in the co-culture, GSEA points especially to increased cell cycle control, energy generation, post translation protein modification/turnover and membrane biogenesis accompanied by a decrease in transcription and replication functions. These broad responses are perhaps expected in the culture stage where N. equitans numbers are high and the host has largely ceased cell division and has to cope with its companion's demand. Further individual analyses of the most abundant Ignicoccus proteins were performed in order to dissect its response to Nanoarchaeum in greater detail.
Among the detected I. hospitalis proteins, the relative abundance of approximately 10% (106 proteins) differed by 2-fold or more between the pure and co-culture (Figures 3, 4 and Tables 3, 4; see Table S1 for the relative fold-change values for all detected proteins). As presented above, several major players involved in energy generation, including the constituents of I. hospitalis' ATP synthase, Ni-Fe hydrogenase, and polysulfide reductase appear to be up regulated upon N. equitans' association. The assumed increased rate of respiration and ATP synthesis in I. hospitalis suggests an environment with increased energy demands, most likely resulting from the metabolic needs of N. equitans which itself does not provide compensation for, at least energetically speaking. Increased levels of key central metabolic enzymes of I. hospitalis support this postulate as well. Several of these enzymes provide key substrates for multiple biosynthetic pathways as well as for carbon fixation, including acetyl-CoA synthase (Iho256), pyruvate ferredoxin-oxydoreductase complex (Iho1256-1259), phosphoenolpyruvate synthase (Iho113), phosphoenolpyruvate carboxylase (Iho341), and fructose 1,6-bisphosphatase (Iho363). Their up-regulation in the co-culture may indicate an increased respiratory and basal metabolic burden brought forth by multiple N. equitans populating each I. hospitalis cell. All these inferences will certainly require biochemical validation and additional quantitative proteomic measurements in various co-culture stages.
Among the genetic information processing steps, transcription was negatively impacted the most in the co-culture sample. Multiple subunits of the RNA polymerase complex were significantly reduced (RpoE0, F, H, Rpb8), as were important general transcriptional regulators, such as the initiation (TFIIB, E) and elongation factors (TFIIS). Almost all transcription factors were decreased, some .5 fold (e.g. Iho1027, 0858), with the most impacted being an AsnC family regulator (Iho0308) which was reduced from an RSpC of 120 in the pure I. hospitalis culture to below detection in the co-culture. As follows, perhaps only one transcription factor may be specific for the co-culture state, Iho0122 (AsnC family), although detected at a low level. Gene expression and promoter activity analyses will be required in order to determine the significance of these changes in the transcriptional machinery, specifically in how they affect global transcripts levels and/or correlate to measured protein abundance.
Aside from the variations in transcription factor abundance, several other global cellular regulators are significantly different in the I. hospitalis-N. equitans co-culture. These differences include a two-fold increase in the carbon starvation protein (CstA, Iho1324) and a three-fold increase in a relatively abundant protein, the roadblock/LC7 superfamily protein Iho0929. The function of RLC7 proteins in Bacteria and Archaea is unclear but has been proposed to involve regulation of cellular GTPases [21]. Another important global metabolic regulator, the P II signal transduction protein encoded by glnK (Iho1294), was present at two-fold lower level in the co-culture. GlnK is widely present in Bacteria and Archaea and has complex regulatory functions that integrate energy, carbon and nitrogen metabolism [25]. Binding of 2oxoglutarate and ATP can induce conformational changes that impact the interaction of GlnK with a variety of cellular targets (e.g. enzymes involved in nitrogen metabolism, the ammonium transporter, and transcription factors). By sensing the levels of 2oxoglutarate and glutamate, both substrates for ammonia assimilation, GlnK can induce or repress genes involved in the global nitrogen cycle. Details of this regulatory control in Ignicoccus and the significance of GlnK down-regulation in the presence of N. equitans are still unknown. Potentially linked to that was an observed reduced abundance of several enzymes involved in nitrogen and amino acid/oxoacids metabolism was observed, including carbamoyl-phosphate synthase (Iho1399, 1400), an aminotransferase (Iho0496) and two aldehyde ferredoxin oxidoreductases (Iho25, 1032). As already discussed, the stress response and protein processing machinery of I. hospitalis has a mixed response to the presence of N. equitans, at least with regard to the samples characterized in this study. Most protein chaperones are either unchanged or decreased in co-culture (e.g. hsp20 [Iho1363] and a prefoldin [Iho1176]), perhaps indicating a reduced rate of protein synthesis. Interestingly, the translation initiation factor Iho0676 is reduced as well. However, one prefoldin (Iho0087) is elevated ,2.5 fold, suggesting a potential functional differentiation among the related chaperones. The proteasome (Iho0453, 0616) and a few proteases (Iho0100, 0533) were also modestly elevated (,2-fold) in the co-culture. Most notable, however, was the .5-fold increase in the abundant protein FAD-dependent pyridine nucleotide-disulphide oxidoreductase (Iho0673) and its neighbor (Iho0672), a protein with unknown function. To understand the specificity and the role of these proteins in I. hospitalis association with N. equitans it will be important to determine whether the dynamic response of this pair of gene products is integral to the establishment and/or progression of the co-culture.
The mechanisms that enable the transfer of metabolic products (e.g. amino acids, lipids, nucleotides) from I. hospitalis to N. equitans are completely unknown at this time. Even though the sample analyzed here contained N. equitans in stationary phase, perhaps implying diminished growth and thus metabolic uptake, the proteomic datasets were scrutinized to uncover potential clues to the mechanisms of interaction and/or communication between the two organisms. In particular, we analyzed proteins potentially involved in signaling, defense and transport. Based on homology, the I. hospitalis genome encodes relatively few proteins that can be assigned to these functional categories. Such proteins may be constitutively present or be induced in co-culture as a result of signaling/sensing events. Among the cellular defense mechanisms, the I. hospitalis genome contains nine CRISPR repeat sequences and 37 predicted CAS genes, some of them grouped in distinct operons (Iho0326-0329, Iho0460-0464 and Iho1139-1141). Most of these proteins were detected with several at moderate to high levels (RSpC.100). However, they appear to not be differentially expressed, suggesting the CRISPR defense mechanism is persistently active in this organism or at least does not respond to N. equitans populating its surface. Since the CRISPR system acts in protecting against both phage infection and presence of foreign DNA (a process that is not expected to occur upon N. equitans attachment), the absence of a CRISPR response is not surprising. Another noteworthy protein that responded to the presence of N. Proteins that were only detected in one of the sample types are represented on either axis and an arbitrary spectral count (RSpC) threshold .10 was chosen for coloring. The scatter plot uses the sum of spectral counts for each protein between the three independent measurements for each sample. doi:10.1371/journal.pone.0022942.g003 equitans is Sec61 (Iho1061), a component of the protein secretion apparatus. This protein was only detected (RSpC of 77) in the coculture and, even though there is yet no evidence of protein transfer from Ignicoccus to Nanoarchaeum, the general protein secretion apparatus may be employed during metabolite transfer. Several other membrane proteins with unknown function were also either enriched (Iho1052, 0235) or depleted (Iho0556, 0514) in the co-culture.

Correlations between the proteomic data and genomic operon structure
In Bacteria and Archaea, most genes are organized in cotranscribed and co-regulated clusters of genes or operons. Most operons are not conserved over long evolutionary distances as they are generally under the pressure of genome rearrangement and fragmentation by insertion of mobile elements [26]. In many cases, genes that belong to the same operon encode separate interacting  [15,20]. The localization of most I. hospitalis membrane complexes (outer vs. inner membrane) is unknown and arbitrarily depicted here as spanning both membranes. I. hospitalis proteins detected by proteomics are indicated by yellow diamonds (,1.5 fold change between pure culture and co-culture) or blue or red arrows (.1.5 fold increase or decrease, respectively in co-culture versus single culture). Black diamonds indicate proteins that were not detected. In N. equitans, the variation between the purified sample and the coculture values are not being represented, the yellow/black diamonds only indicate detection/non-detection. doi:10.1371/journal.pone.0022942.g004 subunits of enzymatic complexes or perform coupled metabolic transformations. Even though post-transcriptional and posttranslational regulation could significantly impact the synthesis and steady-state level of the individual proteins encoded by the genes of an operon, intuitively the relative abundance of such proteins would be expected to correlate with the operon structure. The application of multiple complementary approaches to dissect the functional complexity of microbial genomes indicates that additional regulatory processes, including multiple transcription start sites within an operon, antisense and regulatory RNAs, can complicate in silico annotation and functional prediction [27]. Proteomic data has already been successfully used to verify and predict both operons and regulons in some Archaea [9] as well as to pinpoint the existence of significant post-transcriptional regulatory mechanisms that impact the protein abundance of cotranscribed genes [28,29].
One of the genomic characteristics shared by both N. equitans and I. hospitalis is the absence and/or fragmentation of many relatively long and conserved operons that are present in other Archaea and Bacteria, e.g. those that contain ribosomal protein genes. Nevertheless, there are still a sufficient number of predicted, relatively conserved operons, which allows for matching the average abundance of predicted proteins to an operon-specific response brought about by the association of these two Archaea. Considering each operon as a gene set, we used the GSEA [24] to find operons differentially expressed between the two culturing conditions, I. hospitalis grown by itself or in co-culture with Nanoarchaeum. Between those two biological conditions, only a limited number of individual proteins changed by a factor greater than two-fold. Thus, it came as no surprise that only few operons responded significantly to the association. Such operons included those that contain genes strongly up-or down-regulated by the presence of N. equitans (Iho0672-0673, Iho0169-0170, Iho1087-1088), with several encoding interacting subunits of the same enzymatic complex (i.e. NiFe hydrogenase and sulfur reductase). An important result also included operons and/or clusters of transcriptionally independent but adjacent genes that appeared silent, i.e. their encoded proteins were not detected ( Figure 5). Among them is a predicted daunorubicin resistance ABC transporter (Iho0146-0147), an operon contain- ing predicted type II secretion system proteins (Iho1009-1016), and an operon containing five CRISPR-associated protein genes (Iho1132-1135). These sporadically occurring examples could indicate the presence of regulatory mechanisms that prevented expression of these genes under the culture conditions utilized in this study, perhaps precluding their proteomic observation unless exposed to specific environmental challenges. Notable however were multiple operons consisting of clusters of unexpressed genes that encoded hypothetical proteins with no recognizable function (e.g. Iho0262-0264, Iho0432-0437, Iho0806-0811, Iho0893-0896, Iho1023-1026, Iho1312-1316). Future experimental and comparative genomic analyses will be required to determine the evolutionary history and the function of these unrepresented operons and their encoded proteins.

I. hospitalis proteins in N. equitans?
Initial genomic analysis pointed to a strict dependence of N. equitans on I. hospitalis for energetic and biosynthetic precursors (nucleotides, amino acids, lipids, sugars) [15]. Experimental evidence for the transfer of several types of molecules has indeed been obtained [2,3], even though the mechanisms are still completely unknown. A major lingering question regards the transfer of large macromolecules (proteins, RNAs) that would enable N. equitans to perform functions not encoded by its own genome. At present, ultrastructural investigations of the I. hospitalis-N. equitans system have revealed features that could accomplish molecular transfer in two distinct ways [20,30,31]. Small vesicles that engulf cytoplasmic content and migrate through the large inter-membrane compartment (IMC) to the outermembrane, independent of N. equitans presence, represents a relatively non-specific mechanism. This I. hospitalis-based cellular process of yet unknown physiological significance may have been exploited by N. equitans in its evolution as an ectosymbiont/parasite and could cargo not only small molecules but proteins as well, including membrane protein complexes. Alternatively, specialized structures (pores or channels) [20,30] could form at the point of contact between the two organisms, which might facilitate the transfer and, depending on its size and functional mechanism, could control the types of molecules that pass from I. hospitalis to N. equitans. While a previous study focused on identifying protein complexes that form at the point of contact between the two organisms found a significant number of membrane and nonmembrane associated proteins, the molecular specifics of the association remained unclear [7].
In the current study we attempted to address the question of whether or not I. hospitalis proteins may become integrated with the N. equitans cell. It has been shown that towards the end phase of the co-culture, a large fraction of N. equitans cells are released from the surface of I. hospitalis [2]. There is potential that a transfer of host proteins may enable them to survive for a limited time once detached from their host. This possibility remains speculative as it has not been determined if these free N. equitans cells are metabolically active or even able to re-attach to new host cells. Nevertheless, free N. equitans cells can be purified from their hosts. To perhaps provide evidence for I. hospitalis protein transfer, the proteome from a purified N. equitans cell fraction from a late coculture stage was similarly measured. To assess the purity of the N. equitans cell fraction, the level of I. hospitalis cell carry-over was analyzed by both quantitative PCR and fluorescence in situ hybridization [1]; it was determined to be ,1% I. hospitals relative to N. equitans (not shown).
Using the same criteria for peptide identification and assignment as described above, 464 I. hospitalis proteins were observed in at least one of the three independent MS runs performed on purified N. equitans. A total of almost 12,000 spectra were assigned to I. hospitalis proteins, as compared to 120,000 spectra assigned to N. equitans proteins. The relative abundance of most identified I. hospitalis proteins was very low, with only 170 proteins having a combined rebalanced normalized spectral count (RSpC) value of 20 or above, 22 of which having a RSpC of over 100 ( Table 5). Two opposing possibilities may explain the existence of the observed I. hospitalis proteins. The first possibility is that the detected I. hospitalis proteins are contaminants brought through to the purified N. equitans cell sample and, thus have not been transferred specifically from the host. In the late culture phase, numerous I. hospitalis cells have already lysed resulting in their contents being released into the culture medium. Despite the purification steps, a low-level of the soluble host proteins could still remain trapped with the N. equitans cell mass, perhaps by nonspecific binding to the cell surface, and be detected, though at low levels, by the highly sensitive MS approach. I. hospitalis membrane fragments could also co-purify with the N. equitans cells and, since the DNA and ribosomes would remain mostly in solution, the assays used for estimating co-purification of intact I. hospitalis cells may not fully reflect the efficiency of the purification procedure. At the same time, however, some of the free N. equitans cells may have Figure 5. The I. hospitalis proteome and the effect of co-culture with N. equitans at the genome level. Genes are arranged in rows with the coordinates on the left corresponding to the nucleotide position in the genome. Blue indicate RNA genes. In black are genes encoding proteins that were not detected. Green and red correspond to proteins that were up-or down-regulated, respectively, by 1.5 fold or more between the co-culture versus the independent culture samples. Grey indicates proteins with ,1.5 fold variation between the samples. The direction of the slanted rectangle indicates the direction of transcription. Not all consecutive genes transcribed in the same direction are predicted to be in the same transcriptional unit (operon). doi:10.1371/journal.pone.0022942.g005 became isolated by host cell lysis and not by self-detachment. In that case, they would carry host membrane fragments that may be stably and specifically interacting with N. equitans through the hypothetical interacting complex. Finally, the opposing possibility explaining the existence of I. hospitalis proteins suggests that some of the detected I. hospitalis proteins may be a result of specific intercellular transfer and represent integral constituents of N. equitans.
In an attempt to distinguish between these alternatives we compared the relative abundance of the detected proteins between The enrichment factor, w, was calculated as the ratio of the abundance of an individual I. hospitalis protein (P) relative to the total I. hospitalis proteome (T) between the the N. equitans sample and the I. hospitalis-N. equitans co-culture, with a general assumption that the more abundant an I. hospitalis protein is in the host, the more likely that protein could end up as a significant contaminant in the N. equitans fraction. Also, since N. equitans' genome encodes all its necessary genetic information processing systems (replication/repair, transcription, translation, and protein folding), proteins involved in those processes are expected not to be transferred. When the detected I. hospitalis proteins were grouped based on arCOG categories, no changes in relative abundance were observed between the represented functional categories relative to total proteome of the I. hospitalis-N. equitans co-culture. I. hospitalis proteins involved in biosynthetic metabolic functions, prominently expressed in both the pure culture and co-culture, were also detected at very low levels in the purified N. equitans sample. Collectively, these indicate that no specific enrichment of a broad functional category of I. hospitalis proteins occurred in the N. equitans sample and that the bulk of the I. hospitalis proteins represent contaminants. Therefore, these analyses do not support a hypothesis of a broad transfer of metabolic enzymes from the host to N. equitans. There remains, however, the possibility that a few select I. hospitalis proteins may be incorporated by N. equitans either in its membrane or in the cytoplasm. To evaluate this possibility, the individual I. hospitalis proteins detected in N. equitans were analyzed in terms of both abundance and their ''enrichment factor''. We defined that factor, w, as the ratio of the abundance of an individual Ignicoccus protein (P) relative to the total I. hospitalis proteome (T) between the purified N. equitans (N) and the coculture (N) samples (w = [P N /T N ]/[P C /T C ]). A ratio around one indicates no significant enrichment in the N. equitans sample and, therefore, the presence of that protein is likely a result of contamination with I. hospitalis cellular contents. A higher value (w.2) may indicate enrichment, potentially as a result of actual transfer from I. hospitalis to N. equitans. Caveats to this methodology still remain, however. For very low abundance proteins, statistical fluctuations in gene expression and protein measurement may result in high variability. Therefore, proteins for which the normalized spectral counts were below 20 were not taken into account. Less than 50 proteins satisfied these two filtering criteria (w.2, RSpC.20)(Table 5, Figure 6). Over half of these proteins are predicted to be part of membrane complexes, including subunits of the ATP synthase, Ni hydrogenase, polysulfide reductase and its associated ferredoxin, ABC transporters and the I. hospitalis outer-membrane pore oligomer. While these are clearly important components of I. hospitalis' energetic metabolism and molecular transport, the question of whether their presence in purified N. equitans cells indicates transfer to N. equitans or the presence of co-purifying I. hospitalis membrane fragments cannot be answered at this time. Similarly, a few other membraneassociated proteins were detected at significant levels, including the recently described appendage protein Iho0670 [14] and two potential cell surface proteins with no known function (Iho0759, 0981). Further biochemical and cellular localization studies will be required in order to map their location in the I. hospitalis-N. equitans system and to test whether or not they play a role in the cellular interaction between the two Archaea.

Conclusions
The relationship between I. hospitalis and N. equitans remains one of the most intriguing associations in nature, enabled by mechanisms that are still largely elusive. The deep proteomic coverage obtained for both species confirmed the streamlined nature of their genomes, with most genes constitutively expressed and containing relatively few pseudo-or redundant genes. The proteomic study describe here uncovers several of the physiological changes that occur in I. hospitalis when N. equitans colonizes its surface. A somewhat surprising finding was the relatively low level of proteomic perturbation induced by N. equitans on its host's proteome, with only approximately 10% of the proteins exhibiting an increase or decrease in relative abundance by two or more fold. On the other hand, as most I. hospitalis genes are expressed under normal conditions and given its strict chemoautotrophic growth with few metabolic alternatives available, perhaps its association with N. equitans could only be viable with a tight, selective response at the level of specific genes and processes. The biological effects of small quantitative changes in specific proteins are difficult to predict since different metabolic reactions and cellular processes are controlled by a myriad of mechanisms and rate-limiting reactions. Nevertheless, concerted changes in the abundance of proteins that are co-expressed from the same operons or from different chromosomal regions but are part of the same molecular complexes or metabolic pathways suggests that's changes in some cellular processes may indeed occur.
Observed increases in the abundance of subunits of membrane complexes involved in respiration and ATP synthesis (hydrogenase, sulfur reductase, ATP synthase) suggest an increased energetic and metabolic demand imposed by the presence of N. equitans. Since it is not yet possible to measure the respiration and ATP synthesis rates in this system, this remains unverified. At the metabolic level, the co-culture appeared to induce an increase in the abundance of several key enzymes involved in carbon fixation, most notably the pyruvate ferredoxin oxidoreductase, PEP synthase and PEP carboxylase. As every biosynthetic reaction employed by I. hospitalis (including all the metabolites that N. equitans receives from its host) depend on the reactions catalyzed by those enzymes, this response further substantiates the claim that the association imposes an increased metabolic and energetic demand on the host, I. hospitalis.
Other systematic affects induced by I. hospitalis' association with N. equitans include a reduction of the basal transcriptional machinery and most of the transcriptional regulatory factors. This may result in a general decrease in the overall rate of transcription. Noted, however, was an observed increase in the level of several transcription factors. While the expectation is that these regulatory proteins will specifically influence the expression of selected genes, actual data linking them to either the transcription of mRNA or translation to protein is lacking. Complementing the transcriptional response denoted above, a decrease in the abundance of key translational control proteins (e.g. the translation initiation factor 1A) as well as several proteins involved in protein folding seems to indicate that host cellular growth may be reduced at the expense of increased metabolism in order to support N. equitans. This conclusion is similar to microbiological data that indicate a much lower rate of I. hospitalis cellular division after an increasing number of N. equitans cells have populated their cell surface [2]. These data perhaps converge to a conclusion that N. equitans acts much more like a parasite than a commensal, at least under these laboratory culture conditions. Ongoing research focuses on the dynamics of these two organisms, with measurements acquired during progression of a co-culture; a study that incorporates both quantitative proteomics and transcriptomics to garner an enhanced understanding of this relationship.
The proteomic characterization of the purified N. equitans sample, while complicated by the potential co-purification of I. hospitalis proteins and/or cellular debris, indicates that N. equitans does not receive from its host a significant amount (if any) of enzymes that would enable it to perform primary biosynthetic reactions. The mechanism employed by N. equitans to acquire amino acids, lipids, nucleotides, still remains unknown. Based on enrichment of specific I. hospitalis membrane-associated proteins in the N. equitans sample, the potential transfer of complexes involved in generation and maintenance of a membrane potential and ATP synthesis cannot yet be ruled out and will thus require cellular localization studies before a definitive conclusion can be drawn. Several proteins potentially involved in the interaction or molecular transfers between the two organisms emerged and represent targets for further analyses. A combination of systems biology approaches with biochemical and ultrastructural studies are clearly required in order to fully comprehend this fascinating association.

Cell cultures
I. hospitalis and I. hospitalis-N. equitans were cultured at 90uC in a 50 liters enamel-protected fermentor using K SME medium, elemental sulfur and a H 2 /CO 2 gas phase, as previously described [2,32]. The cultures were harvested in stationary phase, rapidly cooled and cells isolated by centrifugation followed by freezing in liquid nitrogen and storage at 280uC. In the I. hospitalis-N. equitans co-culture, essentially all Ignicoccus harbored multiple N. equitans cells on their surface as determined by microscopic analysis [2]. To obtain higher proteome coverage, we also purified an N. equitans fraction from a separate co-culture using differential centrifugation [1]. Based on FISH analysis and quantitative PCR with speciesspecific primers, the purified fraction contained .99% N. equitans cells.

Sample preparation for 2D-LC-MS/MS
Frozen cell pellets (10-50 mg) were re-suspended in 1 ml SDS lysis buffer (4% SDS, 100 mM Tris-HCl pH 8.0, 50 mM DTT) and incubated in boiling water for 5 min. To aid lysis and to completely solubilize the cells, samples were pulse-sonicated (10 s on, 10 s off) for 2 min with an ultrasonic disruptor (Branson) at 20% amplitude. Samples were then boiled again for 5 min, cleared by centrifugation (10 minutes at 21,000 g) and immediately precipitated with 20% trichloroacetic acid (TCA) overnight at 220uC. TCA-precipitated proteins were washed with two additions of ice-cold acetone, air dried, and re-suspended in 8 M urea, 100 mM Tris-HCl pH 8.0. To aid in re-solubilization, Figure 6. I. hospitalis proteins detected in the purified N. equitans sample. The graph shows the normalized spectral count of detected proteins versus the enrichment factor, w. The enrichment factor takes into account the difference in proteome abundance and complexity between the samples. A value of w = 1 corresponds to no enrichment and can explain protein presence based on carry over/contamination. A region with w.2 and NSpC.20 was shadowed as comprising of the most likely proteins transfer candidates or otherwise specifically associated with Nanoarchaeum. Below NSpC of 20 the low abundance may affect accurate measurements and those proteins were not taken into account. The insert cartoons depict possible explanations of detecting Ignicoccus proteins (green dots) in the Nanoarchaeum sample: (1) real inter-species transfer; (2) adherence of Ignicoccus membrane fragments to Nanoarchaeum, potentially mediated by a distinct ''interactome'' (red) or (3)  samples were sonicated as before, taking care to keep the sample below 37uC, and incubated at room temperature (RT) for 30 minutes. After measuring protein concentration via a BCA assay, samples were adjusted to 10 mM DTT (10 minutes, RT) and then 10 mM iodoacetamide (10 minutes at RT in the dark) to block reduced cysteine residues. Sample aliquots containing ,1-2 mg of crude protein were diluted to 4 M urea (1:1) with 100 mM Tris-HCl pH 8.0, 20 mM CaCl 2 , and pre-digested with sequencing-grade trypsin (Promega) at a 1:75 (w/w) enzyme:protein ratio overnight at room temperature. Samples were then diluted to 2 M urea (1:1) and digested with a second aliquot of trypsin (1:75) for an additional 4 hours. Following digestion, each sample was adjusted to 150 mM NaCl, 0.1% formic acid and filtered through a 10 kDa cutoff spin column filter (VWR brand). The peptide-enriched flow through was then quantified by BCA assay, aliquoted, and stored at 280uC until analysis.

Measurement of peptides by 2D-LC-MS/MS
For each sample, 50 ug of peptides were bomb-loaded onto a biphasic MudPIT back column [33] packed with ,5 cm strong cation exchange (SCX) resin for charge-based separation of peptides followed by ,3 cm C18 reversed phase (RP) for online desalting (Luna and Aqua respectively, Phenomenex). Once loaded, the sample columns were washed offline with solvent A (5% acetonitrile, 95% HPLC-grade water, 0.1% formic acid) for 15 minutes to remove residual urea and NaCl followed by a gradient to 100% solvent B (70% acetonitrile, 30% HPLC-grade water, 0.1% formic acid) over 30 minutes to move the peptide population from the RP to the SCX resin. Washed samples were then placed in-line with an in-house pulled nanospray emitter (100 micron ID) packed with 15 cm of C18 RP material and analyzed via 24-hr MudPIT 2D-LC-MS/MS (11 salt-pulses: 5, 7.5, 10, 12.5, 15, 17.5, 20, 25, 35, 50, 100% of 500 mM ammonium acetate followed by a 100 minute gradient to 50% solvent B) with a hybrid LTQ XL/Orbitrap mass spectrometer (Thermo Fisher) operating in data-dependent mode. Full MS1 scans (2 microscans; 5 MS/MS per MS1) were obtained using an Orbitrap mass analyzer set to 30K resolution, while MS/MS scans (2 microscans) were obtained/performed in the LTQ. A total of three replicate measurements were obtained for each sample.

Database searching and protein identification
Peptide fragmentation spectra obtained from each of the 6 MS sample measurements were assigned peptide sequences with SEQUEST [34] using a composite FASTA database containing the proteomes of I. hospitalis, N. equitans (GenBank CP000816.1, AE017199.1), known protein contaminants, as well as reversed entries for all the aforementioned constituents which were used to assess false-discovery rates. In addition, sample processing-induced modifications were taken into account, specifically carboxymethylation of cysteine. Following database searching, DTASelect [35] was used to both filter the data (XCorr: +1 = 1.8, +2 = 2.5, +3 = 3.5, DeltCN 0.08) and consolidate the identified peptides into protein loci. In order for a protein call to be made, it must have been identified by at least two peptides, with one unique to that specific protein entry.

Data analysis
To prepare for semi-quantitative proteome analysis, DTASelect-filtered data was subjected to spectral count balancing and subsequent NSAF (Normalized Spectral Abundance Factor) determination [36,37,38]. With regards to the former, the unique-status of each identified peptide was assessed. If a peptide was deemed unique, it retained 100% of its previously assessed spectral count (SpC). However, if a peptide was found to belong to more than one representative protein, i.e. non-unique, its SpC was recalculated based on the ratio of uniquely identified peptides between the 2 or more proteins that share the non-unique peptide in question. The adjusted SpC of the proteins (aSpC) incorporate these balanced values, which correct for the slight quantitative bias that occurs with homologous proteins and/or proteins with homologous regions, both of which artificially inflate SpC values at the expense of proteins that share no homology. Once spectrally balanced, NSAF values were calculated for each protein in a specific run to normalize individual MS runs based on the total number of spectra collected and protein length, which by itself introduces a bias that favors larger proteins if not corrected for [36]. NSAF values were then multiplied by 46541, the average total spectra count observed across all sample sets. This converts the NSAF decimal value to a theoretical, normalized spectral count (nSpC) value, which is easier to visualize.
As the quantitative focus of this study was to determine how and to what extent the I. hospitalis proteome changes in response to N. equitans, nSpC values for only I. hospitalis proteins were renormalized so that the total nSpC of the co-culture equaled that of the isolate. These re-normalized nSpC values, denoted as RSpC, were then used in the ensuing analysis. This re-normalization effectively corrects for the observed systematic depression of collected spectra for I. hospitalis proteins identified in the coculture, which was a result of the increased proteomic complexity introduced by the addition of N. equitans proteins. Semiquantitative, RSpC data was then binned into two categories: (i) I. hospitalis proteins that were identified in every MS run (36I. hospitalis runs vs. 36co-culture runs) and (ii) proteins that did not have a complete measurement series. The former category, which by default infers a greater confidence of identification, was used to compute the statistical relevance of each protein's abundance change due to the association of the two organisms. Using standard proteomic data processing [36,37], the RSpC values were log 2 -transformed to create a normalized distribution of abundance values from which further statistical testing was performed. In this regard, significant changes in abundance for each identified I. hospitalis protein, dependent on sample type (pure culture vs. co-culture) was assessed by a Student's T test (p#0.05).

Genomic and metabolic inferences
To analyze the proteomic data in a biological context, previous metabolic reconstructions of I. hospitalis and N. equitans [4,15] were used and updated based on more recent advances in understanding this archaeal system [19]. Inferred biochemical activities/biological functions used PFAM, KEGG and arCOG [13] classification. Cellular localization was inferred based on Signal P and TMHHM as part of the annotation deposited in IMG [39] as well as based on Pred-Signal, an algorithm optimized for Archaea [40] The role and cellular localization of several proteins has been validated previously [2,7,20,41]. To visualize the relative abundance of the proteins in both genomic and metabolic context, the ''Pathways Tools Omics Viewer'' from BioCyc was utilized [42,43]. Operon prediction relied on both BioCyc and DOOR [44]. To identify significant variation in the protein levels between the two culture types and also link them to I. hospitalis individual genes and operon structure, we used a Gene Set Enrichment Analysis, GSEA [24]. The Ignicoccus proteins were ranked according to their signal-to-noise ratio (the difference of means between NSAF values in both conditions scaled by the standard deviation), then the normalized enrichment score (NES) was calculated as described by Subramanian et al. [24]. Table S1 Ignicoccus hospitalis and Nanoarchaeum equitans proteins identified by proteomics and their relative abundance levels. The genes/proteins are arranged in order of gene number based on Genbank. Classifications based on arCOG and membrane association inferences are also shown. For each sample type (Ignicoccus hospitalis culture, Ignicoccus-Nanoarchaeum co-culture or purified Nanoarchaeum equitans sample the sum of the three independent experimental measurements of relative abundances are provided in order of the normalization process: ASpC (Adjusted Spectral Counts); nSpC (Normalized Spectral Counts) RSpC (Re-normalized Spectral Counts). RSpC values were used for all analyses, the other values are only provided as reference.