Comparison of Online Comprehensive HILIC × RP and RP × RP with Trapping Modulation Coupled to Mass Spectrometry for Microalgae Peptidomics

: In this work, two online comprehensive two-dimensional liquid chromatography platforms, namely Hydrophilic interaction liquid chromatography × Reversed phase (HILIC × RP) and Reversed phase × Reversed Phase (RP × RP) coupled to mass spectrometry, were compared for the analysis of complex peptide samples. In the ﬁrst dimension, a HILIC Amide and C18 Bioshell peptide (150 × 2.1 mm, 1.7 and 2.0 µ m) columns were selected, while, in the second dimension, a short C18 (50 × 3.0 mm, 2.7 µ m) Bioshell peptide column was used. Two C18 trapping columns (10 × 3.0 mm, 1.9 µ m), characterized by high retention and surface area, were employed as modulation interface in both HILIC × RP and RP × RP methods. The LC × LC platforms were coupled to UV and tandem mass spectrometry detection and tested for the separation and identiﬁcation of two gastro-intestinal digests of commercial microalgae formulations (Spirulina Platensis and Klamath). Their performances were evaluated in terms of peak capacity, maximum number and properties of identiﬁed phycocyanin peptides. Our results showed that the HILIC × RP approach provided the highest peak capacity values ( n c HILIC × RP: 932 vs. n c RP × RP: 701) with an analysis time of 60 min, while the RP × RP approach was able to identify a slight higher number of phycocyanin derived peptides (HILIC × RP: 88 vs. RP × RP: 103). These results point out the ﬂexibility and potential of HILIC × RP and RP × RP based on trapping modulation for peptide mapping approaches.


Introduction
Liquid chromatography coupled to mass spectrometry (LC-MS) represents the gold standard for the analysis of complex samples, such as in omics studies. These samples can contain hundreds or thousands of analytes with different chemical properties and their characterization is a challenging task. One-dimensional liquid chromatography coupled to mass spectrometry (1D-LC-MS) is not capable of providing an adequate resolving power [1] and, despite the advances of mass spectrometry, high resolution front-end separation techniques are still essential. Comprehensive two-dimensional liquid chromatography (LC × LC), in which the entire sample is subjected to two separation with different selectivity, has emerged as one of the best strategies for the analysis of very complex samples [2].

Sample Preparation
Protein extraction and gastro-intestinal (G.I) digestion were performed as previously developed [12][13][14]. In particular microalgae powders (15 g) were extract with 240 mL of 0.1 M sodium phosphate buffer (ratio w/v of 1:16), with 3 freeze-thaw cycles of 4 h (from −20 • C to 37 • C), finally the suspension was Separations 2020, 7, 25 3 of 12 centrifuged and the supernatant recovered. For G.I, the sample in a concentration of 10 mg/mL was kept under stirring at 37 • C, the pH was adjusted to 2 with HCl 0.1 N and pepsin (enzyme/protein ratio 1:100 w/w) was added. The reaction was blocked after 2 h increasing the temperature to 90 • C for 10 min. The solvent was thus removed under reduced pressure, and 10 mL of a 10 mM HCOONH 4 solution (pH adjusted to 7.5 with NH 4 OH), chymotrypsin, pancreatin bile salts were then added (same enzyme/protein ratio). The reaction, under stirring, was terminated after 2 h, by lowering the pH to 2 with TFA. Peptide digest was solubilized in 0.1% (v/v) aqueous TFA (final concentration 10 mg/mL), and loaded on a Strata ® X Polymeric 100 mg/mL RP cartridge (Phenomenex ® Castel Maggiore, Bologna, Italy), previously equilibrated in 0.1% TFA. After loading, the sample was eluted with MeCN:H 2 O:TFA (70:30:0.1 v/v/v).

Instrumentation
(1D-LC) and LC × LC analyses were performed on a Shimadzu Nexera UHPLC (Shimadzu, Milan, Italy), consisting of a CBM-20A controller, four LC-30AD dual-plunger parallel-flow pumps, a DGU-20A5 degasser, an SPD-M20A PDA detector with a 2.5 µL flow cell, a CTO-20AC column oven, a SIL-30AC autosampler, an external oven (ThermaSphere TMTS-130 HPLC Column) was used for the 1 D column. An additional LC-20AT (Shimadzu) pump was used to deliver the post 1 D dilution flow by means of a stainless steel Tee union, 1/16 in. 0.15 mm bore (Vici-Valco ® Houston, TX 77255, USA). The two dimensions were connected through an ultra high pressure 10 port-two position switching valve with micro-electric actuator (model FCV-12 AH, 1.034 bar, Shimadzu, Kyoto, Japan), placed inside the column oven. The trapping columns were connected to the valve by Viper capillaries of 100 × 0.130 mm I.D (Thermo Fisher Scientific, Milan, Italy). All connections were 0.130 mm I.D. and kept of the shortest length possible. Both dimensions and the valve switching were controlled by the LCMS solution ® software (Version 5.54, Shimadzu). The LC × LC was coupled online to an LTQ-Orbitrap XL (Thermo Scientific, Bremen, Germany) equipped with an electrospray source operated in positive mode. The LC × LC data were visualized into two and three dimensions using Chromsquare ® ver. 1.5.01 software (Chromaleont, Messina, Italy). LC × LC-MS/MS data were aligned by the open source MZmine2.

LC and HRMS Parameters
The two dimensions were optimized in previous experiments, detailed parameters are reported in supporting material. For LC × LC analyses the LC parameters were: 1  The modulation time was 45 s. The flow from the 2 D was split by a Tee union prior the electrospray (ESI) source (approximately 0.4 mL/min to MS). Detection was performed positive mode as follows: spray voltage was set at +4.5 kV, sheath gas arbitrary units 40, auxiliary gas arbitrary units 12, and capillary temperature 250 • C. MS/MS spectra were collected in data-dependent mode (DDA), over the m/z range of 300-2000, at 15,000 resolution. All MS/MS spectra were collected using a normalized collision energy of 35% and an isolation window of 2 m/z, minimum signal threshold 100, and monoisotopic precursor enabled. Ion trap and Orbitrap maximum ion injection times were set to 50 and 100 ms, respectively. Automatic gain control was set to 2 × 10 5 for full Fourier transform mass spectrometry (FTMS) scan and 3 × 10 4 ions for IT. Dynamic exclusion was enabled with a repeat count of 1 and a repeat duration of 30 s. Preview mode for FTMS master scan was enabled. Thermo RAW datafiles were converted in mzXML format by ProteoWizard open source software and elaborated with a free trial of Peaks 8.5 (Bioinformatic Solution, Waterloo, Canada) by using the (database: Aphanizomenon flos-aquae and Arthrospira Platensis, release UniProt 2017) mass accuracy tolerance set at 15 ppm for MS and 0.5 Da for MS/MS. The oxidation of methionine was selected as variable modification.

Results and Discussion
In peptide mapping approaches, the sample is usually subjected to trypsin enzymatic proteloysis to generate multiple peptides per protein, and for a complex sample, this results in hundreds or thousands of peptides. When multiple enzymes are used, such as during the gastro-intestinal digestion, this situation is further complicated by subsequent cleavages, and the exact number of derived peptides cannot be calculated. 1D-LC-MS is unable to provide adequate peak capacity values and, as result the MS is overwhelmed by coeluting peptides entering into the MS at the same time. LC × LC is one of the best strategies to simplify the sample prior to MS analysis, expanding the potential of the MS itself in terms of sensibility and identification capability [15].

Evaluation of Chromatographic Conditions in 1 D (HILIC and RP) and 2 D (RP)
The selection of stationary phase, as well as the chromatographic method development is an essential aspect in LC × LC. In this study five columns were investigated as 1 D (three HILIC and two RP). All the HILIC columns were characterized by same geometry (Length and internal diameter) but different chemistry (amide, bare silica, and diol) and particle size (fully or superficially porous particles). Contrariwise, RP columns were also different in geometry (2.1 mm vs. 1.0 mm I.D). Flow rate, buffer selection, column temperature, gradient, and injection volume were based on previously optimized conditions [11]. It can be observed from Figure 1, that reports the separation of the entire gastro-intestinal digest of Klamath formulation, how Amide stationary phase showed higher resolution and peak capacity (n c amide: 209, n c bare silica: 180, n c diol: 150), together with an increased retention, and a different selectivity over the elution window.
Differently from HILIC, in RP, we compared both a microbore and narrowbore column, in this case only basic pH was investigated, since selectivity can only be achieved by changes in charge on peptide functional groups [4]. While microbore columns are largely used in LC × LC to reduce the volume injected onto the 2 D [2], they can be easily overloaded, furthermore, without a micro LC pump, the very low flow rate needed requires an additional flow splitter after the 1 D mixer, which adds extra-column volume and gradient delay. As a result, the 2.1 mm I.D column outperformed the 1.0 mm I.D and was selected as candidate (Supplementary Figure S1a compromises should be made, such as adapt the flow rate and gradient to facilitate the sampling of peaks, thus HILIC and RP were further adapted lowering the flow rate and adjusting the gradient, the final results are depicted in Figure 2. Differently from HILIC, in RP, we compared both a microbore and narrowbore column, in this case only basic pH was investigated, since selectivity can only be achieved by changes in charge on peptide functional groups [4]. While microbore columns are largely used in LC × LC to reduce the volume injected onto the 2 D [2], they can be easily overloaded, furthermore, without a micro LC pump, the very low flow rate needed requires an additional flow splitter after the 1 D mixer, which adds extra-column volume and gradient delay. As a result, the 2.1 mm I.D column outperformed the 1.0 mm I.D and was selected as candidate (Supplementary Figure S1a,b). Clearly in LC × LC some compromises should be made, such as adapt the flow rate and gradient to facilitate the sampling of peaks, thus HILIC and RP were further adapted lowering the flow rate and adjusting the gradient, the final results are depicted in Figure 2.   Differently from HILIC, in RP, we compared both a microbore and narrowbore column, in this case only basic pH was investigated, since selectivity can only be achieved by changes in charge on peptide functional groups [4]. While microbore columns are largely used in LC × LC to reduce the volume injected onto the 2 D [2], they can be easily overloaded, furthermore, without a micro LC pump, the very low flow rate needed requires an additional flow splitter after the 1 D mixer, which adds extra-column volume and gradient delay. As a result, the 2.1 mm I.D column outperformed the 1.0 mm I.D and was selected as candidate (Supplementary Figure S1a,b). Clearly in LC × LC some compromises should be made, such as adapt the flow rate and gradient to facilitate the sampling of peaks, thus HILIC and RP were further adapted lowering the flow rate and adjusting the gradient, the final results are depicted in Figure 2.  The 2 D RP separation was optimized with a mixture of five standard peptides ranging from a dipeptide to an octapeptide (Supplementary Figure S2). For UV analysis the acidic modifier was 0.1% phosphoric acid on the other hand 0.1% formic acid was employed only for MS detection. The employment of a core shell particle column especially designed for fast and efficient separation Separations 2020, 7, 25 6 of 12 of peptide was adapted from a previous work [11], together with flow rate, column temperature, and modulation time, that have been kept fixed between HILIC × RP and RP × RP.

HILIC × RP and RP × RP with Trapping Modulation: System Performance Evaluation
In HILIC × RP, the transfer of the acetonitrile rich mobile phase is the main drawback and, as stated before, different approaches can be used to overcome this challenge. While the amount of organic solvent transferred in RP × RP is lower, at least at the beginning of 1 D gradient, poor peak focusing can still occur resulting in reduced sensibility and peak distortion. Among the possible approaches, active solvent modulation has been recently proposed also for peptide mapping [16]. In this regard, based on our previously developed HILIC × RP system we demonstrated that the employment in the modulation interface of highly efficient and retentive trapping columns is highly beneficial [11]. With the aim to compare both techniques for peptide mapping, we used the same trapping modulation interface also in a RP × RP approach. The dilution flow ratio prior the trapping was 1:10 for HILIC × RP and was adapted to the different percentage of acetonitrile among the two set-ups, with a final ratio of 1:2 for RP × RP (data not shown). Trapping columns were operated in forward flush mode, to extend system lifetime, given the high backpressure generated by the 1.9 µm particles. The 2D and 3D-UV and MS plot relative to the separation obtained with the HILIC × RP and RP × RP approaches are depicted in Figure 3a-d respectively. Sharp and symmetrical peaks were obtained with both approaches, the difference in peak distribution across the 2D separation space is clearly related to the different orthogonality of the methods. In RP × RP, peaks tend to be more concentrated along a diagonal line, even if the shifted gradient help to maximize the 2D space occupation [17], on the contrary in HILIC × RP there is a better, and partially random, occupation of 2D separation area. The performance of the two LC × LC platforms are summarized in Table 1. The peak capacity were calculated taking into account both Sharp and symmetrical peaks were obtained with both approaches, the difference in peak distribution across the 2D separation space is clearly related to the different orthogonality of the methods. In RP × RP, peaks tend to be more concentrated along a diagonal line, even if the shifted gradient help to maximize the 2D space occupation [17], on the contrary in HILIC × RP there is a better, and partially random, occupation of 2D separation area. The performance of the two LC × LC platforms are summarized in Table 1. The peak capacity were calculated taking into account both undersampling and orthogonality. For both 1 D HILIC and RP, peak capacity has been calculated with Equation (1).
where t r,l and t r,f are the retention times of the last and first eluting peaks in the 2D maps, and 1 w avg is the average 4σ peak width ( 1 D) of five peaks across the 2D space, and covering the whole separation window. The 2 D RP peak capacity was calculated according to Equation (2): where 2 t g is the 2 D gradient time, without the re-equilibration time and 2 w avg is the average 2 D 4σ peak width, relative to the same five peaks considered for the 1 D calculation. The theoretical peak capacity of the two LC × LC systems was calculated as follows: and subsequently corrected for the undersampling effect [18] with the Equation (4).
in which the term β is the correction factor described in the Equation (5): where 2 t c is second dimension cycle time, and w the average 1 D peak width. This value must be corrected taking into account the correlations between the two dimensions. The asterisk method [19] has been employed for the estimation of the orthogonality (A 0 ). The maximum value of A 0 = 0.70 was obtained with the HILIC × RP system, which is as expected higher than the RP × RP (0.59), where the selectivity is only given from the different mobile phase pH, and thus limited. Interestingly, these values are very similar to the model of Gilar [7]. Finally, the values of 932 and 701 were obtained for HILIC × RP and RP × RP. It should be noted that we fixed the total analysis time to 60 min, and the employment of longer gradients in 1 D, which is very common in peptide mapping, can lead to higher peak capacity values.

Comparison of HILIC × RP and RP × RP Coupled to Tandem Mass Spectrometry for Peptide Mapping
While a huge number of peptides belonging to different phycobiliproteins have been detected by both approaches, in this study we focused only on phycocyanins, since these proteins are the most interesting for human health [14], representative spectra are reported in Figure S4.
One of the main challenges in the coupling of LC × LC to MS is the narrow peak widths, and the acquisition speed of the Orbitrap is not suited for this task, thus the resolution should be drastically reduced for data dependent acquisition. Despite this mass accuracy was well below 5 ppm (average −2.76 ppm, see supporting material). Figure 4 reports the origin of peptides detected with the two approaches, and, as can be observed in both approaches the half of identified peptides (HILIC × RP: 46%, RP × RP: 44%) came from C-phycocyanin alpha chain, a similar value was obtained for C-phycocyanin beta chain (HILIC × RP: 27%, RP × RP: 30%) while opposite values were obtained from Allophycocyanin peptides (alpha: HILIC × RP: 10%, RP × RP: 18% and beta: HILIC × RP: 17%, RP × RP: 8%).  The highest number of peptides, without considering de novo sequencing peptides, was detected obtained with the RP × RP approach (103 vs. 88), the detailed list of peptides identified is reported in Supplementary Table S1. The difference among the two approaches can be related to the chemical nature of some peptides that, because of the effect of highly eluotropic 1 D mobile phase,  The highest number of peptides, without considering de novo sequencing peptides, was detected obtained with the RP × RP approach (103 vs. 88), the detailed list of peptides identified is reported in Supplementary Table S1. The difference among the two approaches can be related to the chemical nature of some peptides that, because of the effect of highly eluotropic 1 D mobile phase, cannot be efficiently retained even with the high retentive trapping columns, and thus are lost as breakthrough effect in the HILIC × RP approach The largest part of peptides were comprised between penta and octapeptides ( Figure 5) and their molecular weight ranged between 417 and 1678 Da (Figure 6), as can be appreciated very similar results were obtained with the two approaches. In fact, for both Spirulina and Klamath peptides, a high degree of overlap between the identifications with the two separation approaches was obtained, with ≥ 60% in common between the two techniques ( Figure 7).  The highest number of peptides, without considering de novo sequencing peptides, was detected obtained with the RP × RP approach (103 vs. 88), the detailed list of peptides identified is reported in Supplementary Table S1. The difference among the two approaches can be related to the chemical nature of some peptides that, because of the effect of highly eluotropic 1 D mobile phase, cannot be efficiently retained even with the high retentive trapping columns, and thus are lost as breakthrough effect in the HILIC × RP approach The largest part of peptides were comprised between penta and octapeptides ( Figure 5) and their molecular weight ranged between 417 and 1678 Da (Figure 6), as can be appreciated very similar results were obtained with the two approaches. In fact, for both Spirulina and Klamath peptides, a high degree of overlap between the identifications with the two separation approaches was obtained, with ≥ 60% in common between the two techniques ( Figure 7).     The suitability of HILIC in online 2D-LC coupling has been demonstrated in several papers. In particular, with respect to the employment of strong cation exchange (SCX) as 1 D for neuropeptides [20] as well as for phosphopeptides [21]. In this work, we carried out a direct comparison between HILIC × RP and RP × RP with a trapping columns modulation interface, even though for a limited number of proteins. These data evidence that both strategies are highly suitable and flexible for peptide mapping. Modifications such as the scale down to capillary or nano-LC conditions could further boost the sensitivity, while the employment of faster MS devices could result in higher The suitability of HILIC in online 2D-LC coupling has been demonstrated in several papers. In particular, with respect to the employment of strong cation exchange (SCX) as 1 D for neuropeptides [20] as well as for phosphopeptides [21]. In this work, we carried out a direct comparison between HILIC × RP and RP × RP with a trapping columns modulation interface, even though for a limited number of proteins. These data evidence that both strategies are highly suitable and flexible for peptide mapping. Modifications such as the scale down to capillary or nano-LC conditions could further boost the sensitivity, while the employment of faster MS devices could result in higher identification capability. The comparison of a non-trapping (loop) modulation interface with the current strategy will be presented in a subsequent paper.

Conclusions
Microalgae represent a growing market in the nutraceutical field, especially for the high protein content, and the bioactivity which is related to the peptides derived from their hydrolysis. While the healthy properties are constantly being reported in literature, there is a lack of knowledge about the nature of G.I derived peptides, thus peptidomic approaches are necessary. Comprehensive LC × LC has a great potential in peptide mapping, but can suffer from sensitivity and solvent incompatibility problems related to the modulation process. In this work we focused on this aspect, and by means of a recently developed trapping columns modulation interface we described a comparison between HILIC × RP and RP × RP for the peptide mapping of microalgae formulations. Both methods showed that the employment of two highly retentive trapping columns as modulation interface deliver high peak focusing and boost sensitivity and efficiency. The results showed similar performances between the two strategies, with slightly better identification capability for the RP × RP setup. It is noteworthy that both setups provided high peak capacity in relative short time for bottom-up approaches and a good coverage of phycocyanins, which were the main target of this study. These data highlight the potential of trapping column modulation with both stationary phase combinations and further enforce the flexibility and potential of LC × LC for complex sample analysis such as peptide gastro-intestinal digests.