Novel Phenolic Constituents of Pulmonaria officinalis L. LC-MS/MS Comparison of Spring and Autumn Metabolite Profiles

Lungwort (Pulmonaria officinalis L., Boraginaceae) is considered to possess therapeutic properties and it has been traditionally used as a remedy against various lung disorders in many countries. Nevertheless, very few data concerning its phytochemical composition are available. This research aims to provide a detailed description of specialized metabolites from the aerial parts of lungwort. Nine previously undescribed and 36 known phenolic compounds were detected in the 50% methanolic extract. Following multistep preparative procedures, structures of newly discovered compounds were determined using one- and two-dimensional techniques of NMR spectroscopy. Among the identified compounds were caffeic acid esters with aliphatic hydroxycarboxylic acids, conjugates of dicaffeic acid with rosmarinic acid, and previously unknown isomers of isosalvianolic acid A and yunnaneic acid E, as well as other lignans. Concentrations of all identified phenolic derivatives in the investigated herbal material were estimated using a method based on liquid chromatography with high-resolution mass spectrometry detection. Seasonal changes in the concentration of metabolites were also investigated using targeted and untargeted metabolomics techniques.


Introduction
Pulmonaria officinalis L (lungwort), belonging to the Boraginaceae family, is a herbaceous perennial plant, widely spread in Europe and western Asia. It has a long tradition of use in folk medicine of many countries as a remedy against various respiratory diseases including asthma, chronic bronchitis, tuberculosis, laryngitis, and coughs. It also has expectorant, antitussive, and diaphoretic properties [1][2][3][4][5][6]. Other ethnomedicinal sources indicate that infusions or decoctions of P. officinalis are administrated as astringent, anticoagulant, anti-microbial, and anti-inflammatory herbs, as well as a remedy for urinary disorders, cystitis, moreover, it shows diuretic and anti-lithiasis activities [2,7,8]. Applied externally, it can be very beneficial in the treatment of burns, wounds, cuts, and eczema [1,2]. P. officinalis extract was tested as a component of bioactive hydrogels, which can be used in the treatment of wounds with heavy and medium exudates [9]. Aerial parts of P. officinalis, commercially available as Pulmonariae Herba, in combination with Tussilago farfara (coltsfoot), are often used to treat chronic cough, including
The 13 C-NMR (DEPTQ-135) (distorsionless enhancement by polarization transfer including the detection of quaternary nuclei) spectrum of 38 contained 28 signals, sorted by HSQC and HMBC spectra, as one CH 3 , two CH 2 , 11 CH, and 14 quaternary carbon atoms ( Table 3). The high-field shifted resonances at δ C 174.5, 172.7, and 164.2 suggested the presence of three carboxyl groups (C-9, C-9 , and C-9 , respectively). The aromatic region of the 1 H and COSY spectra of 38 revealed the presence of three aromatic rings in the structure. Two of them were 1,3,4-trisubstituted benzene rings-first at δ H 6.98 (br s, H-2), 6 , forming AMX and ABX spin systems, respectively. Proton H-8 correlated in the HSQC spectrum with a carbon at δ C 75.3, while proton H-8 correlated with a carbon at δ C 39.3, suggesting that the first one was an oxygenated methine. The 1 H-NMR spectrum also revealed two sharp singlets, at δ 7.13 (H-7 ) and 3.54 (9-OMe), correlating in the HSQC spectrum with carbons at δ 127.0 and 52.1 (respectively), suggesting the presence of a tri-substituted (Z)-double bond [23] and a methyl ester function. This evidence suggested that 38 was a dicaffeic acid-(3,4-dihydroxyphenyl lactic acid) conjugate. The long-range correlations visible in the HMBC spectrum between H-7 and C-8 (δ C 140.2)/C-9 and H-8 and C-9 /C-9 suggested that the core structure of 38 was rosmarinic acid (27). Furthermore, the long-range correlations observed in the HMBC spectrum between H-8 and carbons C-1, C-5 (δ C 132.8), and C-9 unambiguously proved the presence of a dihydrocaffeic acid methyl ester attached to C-5 and C-8 of the rosmarinic acid, forming a 14-carbon close-ring structure ( Figure 3). It was further supported by the NOE effect observed in the NOESY (nuclear Overhauser effect spectroscopy) spectrum between protons H-2 and H-6 , and substantial broadening of their signals, as it suggested the presence of steric hindrance to free rotation, and their close vicinity. Compound 38 was named pulmitric acid A.  [23] and a methyl ester function. This evidence suggested that 38 was a dicaffeic acid-(3,4dihydroxyphenyl lactic acid) conjugate. The long-range correlations visible in the HMBC spectrum between H-7′′ and C-8′′ (δC 140.2)/C-9′′ and H-8′ and C-9′′/C-9′ suggested that the core structure of 38 was rosmarinic acid (27). Furthermore, the long-range correlations observed in the HMBC spectrum between H-8 and carbons C-1, C-5′ (δC 132.8), and C-9 unambiguously proved the presence of a dihydrocaffeic acid methyl ester attached to C-5′ and C-8′′ of the rosmarinic acid, forming a 14-carbon close-ring structure ( Figure 3). It was further supported by the NOE effect observed in the NOESY (nuclear Overhauser effect spectroscopy) spectrum between protons H-2 and H-6′, and substantial broadening of their signals, as it suggested the presence of steric hindrance to free rotation, and their close vicinity. Compound 38 was named pulmitric acid A. The 13 C-NMR (DEPTQ-135) spectrum of 39 contained 27 signals, sorted by HSQC and HMBC spectra, as one CH2, 12 CH, and 14 quaternary carbon atoms ( Table 3). The NMR spectra suggested the presence of three phenylpropanoid moieties, similarly to 38, with rosmarinic acid being the core structure. One ABX spin system at δH 7.37 (dd, J = 8. Hz, H-8) appeared in the aromatic part of the COSY spectrum. The 13 C-NMR spectrum also showed two aliphatic carbons at δC 74.7 (C-8′), 37.9 (C-7′) and three carboxyl groups at δC 173.4 (C-9′), 168.1 (C-9) and up-field shifted one at 160.2 (C-9′′). Additionally, the 1 H-NMR spectrum contained three sharp singlets at δH 7.03 (H-7′′), 6.81 (H-6′′) and 6.79 (H-3′′). Altogether, compound 39 seemed to possess a structure similar to that of lycopic acid, but with 3,4,5trihydroxycinnamic instead of caffeic acid [24]. The proton H-7′′ showed a correlation in the NOESY The 13 C-NMR (DEPTQ-135) spectrum of 39 contained 27 signals, sorted by HSQC and HMBC spectra, as one CH 2 , 12 CH, and 14 quaternary carbon atoms ( Table 3). The NMR spectra suggested the presence of three phenylpropanoid moieties, similarly to 38, with rosmarinic acid being the core structure. One ABX spin system at δ Hz, H-8) appeared in the aromatic part of the COSY spectrum. The 13 C-NMR spectrum also showed two aliphatic carbons at δ C 74.7 (C-8 ), 37.9 (C-7 ) and three carboxyl groups at δ C 173.4 (C-9 ), 168.1 (C-9) and up-field shifted one at 160.2 (C-9 ). Additionally, the 1 H-NMR spectrum contained three sharp singlets at δ H 7.03 (H-7 ), 6.81 (H-6 ) and 6.79 (H-3 ). Altogether, compound 39 seemed to possess a structure similar to that of lycopic acid, but with 3,4,5-trihydroxycinnamic instead of caffeic acid [24]. The proton H-7 showed a correlation in the NOESY spectrum with H-6 , but quite surprisingly also with H-2, confirming the (E)-stereochemistry of the double bond and the site of conjugation with a rosmarinic acid moiety. It was further supported by the substantial downfield shift of C-2 (∆δ + 6.3) and C-4 (∆δ + 2.7), and the upfield shift of C-3 (∆δ − 2.7), when compared to that of rosmarinic acid (see Supplementary Materials Figure S48). Compound 39 was named pulmitric acid B.
Compounds 40 and 41 were isolated as separated chromatographic peaks using preparative HPLC. However, they presented a set of identical 1 H and 13 C-NMR resonances, in agreement with the structure of isosalvianolic acid A, found in Mentha species [25,26]. Compound 40 was recognized as isosalvianolic acid A (7R,8 R), from the value of its optical rotation ([α] 23 D = +39.7 • ), similar to that of rosmarinic acid. To investigate that in more detail CD (circular dichroism) spectra of 40 were recorded. As the C-8 chiral center and its environment are identical to that of rosmarinic acid, one would expect a similar CD spectrum if the configuration around C-8 is the same as in 27. In fact, the CD spectra were only quantitatively different (see Supplementary Materials Figure S48 and S96). The values of the chemical shifts of compounds 40 and 41 were identical with the literature data [27]. On the other hand, optical rotation ([α] 23 D = +18.5 • ) and the negative Cotton effect observed in the CD spectrum at 240-270 nm, suggested that the absolute stereochemistry of C-7 in 41 was (S). Therefore, compound 41 was named isosalvianolic acid A-1.
The 13 C-NMR of compound 18 showed 27 signals that were classified as one CH 2 , 12 CH, and 14 quaternary carbon atoms ( Table 4). The UV and MS spectral properties, as well as the 1 H and 13 C-NMR chemical shifts, were almost superimposable with that of salvianolic acid R (note the different enumeration of carbons) in [28]. However, the chemical structure of 18 was in agreement with globoidnan B, as presented in [29]. The structures differed in the location of a 3,4-dihydroxyphenyllactic moiety, connected with the core moiety of epiphyllic acid through an ester bond, either with C-9 or C-10 [30]. Our findings suggest that NMR chemical shifts are in agreement with globoidnan B, and are based on the careful study of both HMBC correlations and the analysis of 1,1-ADEQUATE (adequate double-quantum transfer experiment) spectrum and unambiguously identify the site of esterification ( Figure 4). Additionally, we presented the optical rotation and CD spectra of this compound. spectrum with H-6′′, but quite surprisingly also with H-2, confirming the (E)-stereochemistry of the double bond and the site of conjugation with a rosmarinic acid moiety. It was further supported by the substantial downfield shift of C-2 (Δδ + 6.3) and C-4 (Δδ + 2.7), and the upfield shift of C-3 (Δδ − 2.7), when compared to that of rosmarinic acid (see Supplementary Materials Figure S48). Compound 39 was named pulmitric acid B. Compounds 40 and 41 were isolated as separated chromatographic peaks using preparative HPLC. However, they presented a set of identical 1 H and 13 C-NMR resonances, in agreement with the structure of isosalvianolic acid A, found in Mentha species [25,26]. Compound 40 was recognized as isosalvianolic acid A (7R,8′′R), from the value of its optical rotation ([ ] = +39.7°), similar to that of rosmarinic acid. To investigate that in more detail CD (circular dichroism) spectra of 40 were recorded.. As the C-8′′ chiral center and its environment are identical to that of rosmarinic acid, one would expect a similar CD spectrum if the configuration around C-8′′ is the same as in 27. In fact, the CD spectra were only quantitatively different (see Supplementary Materials Figure S48 and S96). The values of the chemical shifts of compounds 40 and 41 were identical with the literature data [27]. On the other hand, optical rotation ([ ] = +18.5°) and the negative Cotton effect observed in the CD spectrum at 240-270 nm, suggested that the absolute stereochemistry of C-7 in 41 was (S). Therefore, compound 41 was named isosalvianolic acid A-1.
The 13 C-NMR of compound 18 showed 27 signals that were classified as one CH2, 12 CH, and 14 quaternary carbon atoms ( Table 4). The UV and MS spectral properties, as well as the 1 H and 13 C-NMR chemical shifts, were almost superimposable with that of salvianolic acid R (note the different enumeration of carbons) in [28]. However, the chemical structure of 18 was in agreement with globoidnan B, as presented in [29]. The structures differed in the location of a 3,4dihydroxyphenyllactic moiety, connected with the core moiety of epiphyllic acid through an ester bond, either with C-9 or C-10 [30]. Our findings suggest that NMR chemical shifts are in agreement with globoidnan B, and are based on the careful study of both HMBC correlations and the analysis of 1,1-ADEQUATE (adequate double-quantum transfer experiment) spectrum and unambiguously identify the site of esterification ( Figure 4). Additionally, we presented the optical rotation and CD spectra of this compound.  The 13 C-NMR of compound 32 showed 47 signals that were classified as three CH 3 , three CH 2 , 26 CH, and 15 quaternary carbon atoms ( Table 4). The aromatic region of the 1 H and COSY spectra of 32 exhibited the presence of two sets of aromatic protons, in accordance with AMX spin systems, at δ , and correlated in the HMBC spectrum with the carbonyl group at δ C (168.9, C-9 ), which suggested the possible site of esterification. The other set was evidenced as part of an 1,2-dihydronaphtalene moiety-similarly to 18, additionally methylated in positions C-6 (δ C 147.7) and C-3 (δ C 149.0), which was further supported by the HMBC correlations and the NOE effects visible in the TROESY (transverse rotating-frame Overhauser enhancement spectroscopy) spectra between 6-OMe (δ H 3.79, s) and H-5 (δ H 6.73, s), and between 3 -OMe (δ H 3.84, s), and H-2 ( Figure 5). This part of the molecule can also be recognized as the 8,8 -diferulic acid, in its aryltetralin form, often present in the plant cell-walls [31]. However, the relative configuration of H-1/H-2 must be different when compared to globoidnan B (gauche). This assumption was based on the large (J = 15.3 Hz) coupling constant between H-1 (δ 4.34, dd, J = 15.3, 1.0 Hz, 1H) and H-2 (δ 4.21, dd, J = 15.3, 2.5 Hz, 1H), and allylic coupling ( 4 J) between H-2 and H-4 (δ 7.39, d, J = 2.5 Hz, 1H), suggesting that orientation of C-2-H-2 is probably quasi-axial [32]. Additionally, the NOE effect observed in the TROESY spectrum between protons H-8 (δ 6.11, s)/H-2 and H-8/H-6 indicated that the orientation of C-1-H-1 is quasi-axial too and therefore the relative configuration of H-1/H-2 is anti [32]. Moreover, two anomeric proton signals at δ H 5.50 (d, J = 1.8 Hz, H-1 ) and 5.35 (d, J = 3.6 Hz, H-1 ) were observed, indicating the presence of two sugar units. Based on the values of coupling constants, and the analysis of 1D TOCSY and 1D TROESY, COSY, HSQC, H2BC, HSQC-TOCSY, F2-coupled HSQC [33], and HMBC data, the two sugar units were elucidated as α-rhamnopyranoside δ H/C 5.50 (H-1 )/100.7 (C-1 ), and α-glucopyranoside δ H/C 5.35 (H-1 )/94.4 (C-1 ) ( Table 4). The α orientation of anomeric protons was supported by their 1 J CH coupling constants, with values of~172 and~169 Hz [34], respectively, measured in the F2-coupled HSQC experiment. Additionally, the 3 J correlations observed in the HMBC spectrum between the anomeric proton of the Glc (H-1 ) and C-2 of a β-fructofuranose unit (δ C 110.1) indicated the presence of interglycosidic linkage between these sugar units (1→2), forming a sucrose unit. This was further supported by the NOE effect detected in the TROESY spectrum between H-1 and CH 2 -1 (δ H 3.90, d, J= 12.4 Hz/3.72, d, J = 12.4 Hz). Careful examination of correlations visible in the HMBC spectrum allowed to assign connectivities between the glycosidic and aromatic parts of 32 unambiguously. Therefore, the correlation between H-1 and C-4 (δ C 147.9) of caffeoyl moiety proved a Rha unit to be the last part of the glycosidic chain. The 1 H and 13 C-NMR spectra of compound 35 showed an almost identical set of atoms, with only one additional CH3 group located at C-3′′′′ (δC 152.0), when compared to 32 (Table 4). In a chain of glycosidation, type of sugar units was precisely the same (with 8,8′-diferulic acid as the core of the molecule), the only difference was the presence of the (E)-ferulic acid instead of caffeoyl moiety. This was supported by the long-range correlation that was visible in the HMBC spectrum of 35 between 3′′′′-OMe (δH 3.92, s) and C-3′′′′, and the NOE effect was visible in the TROESY spectrum between proton mentioned above and H-2′′′′ (δH 7.26, d, J = 2.0 Hz). Hence, the structure of 35 was elucidated, and we proposed its trivial name to be pulmonarioside B.
Compound 30 was isolated as the minor constituent of P. officinalis, and its 13 C-NMR showed 26 signals that were classified as three CH2, 10 CH, and 13 quaternary carbon atoms ( Table 5). The The 1 H and 13 C-NMR spectra of compound 35 showed an almost identical set of atoms, with only one additional CH 3 group located at C-3 (δ C 152.0), when compared to 32 (Table 4). In a chain of glycosidation, type of sugar units was precisely the same (with 8,8 -diferulic acid as the core of the molecule), the only difference was the presence of the (E)-ferulic acid instead of caffeoyl moiety. This was supported by the long-range correlation that was visible in the HMBC spectrum of 35 between 3 -OMe (δ H 3.92, s) and C-3 , and the NOE effect was visible in the TROESY spectrum between proton mentioned above and H-2 (δ H 7.26, d, J = 2.0 Hz). Hence, the structure of 35 was elucidated, and we proposed its trivial name to be pulmonarioside B.
Compound 30 was isolated as the minor constituent of P. officinalis, and its 13 C-NMR showed 26 signals that were classified as three CH 2 , 10 CH, and 13 quaternary carbon atoms ( Table 5). The spectral features of this compound suggested that it possessed both 3,4-dihydroxyphenyllactic acid and 4-biphenylpropionic acid moieties conjugated through an ester linkage. It was evidenced by the presence of three sets of aromatic protons in the 1 H and COSY NMR spectra. The one was in accordance with the ABX spin system at δ Hz) and C-9 (δ C 173.8) presenting the ester linkage, and H-6 to C-1 (δ C 132.4) with H-2 /H-6 to C-1 (δ C 141.2). It was consistent with the NOE correlations that were visible in the TROESY spectrum between pairs H-2/H-6 and H-2 /H-6 , confirming the presence of the biphenyl moiety. One ambiguity arose during the elucidation of the structure of 30-the 13 C-NMR spectrum showed the presence of resonances at δ C 167.8 and 192.0. The latter was attributable to the oxo-group in the C-10 position from the long-range correlation visible in the HMBC spectrum between H-2 and C-10. However, there was no visible correlation to the resonance at 167.8 ppm, and no aldehyde-type proton that was visible in the 1 H-NMR spectrum, but the chemical formula established through HR-MS required the presence of 26 carbons. Therefore, we suggest that the α-oxoacid moiety is present in the structure of 30, located in the carbons C-10 and C-11. This compound showed several similarities to the already known compound-yunnaneic acid E [36], and we proposed its trivial name to be yunnaneic acid E-1.

The Main Phytochemical Constituents of P. officinalis
For quantitative analysis of P. officinalis extracts, we employed a method based on liquid chromatography with high resolution mass spectrometric detection. Metabolites isolated for structure determination and confirmation were used as reference standards, and digoxin was used as the internal standard. Results of these measurements are shown in Table 6.
Rosmarinic acid (27)-a depside composed of (2R)-3-(3,4-dihydroxyphenyl)-lactic and caffeic acid residues, was confirmed as the chief constituent of examined extracts. Although the observed content of rosmarinic acid was relatively high, ranging from 7 to 12 mg/g of dry weight (DW), we previously observed much higher levels (approaching 60 mg/g DW) in similar samples [18]. Reasons for such a broad diversity in the concentration of this compound are unclear. However, comparable levels were reported from other species [37,38]. Rosmarinic acid exerts a variety of well-documented pharmacological properties, such as antioxidant, anti-inflammatory, antibacterial, anti-angiogenic, anti-mutagenic, antidepressant, and neuroprotective, as well as antiallergenic [39][40][41][42][43]. Due to a high content of rosmarinic acid, such properties can also be attributed to whole extracts or tinctures of lungwort. Investigated samples also contained small amounts of methyl rosmarinate (43) [44].
Shimobashiric acid C (26) was another phenolic acid derivative that was abundantly present in the P. officinalis extract (1.2-1.8 mg/g DW). It is a dimer of rosmarinic acid containing cyclobutane scaffold (truxillic acid) formed presumably by a [2 + 2] photocycloaddition of two olefinic moieties [45]. Shimobashiric acid C was isolated from Keiskea japonica [46], and recently from Plectranthus amboinicus [45]. However, its activity has been poorly studied. It can act as a hyaluronidase inhibitor [46] and also possesses anti-inflammatory properties. Chen et al. confirmed that this molecule inhibited the binding of the AP-1 transcription factor to its consensus DNA sequence, and showed TNF-α inhibitory activity as well [45].
Salvianolic acid H (33, 3 -O-(8 -Z-caffeoyl) rosmarinic acid) and its methyl ester (44) were also detected at low levels (0.03-0.26 and 0.006-0.03 mg/g DW, respectively). However, their presence can be, at least partially, attributed to the transformation of another compound, lycopic acid C (45), during the extraction and isolation processes [24]. Treatment of lycopic acid C with water for three days was reported to produce 3-O-(caffeoyl) rosmarinic acid [65], and similar treatment with methanol produced its methyl ester. Nevertheless, the presence of salvianolic acid H was confirmed in some species from genus Salvia, such as Salvia cavaleriei [51], and Salvia miltiorrhiza [48,50].
In addition to rosmarinic acid derivatives, Pulmonaria also appears to have an ability to synthesize chlorogenic acids, as evidenced by the presence of chlorogenic (6, 0.2-0.3 mg/g DW), crypto-chlorogenic (9, 0.005-0.03 mg/g DW), and neo-chlorogenic (13) acid (0.02-0.03 mg/g DW) in the investigated extracts. All these compounds showed the same ion [M − H] − at m/z 353. Compound 6 was identified as 3-O-caffeoylquinic acid, whereas compound 9 was identified as 4-O-caffeoylquinic acid, by comparisons of MS/MS fragmentation patterns and retention times with that of authentic standards. The last isomer was only tentatively identified as a 5-O-caffeoylquinic acid (13) by its fragmentation pattern alone [66,67].
3-O-p-coumaroylquinic acid (15), preliminarily identified using HR-QTOF-MS/MS (high resolution quadrupole-time of flight tandem mass spectrometry) and subsequently isolated and confirmed using NMR techniques, was detected at levels (0.1-0.36 mg/g DW) that were comparable to that of its sister molecule, chlorogenic acid. Isomeric 4-O-p-coumaroylquinic (16) and 5-O-p-coumaroylquinic (17) acids, tentatively identified by the MS/MS fragmentation patterns [66], also occur in Pulmonaria, although the former is a relatively minor component, detected only in samples collected in autumn.
Lignans were abundantly represented by globoidnan B (18, 3.8-6.8 mg/g DW), and a few other components detected at trace levels. That latter group, included also newly discovered pulmonariosides A and B, as well as globoidnan A (0.02 mg/g DW). Nothing is known about biological activities of globoidnan B. However, globoidnan A was previously isolated from Eucalyptus globoidea [69], Origanum minutiflorum [70], and Thymus praecox [71] and extensively investigated. It is known to inhibit the action of HIV integrase, an enzyme which is responsible for the introduction of HIV viral DNA into a host s cellular DNA [69]. Moreover, it revealed anti-proliferative activity against HeLa (human cervix carcinoma) or C6 (rat brain tumor) cell lines [71].
The presence of menisdaurin (2), a nitrile glucoside, was confirmed by extensive spectroscopic HR-MS and NMR analyses and comparison with literature data [72]. The occurrence of this compound in Pulmonaria officinalis was so far unknown. Menisdaurin was detected in a moderate amount, approximately 0.1 mg/g DW, exclusively in the samples collected during spring. Very few members of Boraginaceae family are known to contain nitrile glucosides. Compounds structurally related to menisdaurin glucosides containing cyclohexenylcyanomethylene aglycones, such as lithospermoside (griffonin) and ehretiosides, were reported from various species of Lithospermum and Ehretia. Menisdaurin itself was also reported from genus Tiquilia. However, these plants are only distantly related to Pulmonaria. Although generally considered to be non-cyanogenic, menisdaurin gave positive results for HCN release in the tests carried out by Siegler et al. [72]. Menisdaurin elicits moderate anti-hepatitis B virus activity by inhibiting HBV DNA replication [73,74]. Its anti-tumor [75], as well as anti-inflammatory [75] activity were also reported.

Spring Autumn
Lycopic acid C NA NA TR-traces, indicates level below the limit of quantification; NA-not analyzed; ND-not detected.

Seasonal Fluctuation in Phytochemical Composition of P. officinalis
Numerous reports describe seasonal changes in the contents of flavonoid and phenolic acid derivatives in various plant species [81][82][83][84][85]. Although a variety of different accumulation patterns were observed, it is often stated in the literature that seasonal changes in phenolic acids contents follow a trend that is frequently the opposite to the direction of changes in flavonoid glycosides concentration. To investigate the seasonal variability of flavonoids and phenolic acids in P. officinalis, we first applied the untargeted metabolomics approach.
Univariate volcano plot analysis, combining t-tests and fold change examination, determined several features as linked with spring and autumn stages of Pulmonaria life cycle. As shown in Figure 6, the features on the right side of the volcano plot were the most significant for the spring samples, whereas the elements on the left side are connected with the autumn samples. Among the features characteristic for spring samples, a few compounds isolated and characterized in this study were identified. Chief among them is menisdaurin (2), which nearly completely disappeared in the autumn samples. Similarly, spring samples could be described by high levels of rutin (19), nicotiflorin (24) as well as 3-O-caffeoyl-threonic acid (3). Presence of the latter compound can have some ecological significance, as this type of hydroxycinnamic acid derivative is known as oviposition stimulant for Papilio sp. butterflies [86]. On the other hand, high levels of several other compounds, but mainly salvianolic acid H (33), actinidioionoside (7), three isomers of coumaroylquinic acid (15,16,17), as well as pulmonarioside B (35), were linked to the autumn samples. Numbers indicate compounds isolated and characterized in this study ( Figure 1 and Table 1).
Unsupervised principal component analysis (PCA) indicated a clear separation of samples according to their phenological stage along the first component, encompassing 74.2% of the dataset variability ( Figure 7). Investigation of PCA loadings (Figure 7) revealed that the variables responsible for grouping spring samples were, in large part, virtually the same as these indicated by univariate analyses: rutin (19), menisdaurin (2), O-caffeoyl-L-threonic acids (3,4), and nicotiflorin (24). Additionally, O-caffeoyl glyceric acid (11) and malonyl glucoside of quercetin (23) were also indicated as being characteristic for spring samples. Similarly, autumn samples were again characterized by high amounts of salvianolic acid H, actinidioionoside, and coumaroylquinic acids, but also globoidnan B (18) astragalin (25), shimobashiric acid C (26), and malonyl glucoside of kaempferol (28).  (Figure 7). Investigation of PCA loadings (Figure 7) revealed that the variables responsible for grouping spring samples were, in large part, virtually the same as these indicated by univariate analyses: rutin (19), menisdaurin (2), O-caffeoyl-L-threonic acids (3,4), and nicotiflorin (24). Additionally, O-caffeoyl glyceric acid (11) and malonyl glucoside of quercetin (23) were also indicated as being characteristic for spring samples. Similarly, autumn samples were again characterized by high amounts of salvianolic acid H, actinidioionoside, and coumaroylquinic acids, but also globoidnan B (18) astragalin (25), shimobashiric acid C (26), and malonyl glucoside of kaempferol (28). Multivariate analyses carried out on the dataset derived from targeted measurements of selected 39 metabolites produced very similar results ( Figure 8). Again, clear separation of the two groups of samples can be observed along the first principal component (87.7% variability). Likewise, loadings still indicated nearly the same set of metabolites as responsible for grouping spring and autumn samples ( Figure 8). However, one notable exception is rosmarinic acid (27), which in the untargeted analysis, was characteristic for the spring samples. In the targeted study, it conversely appeared to be marker for autumn samples. The reason for this discrepancy underlays a problem with metabolomics based on peak intensities. Both spring and autumn samples were so abundant in rosmarinic acid that its signal during MS analyses was overloaded. Thus, in both cases, registered intensities of the m/z 359.07 ion had similarly high levels, and even slight variabilities of the signal produces differences that may seem relevant. However, as clearly seen in the volcano plot (Figure 6), where the point representing rosmarinic acid is located in the lower central part, these changes were lesser in the level and have higher p-value. A signal overload can be easily corrected using appropriate dilutions; however, high dilutions generate a risk of loss from the analysis of some other less abundant but also essential components of the sample. The trade-off dilution that we applied-10-fold, was unfortunately not sufficient enough to include the rosmarinic acid in the multivariate PCA model in the untargeted analysis. On the other hand, the targeted analysis was based on a series of three dilutions (2, 10, and 50 times), to allow for proper quantitation of the analyzed compounds within the range of the calibration curves. Multivariate analyses carried out on the dataset derived from targeted measurements of selected 39 metabolites produced very similar results ( Figure 8). Again, clear separation of the two groups of samples can be observed along the first principal component (87.7% variability). Likewise, loadings still indicated nearly the same set of metabolites as responsible for grouping spring and autumn samples ( Figure 8). However, one notable exception is rosmarinic acid (27), which in the untargeted analysis, was characteristic for the spring samples. In the targeted study, it conversely appeared to be marker for autumn samples. The reason for this discrepancy underlays a problem with metabolomics based on peak intensities. Both spring and autumn samples were so abundant in rosmarinic acid that its signal during MS analyses was overloaded. Thus, in both cases, registered intensities of the m/z 359.07 ion had similarly high levels, and even slight variabilities of the signal produces differences that may seem relevant. However, as clearly seen in the volcano plot (Figure 6), where the point representing rosmarinic acid is located in the lower central part, these changes were lesser in the level and have higher p-value. A signal overload can be easily corrected using appropriate dilutions; however, high dilutions generate a risk of loss from the analysis of some other less abundant but also essential components of the sample. The trade-off dilution that we applied-10-fold, was unfortunately not sufficient enough to include the rosmarinic acid in the multivariate PCA model in the untargeted analysis. On the other hand, the targeted analysis was based on a series of three dilutions (2, 10, and 50 times), to allow for proper quantitation of the analyzed compounds within the range of the calibration curves. With the results of the targeted analysis, it was clear that phenolic acids and flavonoids accumulation patterns during the life cycle of Pulmonaria were quite multifaceted ( Table 6, heat map in Figure 9). Flavonol glycosides in other species are known to be at their highest levels during spring, then their concentrations decline, reaching a minimum at the end of the phenological cycle. Such a pattern was observed for unidentified flavonoids in Rosmarinus officinalis, although rutin displayed the opposite trend, accumulating in the highest amount at the vegetative stage of the life cycle [81]. Similarly, analyses of flavonoids in Hypericum perforatum also showed the accumulation of hyperoside and apigenin glucoside in the pre-flowering and flowering stages, followed by a decline during mature fruiting, and a slight increase in the late vegetative stage. Again, rutin accumulated only during the vegetative stage [82]. The opposite trend for rutin was, however, observed in the case of Melittis melissophyllum, where it accumulated in leaves in May, and sharply declined in the same tissues in September [85]. This pattern is mostly the same as observed in the current study for rutin, nicotiflorin, and to a slightly lesser extent, also for malonyl glucoside, as well as for glucosides of quercetin. On the other hand, both malonyl glucoside and a glucoside of kaempferol seemed to accumulate during the vegetative stage. Considering that P. officinalis is one of the early-spring plants that flowers before tree-canopy leaves develop, increased levels of quercetin derivatives (19,21,23) are probably a reflection of high light irradiance that up-regulates their biosynthesis [87,88]. On the contrary, higher levels of kaempferol derivatives at the end of the vegetative cycle, when Pulmonaria plants usually live under dense tree canopies, may reflect an association of these derivatives with shaded growth conditions [88]. However, flavonoid glycosides with B-ring diols, such as quercetin derivatives, were hypothesized to prevent high-intensity light damage by scavenging reactive oxygen species (ROS). UV-induced ROS generation can also be inhibited by compounds that merely absorb relevant light wavelengths, and phenolic acids appear to be much more suitable for this function than flavonoids [89]. This can be a potential function of several caffeic acid derivatives with high concentrations in spring samples that decline in levels during life cycle progression. In particular, O-caffeoyl threonic and glyceric acids, and globoidnan B may serve UV-B protective functions during the spring development of Pulmonaria. Nonetheless, numerous other phenolic acids derivatives, including the main metabolite-rosmarinic acid, appear to accumulate in aerial parts in autumn. An increase in the levels of simple compounds such as caffeic acid (8) and danshensu (1), can be a result of the turnover and degradation of other derived composite metabolites. However, yunnaneic acid B, a relatively complex molecule, also accumulates nearly 9-fold compared to spring samples. The role of these metabolites at the vegetative stage of Pulmonaria life cycle is yet to be explained. With the results of the targeted analysis, it was clear that phenolic acids and flavonoids accumulation patterns during the life cycle of Pulmonaria were quite multifaceted ( Table 6, heat map in Figure 9). Flavonol glycosides in other species are known to be at their highest levels during spring, then their concentrations decline, reaching a minimum at the end of the phenological cycle. Such a pattern was observed for unidentified flavonoids in Rosmarinus officinalis, although rutin displayed the opposite trend, accumulating in the highest amount at the vegetative stage of the life cycle [81]. Similarly, analyses of flavonoids in Hypericum perforatum also showed the accumulation of hyperoside and apigenin glucoside in the pre-flowering and flowering stages, followed by a decline during mature fruiting, and a slight increase in the late vegetative stage. Again, rutin accumulated only during the vegetative stage [82]. The opposite trend for rutin was, however, observed in the case of Melittis melissophyllum, where it accumulated in leaves in May, and sharply declined in the same tissues in September [85]. This pattern is mostly the same as observed in the current study for rutin, nicotiflorin, and to a slightly lesser extent, also for malonyl glucoside, as well as for glucosides of quercetin. On the other hand, both malonyl glucoside and a glucoside of kaempferol seemed to accumulate during the vegetative stage. Considering that P. officinalis is one of the early-spring plants that flowers before tree-canopy leaves develop, increased levels of quercetin derivatives (19,21,23) are probably a reflection of high light irradiance that up-regulates their biosynthesis [87,88]. On the contrary, higher levels of kaempferol derivatives at the end of the vegetative cycle, when Pulmonaria plants usually live under dense tree canopies, may reflect an association of these derivatives with shaded growth conditions [88]. However, flavonoid glycosides with B-ring diols, such as quercetin derivatives, were hypothesized to prevent high-intensity light damage by scavenging reactive oxygen species (ROS). UV-induced ROS generation can also be inhibited by compounds that merely absorb relevant light wavelengths, and phenolic acids appear to be much more suitable for this function than flavonoids [89]. This can be a potential function of several caffeic acid derivatives with high concentrations in spring samples that decline in levels during life cycle progression. In particular, O-caffeoyl threonic and glyceric acids, and globoidnan B may serve UV-B protective functions during the spring development of Pulmonaria. Nonetheless, numerous other phenolic acids derivatives, including the main metabolite-rosmarinic acid, appear to accumulate in aerial parts in autumn. An increase in the levels of simple compounds such as caffeic acid (8) and danshensu (1), can be a result of the turnover and degradation of other derived composite metabolites. However, yunnaneic acid B, a relatively complex molecule, also accumulates nearly 9-fold compared to spring samples. The role of these metabolites at the vegetative stage of Pulmonaria life cycle is yet to be explained. . Result of the aggregative hierarchical clustering (Euclidian distance measure, Ward′s clustering algorithm) of Pulmonaria officinalis samples and metabolites, shown as a heatmap. Numbers correspond to compounds isolated in this study as shown in Figure 1 and Table 1, heatmap colors represent the relative concentrations in the samples from high (red) to low (blue).
There is still need to explore the usefulness of many of traditionally used herbs for modern therapy, pharmacy, pharmacology, or medicine. Presented research explores the phytochemical composition of commonly occurring and well known medicinal plant that nowadays is not intensively used for therapeutic purposes. For the first time, specialized metabolites of Pulmonaria, especially phenolic compounds, were thoroughly investigated by a combination of NMR, HR-MS, and other spectral techniques, providing full-scale qualitative and quantitative information. The results of this study will contribute to qualitative and quantitative method development and quality control or standardization purposes. Figure 9. Result of the aggregative hierarchical clustering (Euclidian distance measure, Ward s clustering algorithm) of Pulmonaria officinalis samples and metabolites, shown as a heatmap. Numbers correspond to compounds isolated in this study as shown in Figure 1 and Table 1, heatmap colors represent the relative concentrations in the samples from high (red) to low (blue).
There is still need to explore the usefulness of many of traditionally used herbs for modern therapy, pharmacy, pharmacology, or medicine. Presented research explores the phytochemical composition of commonly occurring and well known medicinal plant that nowadays is not intensively used for therapeutic purposes. For the first time, specialized metabolites of Pulmonaria, especially phenolic compounds, were thoroughly investigated by a combination of NMR, HR-MS, and other spectral techniques, providing full-scale qualitative and quantitative information. The results of this study will contribute to qualitative and quantitative method development and quality control or standardization purposes.

Chemicals and Reagents
Acetonitrile and methanol, LC-MS grade and HPLC grade respectively, were purchased from Merck (Darmstadt, Germany). MS-grade formic acid was purchased from Sigma Aldrich (Steinheim, Germany). Ultrapure water was prepared using a Milli-Q water purification system (Millipore, Milford, MA, USA).

Plant Material
Dried aerial parts of P. officinalis used for compound isolation were purchased from a local herb supplier (Kania, Czestochowa, Poland). The voucher sample (POFF./EXTR/2013/1) has been deposited at the Department of Biochemistry and Crop Quality of the Institute. Quantitative analyses of isolated compounds during the growing season have been carried out for P. officinalis L., and rhizomes were donated from The Botanical Garden of Maria Curie-Skłodowska University in Lublin. Rhizomes were planted at the experimental plots of the Institute of Soil Science and Plant Cultivation, Puławy, Poland (N51 • 24.767 E021 • 57.924 ). The aerial parts of lungwort were collected from early spring to late autumn at weekly intervals. The meteorological conditions for April and September are presented in Table 7. The voucher samples have also been deposited at the Department of Biochemistry and Crop Quality of the Institute.

Extraction and Isolation
Aerial parts of P. officinalis were finely powdered with an electric grinder and sieved through a 0.5 mm sieve. Plant material was defatted with chloroform in a Soxhlet apparatus. Afterward, it was extracted twice with 80% methanol (v/v) using ultrasonic bath, at room temperature for 24 h. The obtained extract was filtered through filter paper (Whatmann No.1), concentrated under reduced pressure/vacuum, and freeze-dried. The yield of the extraction was 24.4%.
The crude methanol extract was then purified in a stepwise manner by different chromatographic methods. First, the extract was applied to a preconditioned RP-C 18 column (80 × 100 mm, Cosmosil 140C 18 -PREP, 140 µm; Nacalai Tesque, INC., Kyoto, Japan), followed by removal of polar constituents (1% MeOH v/v), while phenolic-rich fraction was eluted with 50% methanol. Collected fractions were monitored using thin layer chromatography (TLC) techniques. TLC was performed on silica gel plates with a solvent system consisting of MeCN:H 2 O:CHCl 3 :HCOOH (100:10:10:5) as a mobile phase. The plates were visualized under UV light at 254/360 nm, and then sprayed with a methanol/sulfuric acid reagent and heated on a hot plate. The 50% methanol fraction (32.8 g) was further purified by low-pressure chromatography on Sephadex LH-20 (Sigma-Aldrich, Steinheim, Germany) column (48 × 400 mm) and eluted with a gradient of MeOH (5-100%, v/v). As a result of this separation, 10 fractions (Fr 0-9) were collected. Due to the high similarity of their composition, fractions 4 and 5, as well as 6 and 7, were combined with each other. The fractions were subsequently subjected to a reversed phase column (32 × 300 mm, Cosmosil 40C 18 -PREP, 40 µm; Nacalai Tesque, INC., Kyoto, Japan), which yielded several sub-fractions. The compositions of the fraction and sub-fractions were monitored by LC-MS technique as described below. Individual compounds were further purified by a semi-preparative HPLC chromatographic system.

Semi-Preparative HPLC
The final purification steps utilized the semi-preparative HPLC Gilson chromatographic system (Gilson Inc., Middleton, WI, USA), equipped with an evaporative light scattering detector (ESLD, Gilson PrepELS II). The drift tube of the ELSD detector was maintained at 65 • C, and the pressure of the nebulizer gas (nitrogen) was 47 psi. Sub-fractions obtained from low-pressure reversed phase liquid chromatography were further purified on a variety of columns: Atlantis T3 Prep OBD (10 × 250 mm, 5 µm, Waters, Milford, MA, USA), COSMOSIL π-NAP (10 × 250 mm, 5 µm, Nacalai Tesque, INC., Kyoto, Japan) or RP-18 Kromasil (10 × 250mm, 5 µm, AkzoNobel, Bohus, Sweden). The conditions of chromatographic separation were individually optimized for each fraction. Separations were carried out in isocratic or in gradient mode, using aqueous acetonitrile or methanol solutions, containing 0.1% formic acid. The mobile phase flow rate was from 3 to 4 mL/min, and the column was held between 35 and 55 • C. The effluent from the HPLC system was diverted through a passive splitter to ELSD with a split ratio of 1:100.

High-Resolution LC-MS and Qualitative Analysis
For quantitative analyses, harvested plants were freeze-dried, finely powdered, and used for extraction with an automated accelerated solvent extractor, ASE 200 (Dionex, Sunnyvale, CA, USA). One hundred milligrams of each sample was extracted three times with 80% aqueous MeOH (three static cycles, 5 min each), at 1500 psi (10.3 MPa) solvent pressure, 100 • C temperature of extraction cells, flush 150%. Obtained crude extracts were evaporated to dryness, dissolved in 1 mL of Milli-Q water (Millipore Corp., Billerica, MA, USA), containing 25 µg/mL of digoxin (internal standard, IS) and purified by solid phase extraction (SPE) using Oasis HLB columns (500 mg, Waters Corp., Milford, MA, USA). The extracts were loaded on preconditioned cartridges, which were washed with 0.5% MeOH to remove unbound material, and then with 85% MeOH to elute specialized metabolites. These fractions were evaporated and dissolved in 1 mL of 85% MeOH acidified with 0.1% HCOOH. All analyses were performed in triplicate and samples were stored in a freezer at −20 • C before analysis. Before spectrometric analyses samples were centrifuged (15 min, 23,000× g) and appropriately diluted with distilled water.
High-resolution LC-MS analyses, e.g.,: exact masses, MS/MS fragmentation patterns, molecular formulae, as well as quantitative determinations, were performed on a Thermo Scientific Ultimate 3000 RS chromatographic system coupled/hyphenated with a Bruker Impact II HD (Bruker, Billerica, MA, USA) quadrupole time-of-flight (Q-TOF) mass spectrometer.
Chromatographic separations were carried out on a Waters CORTECS T3 column (2.1 × 150 mm, 2.7 µm, Milford, MA USA) equipped with pre-column. The mobile phase A was 0.1% (v/v) formic acid, and the mobile phase B was acetonitrile containing 0.1% (v/v) of formic acid. A concave-shaped gradient (Dionex gradient curve nr. 6) from 5% to 60% of phase B over 25 min was used for separation. The flow rate was 0.6 mL/min, and the column was held at 35 • C. Between the injections column was equilibrated with 10 volumes of 5% phase B. Injection volume was 5 µL.
Analyses were carried out in both positive and negative ion mode with electrospray ionization. Measurements in the negative ion mode were accompanied with detection based on charged aerosol (CAD). A flow splitter was used to divert the column effluent in 1:3 proportion between Q-TOF MS and charged aerosol detector connected in parallel. CAD acquisition frequency was 10 Hz. Analyses in the positive ion mode employed additional UV absorbance detection in the 200-600 nm wavelength range with 5 nm bandwidth and 10 Hz acquisition frequency.
Linear (centroid) mass spectra were acquired over a mass range from m/z 50 to m/z 2000 with the following parameters of mass spectrometer: positive ion capillary voltage 4.5 kV; negative ion capillary voltage 3.0 kV, dry gas flow 6 L/min; dry gas temperature 200 • C; nebulizer pressure 0.7 bar; collision RF 700.0 V; transfer time 90 µs; prepulse storage 7.0 µs. Two precursor ions with intensities over 2000 counts were fragmented in each scan. The collision energy and the ion isolation width were set automatically depending on the m/z of the fragmented ion, in the range of 5 to 100 eV and from 3 to 8 mass units, respectively. The acquired data were calibrated internally with sodium formate introduced to the ion source via a 20 µL loop at the beginning of each separation.
Calibration curves in the range from 0.05 to 50 µg/mL were prepared from the 1 mg/mL methanolic stock solutions of investigated compounds and analyzed in the conditions specified above. Extracted ion chromatograms were made from full scan data with a 0.005 Da width.
Smoothing using the Savitzky-Golay algorithm (window width 5 points, one iteration) was applied and peaks corresponding to deprotonated molecules (or, in the case of compounds 2,7 and the internal standard, deprotonated formic acid adducts) were integrated. Ratios between the analyte peak area and the IS peak area were used for calculations. Details of calibrations are shown in Table S1 (see Supplementary Materials). Data processing was performed using Bruker DataAnalysis 4.4 SR1 software. All quantitative results were calculated per dry weight (DW).

Untargeted Metabolomics Analyses.
For untargeted metabolomic analyses, data obtained from LC-QTOF-MS/MS runs were processed using Find Molecular Features function of Bruker DataAnalysis ver. 4.4 SR1 software. The following parameters were used: a signal-to-noise threshold of 9, correlation coefficient threshold of 0.7, and the minimum compound length of seven spectra. For each detected compound, in addition to the deprotonated ion, all typical adduct and composite ions were grouped into a single feature. Next, features from the entire dataset were subjected to advanced bucket generation with Bruker ProfileAnalysis ver. 2.3 software, using retention time range 1.2-22.0 min and m/z range 50-1500. Time alignment was also applied. Resulting data matrix, consisting of intensity values for 283 peaks indexed by feature parameters (m/z, retention time) and sample name (12 samples), was exported and uploaded to MetaboAnalyst [90]. Data were normalized using internal standard (by the peak intensity of a feature containing m/z 825.428, corresponding to [M + HCOOH-H] − ion of digoxin at RT 17.48 min). At this stage of data processing, standard t-test and fold-change analyses were carried out to provide a preliminary overview of features potentially characteristic for the two phenological stages under study. Next, all missing intensity values and intensities that were equal to zero in the data matrix were replaced by the half of the minimum positive value that was found within the data, and multivariate PCA was applied to investigate systematic variation in the data matrix and to identify potential groups in an unsupervised manner.
The bias related to instrumental drift was minimized by randomization of the sample list injection order. It was achieved using an in-house-developed VBA (Visual Basic for Applications) script in Microsoft Excel.
The performance of the LC-MS and the data processing systems were monitored with two types of quality control (QC) samples. Class-specific QC samples (eight in total, four for each group) were prepared by mixing 10 µL aliquots of each sample within the group and diluting them 10 times, just like normal samples. Class QC samples were analyzed after every six analyses of normal samples. During data processing, features detected in less than 75% of class-specific QC samples were removed from the dataset. The frequency distribution of the relative standard deviation for the peak intensities and peak numbers in the QC samples are given in Figure S107 (see Supplementary Materials). The general QC sample consisted of a 8 µg/mL mixture of each of the 38 compounds from P. officinalis that had been isolated in the course of this study, and 25 µg/mL of the internal standard. This QC sample was used to monitor the quality and stability of the data acquisition and was analyzed in replicate after each block of 20 analyses.

Optical Rotation [α]
The optical rotation of isolated compounds was measured on an automatic polarimeter (P-2000, JASCO, Tokyo, Japan).

Circular Dichroism Spectroscopy
Circular dichroism measurements of optically active compounds were carried out on a J-815 circular dichroism spectrometer (JASCO, Tokyo, Japan), using a quartz cell of 1 cm path length. Spectra of analyzed molecules were recorded at 21 • C from 220 to 400 nm at a 0.2 nm resolution, with a scan rate of 50 nm/min. Raw data were smoothed using the Savitzky-Golay method with a window of 11 data points. CD spectra are presented in the Supplementary Materials.

Characteristic Data of Lungwort Compounds
HR-QTOF-MS/MS data in negative ion mode for all compounds has been shown in Table 1 (Table 3).  (Table 3).

Conclusions
To the best of our knowledge, this is the first comprehensive study of specialized metabolites in the aerial parts of Pulmonaria officinalis. The presented research may provide insights for the potential applications of lungwort as a dietary supplement or a nutraceutical, and may it also contribute to the broader application of Pulmonariae Herba. Extracts of P. officinalis may serve as a prominent supply of rosmarinic acid and related compounds, as well as a source of several others metabolites. Among the 45 identified metabolites, we found many compounds with well-established therapeutic properties, although none of them alone can be directly associated with the ethnomedicinal use of lungwort. Our results also show progressive changes in the phytochemical composition of P. officinalis during the phenological cycle, presumably reflecting both changes in the physiological state of plants, as well as varying intensity of different abiotic factors.