Comparison of glucosinolate diversity in the crucifer tribe Cardamineae and the remaining order Brassicales highlights repetitive evolutionary loss and gain of biosynthetic steps Author Agerbirk, Niels * & Department of Plant and Environmental Sciences, University of Copenhagen, Thorvaldsensvej 40, 1871, Frederiksberg C, Denmark Author Hansen, Cecilie Cetti Author Kiefer, Christiane Author Hauser, Thure P. Author Ørgaard, Marian Author Lange, Conny Bruun Asmussen Author Cipollini, Don Author Koch, Marcus A. text Phytochemistry 2021 112668 2021-05-31 185 1 26 http://dx.doi.org/10.1016/j.phytochem.2021.112668 journal article 10.1016/j.phytochem.2021.112668 1873-3700 8259764 2.1. Updated phylogeny of the tribe Cardamineae A molecular phylogeny based on ITS sequences was obtained for 171 taxa out of the total 386 species (44%) in the tribe ( Koch et al., 2017 ). The phylogeny (Supplementary Fig. S1 ) generally agrees with an earlier phylogeny based on fewer species ( Olsen et al., 2016 ). The genus Aplanodes is resolved as sister to all other species of the tribe included in the phylogeny. However, there is either low or no bootstrap support for the backbone of the phylogeny, which might indicate rapid radiation, conflicting phylogenetic signal due to reticulate evolution (high number of polyploids may be seen as an indicator) or a lack of sequence information due to a limited number of single nucleotide polymorphisms (SNPs) analyzed. We tentatively conclude that the genus Cardamine in its current taxonomic circumscription is not monophyletic and that most other genera or larger alliances are resolved as monophyletic with high bootstrap support. These genera or larger clades are Leavenworthia / Selenia (87% bootstrap support), the C. pratensis alliance (97% bootstrap support), the genus Barbarea (100% support) including two chemotypes of B. vulgaris (Agerbirk et al., 2021) , and Nasturtium (100% support). The C. pratensis alliance and the genus Barbarea are a focus of this work because of their deviating GSL profiles, but species in less well-resolved and less well-supported positions in the tree are also considered. Non- Cardamine genera are indicated with grey boxes, and species discussed in some detail are indicated in bold and are marked with grey boxes, too (Supplementary Fig. S1 ). 2.2. Glucosinolate diversity in the tribe Cardamineae 2.2.1. Critical evaluation of literature We previously reviewed GSL profiles of all members of the tribe Cardamineae ( Olsen et al., 2016 ) . Although that review provided a detailed overview of the known diversity, several genera and multiple species were not suitably investigated. In an accompanying paper (Agerbirk et al., 2021), we tested GSL occurrences considered questionable in the previous review, and specifically tested several cases of “not analyzed” for cases where presence or absence of a particular biosynthetic group had never been reliably tested for an entire genus. In the accompanying paper, we provide evidence for Trp-derived GSLs and various biosynthetic groups in Nasturtium , Rorippa and Planodes , conclusive evidence for several questioned GSLs from Armoracia , lack of thioGlc-acylated GSLs in the tribe apart from Barbarea , lack of aliphatic GSLs in two types of B. vulgaris , and lack of Phe-derived GSLs (not chain elongated) in Nasturtium . Including the accompanying paper, we brought our previous literature review up to date. For this purpose, we inspected relevant papers (until around 2019) that were published since the former review (and a single paper from 1983 that had been overlooked), and compiled GSL diversity by species ( Louda and Rodman, 1983 ; Giallourou et al., 2016 ; Pellissier et al., 2016 ; Dekic´et al., 2017; Ciska et al., 2017 ; Jeon et al., 2017 ; Bakhtiari et al., 2018, 2019; Liang et al., 2018 ; Cuong et al., 2019 ; Okamura et al., 2019 ; Badenes-P´erez et al., 2020; Montaut et al., 2020b ). A recent claim without evidense of identification ( Hussain et al., 2020 ), concerning proposed aliphatic GSLs in B. vulgaris , was ignored for reasons detailed in the accompanying paper (Agerbirk et al., 2021). We also re-checked and occasionally revised our scoring of GSLs in older references as either conclusive, tentative or not sufficiently evidenced (Supplementary Table S1A ). In total, the newly included papers concerned horseradish ( A. rusticana ), multiple Cardamine spp. , watercress ( N. officinale ), Rorippa sarmentosa and an erratum on Rorippa austriaca . Poorly evidenced suggestions of GSLs never convincingly characterized from any plant were ignored, relying on the recent compilation by Blaˇzevi´c et al. (2020). As done previously, we judged the reports based on the criterion that conclusive compound identification should be based on use of reference compounds, validated reference materials or NMR structure elucidation. We also required suitably specific peak detection by means of an MS detector (or a diode array detector in case of some Trp-derived GSLs), and when relevant one additional spectral characteristic, either complete MS (typically MS2) or a characteristic UV spectrum (in case of Trp-derived GSLs only). In the case of indirect identification by GC-MS of GSL products, we also regarded identification by reliable retention index (RI) compared to databases as proof of identity because of the general power of GC-MS as an identification tool when using both RI and MS spectral comparison. Reports that did not meet the criteria were registered as tentative by putting the relevant GSLs in parentheses in the table. For our own recent results (Agerbirk et al., 2021), registration as tentative typically resulted from lack of an authentic standard but reasonable t R compared to homologs, or agreement of t R and m / z with an authentic standard but with too low levels to check MS2. However, criteria for registration as tentative or conclusive could not always be as strict as for our own results from inspection of the general literature, because of lack of details in reports. Detailed comments for individual reports are given in Supplementary Table S1 , which also contains a supplement with examples of our arguments. In the majority of recent papers, comparison with previous studies of the same species were notsystematic or frequently missing completely, possibly due to the difficulty in locating previous studies in standard literature search (Section 4.2.). This was the case for 3moBZ suggested from C. cordifolia with faint evidence ( Humphrey et al., 2016 ), without discussion of a previous suggestion with moderate evidence of an isomer (BAR or EBAR) in the same species ( Louda and Rodman, 1983 ). Neither suggestion was accepted. Similarly, the analytical basis of surprising structures was typically not elaborated on or discussed, making it difficult to judge whether several surprising structures should be accepted as existing in the tribe. This was the case for, e.g., unsaturated Met-derived GSLs such as 4mSbuen and 4mSObuen reported by two inconclusive studies ( Pellissier et al., 2016 ; Ciska et al., 2017 ) and suggested 5BzOp and S2hPeen ( Ciska et al., 2017 ; Bakhtiari et al., 2018). Furthermore, 2-3homoMet and 5homoMet derived GSLs were reported from Cardamine amara by Pellissier et al. (2016) based on HPLC-UV Table 2 Diversity of well-established glucosinolates (GSLs) in the tribe Cardamineae , ordered by genus and by amino acid precursor (except the category of glucose 6 ′ -acylated ( “ 6 ′ - Acyl ” ) that is collected for all precursors ) .

Well-identified (and suggested) glucosinolates by apparent amino acid precursor or substitution (and alkyl chain length for n-homoMet-derived GSLs).
Genus	1-3HomoMet	4-6HomoMet	7-8HomoMet	Trp	Phe	HomoPhe	Val	Leu	Ile	1-2HomoIle	6′-
(inv./acc.) a	(3–5C)	(6–8C)	(9–10C)	/2–4homoPhe	Acyl
Armoracia b	12 , 73 , 94 , 95 , 101 ,	66 , 67 , 69 , 88	n.d.	28 , 43 , 48	11	105	( 56 )	62	61	( 54 )	NA
(1/3 = 33%)	107 ( 82 , 84 )	( 87 )	2-4homoPhe: [104] , [106] , [156]
Barbarea c (9/29 = 31%)	n.d.	n.d.	n.d.	28, 43 , 47 , 48 , 138	n.d.	40 S , 40 R , 105 , 139 S , 139 R , 140 , 142 R , 148	n.d.	n.d.	n.d.	n.d.	129 , 130 , 131 S , 131 R , 132 R
Cardamine d (22/239 = 9%)	12 , 24 R , 64 , 72 , 84 , 94 , 101 ( 24 S , 38 S , 107 )	66 , 69 , 92 ( 67 , 87)	n.d.	28, 43 , 47 , 48, 138	11 , 23 , 46	105	56 , 57	31 , 62	30 , 61	29 , 54 , 58 , 141 , 149	n.d.
Leavenworthia (2/8 = 25%)	101	n.d.	n.d.	NA	n.d.	105	57	n.d.	n.d.	–	NA
Nasturtium e (2/5 = 40%)	12 , 72 ( 24 R , 64 , 73 , 84 , 101 )	66 , 67 , 69 , 80 , 87 , 92 ( 88 )	n.d.	28 , 43 , 48	n.d.	40 S , 40 R , 105	n.d.	n.d.	n.d.	n.d.	n.d.
Planodes f (1/2 = 50%)	64 , 72 ( 94 )	66 , 67 , 69 , 87 , 92 ( 80 , 88 )	( 68 , [89] )	43	n.d.	40 S , 40 R , 105 , 139 R	n.d.	n.d.	n.d.	n.d.	n.d.
Rorippa (9/87 = 10%)	12 , 24R , 64 , 72 , 101 , 107	66 , 67 , 69 , 80 87, 92	68 , 77 , 79 , [89] ( 65 )	28, 43 , 47 , 48, 138	n.d.	105	n.d.	62	61	n.d.	n.d.
Selenia (1/5 = 20%)	n.d.	n.d.	n.d.	–	11	40 , 105	n.d.	n.d.	n.d.	–	NA

8 only, without discussion of a previous report using HPLC-MS (Windsor et al., 2005) that reported absence of n-homoMet-derived GSLs in this species. Lack of discussion was even the case for an unexpected GSL structure (“veratryl GSL” = “phenylvinyl GSL”) (Bakhtiari et al., 2018) that had never been suggested from nature before and might be an artifact (a loss of water MS-fragment of BAR or EBAR). Similarly, many incomprehensive names had to be ignored (“trimethoxy GSL”, “hydroxybenzyl-methylether GSL”) as well as some with imprecise names (“hydroxypropyl GSL”, “butyl GSL”, “hydroxymethylbutyl GSL?”, “2-hydroxymethylpropyl GSL”) (Bakhtiari et al., 2018), the latter apparently supposed to be a known GSL although no known GSL could be named in this way. Several additional reports needed discussion before GSL profiles could be listed as either conclusive or tentative (Supplementary Table S1A ). This general tendency for lack of precision, lack of comparison with the previous literature and lack of discussion of analytical uncertainty is in contrast to solid, classic work such as Yamane et al. (1992) using NMR for conclusive peak identification in ecological work, and work by HPLC-MS pioneers ( Griffiths et al., 2001 ), who explicitly concluded that suggested structures based on HPLC-MS/MS without authentic standards are tentative. Despite the mentioned uncertainty and need for critical evaluation, almost all considered papers, including those criticized for details above, mainly fit into the general picture for the relevant genus or species. All in all, by systematically applying strict analytical chemistry principles as well as conservative judgement in cases of unclear reports, we classified each reported GSL as either conclusively identified, tentatively identified (in parentheses) or insufficiently evidenced (labelled in red print) (Supplementary Table S1A ).