New Discorhabdin Alkaloids from the Antarctic Deep-Sea Sponge Latrunculia biformis

The sponge genus Latrunculia is a prolific source of discorhabdin type pyrroloiminoquinone alkaloids. In the continuation of our research interest into this genus, we studied the Antarctic deep-sea sponge Latrunculia biformis that showed potent in vitro anticancer activity. A targeted isolation process guided by bioactivity and molecular networking-based metabolomics yielded three known discorhabdins, (−)-discorhabdin L (1), (+)-discorhabdin A (2), (+)-discorhabdin Q (3), and three new discorhabdin analogs (−)-2-bromo-discorhabdin D (4), (−)-1-acetyl-discorhabdin L (5), and (+)-1-octacosatrienoyl-discorhabdin L (6) from the MeOH-soluble portion of the organic extract. The chemical structures of 1–6 were elucidated by extensive NMR, HR-ESIMS, FT-IR, [α]D, and ECD (Electronic Circular Dichroism) spectroscopy analyses. Compounds 1, 5, and 6 showed promising anticancer activity with IC50 values of 0.94, 2.71, and 34.0 µM, respectively. Compounds 1–6 and the enantiomer of 1 ((+)-discorhabdin L, 1e) were docked to the active sites of two anticancer targets, topoisomerase I-II and indoleamine 2,3-dioxygenase (IDO1), to reveal, for the first time, the binding potential of discorhabdins to these proteins. Compounds 5 and 6 are the first discorhabdin analogs with an ester function at C-1 and 6 is the first discorhabdin bearing a long-chain fatty acid at this position. This study confirms Latrunculia sponges to be excellent sources of chemically diverse discorhabdin alkaloids.


Introduction
Latrunculia species are cold-adapted sponges commonly found in the coastlines of the southern hemisphere [1][2][3]. The genus Latrunculia has proven to be a prolific source of structurally intriguing compounds from different classes, such as norsesterterpenes [4,5], callipeltins [6,7], and various types of pyrroloiminoquinone alkaloids [8][9][10]. Discorhabdins represent a large and unique subclass of pyrroloiminoquinone alkaloids that have been associated with the chemical defense and greenish to brownish coloration of the sponge [11,12]. Discorhabdins exhibit strong anticancer activity against many cancer types, such as human colon cancer, adenocarcinoma, and leukemia [13][14][15]. However, the mechanism of their anticancer action has been poorly studied. Indeed, only the farnesyltransferase enzyme [16] and hypoxia-inducible factor 1α (HIF-1α) and transcriptional coactivator p300 interaction [17] have been shown as potential targets of discorhabdins.
As part of our research interest into deep-sea Latrunculia sponges from Antarctica [18], herein we investigated the in-depth chemistry of Latrunculia biformis, which was collected from the Antarctic Weddell Sea shelf at 291 m depth. The crude organic extract of the sponge exhibited significant in vitro anticancer activity against six cancer cell lines. A molecular networking (MN)-based metabolomics study on fractions obtained from the MeOH-soluble portion of the sponge indicated the presence of a large discorhabdin cluster with many nodes belonging to potentially new discorhabdins. Guided by anticancer activity and MN-based dereplication, six discorhabdin-type alkaloids were isolated from the MeOH subextract, including three known compounds (−)-discorhabdin L (1), (+)-discorhabdin A (2), (+)-discorhabdin Q (3) and three new discorhabdin derivatives, namely (−)-2-bromo-discorhabdin D (4), (−)-1-acetyl-discorhabdin L (5), and (+)-1-octacosatrienoyl-discorhabdin L (6). Since the amounts of the isolated compounds were very minor, only compounds 1, 5 and 6 could be tested for their anticancer activity against the human colon cancer cell line HCT-116. We applied a structure-based docking approach on all isolated compounds and the enantiomer of 1 (1e) against two cancer targets reported for pyrroloiminoquinone alkaloids, i.e., topoisomerase I-II and indoleamine 2,3-dioxygenase IDO1 [19][20][21] to predict their anticancer potential and to suggest potential molecular mechanism(s) of action. This study reports MN and bioactivity-guided isolation of compounds 1-6, their structure elucidation, and biological activities with potential target identification for their anticancer activity.

Bioactivity and Molecular Networking-guided Purification and Structural Elucidation
The olive green-colored sponge material was freeze-dried and successively extracted with water, MeOH, and dichloromethane (DCM) subsequently. The combined organic extract was submitted to bioactivity screening against six cancer cell lines, where it showed significant activity with IC 50 values ranging from 4.0 to 56.2 µg/mL (Table 1). The solvent partitioning of the crude organic extract between MeOH and n-hexane yielded the MeOH (M) and the n-hexane subextracts. The M subextract demonstrated strong anticancer activity ( Figure 1) and was further fractionated over a C18 SPE cartridge. The anticancer activity was tracked to six SPE fractions, M2-M5, M7, and M8 ( Figure 1). In order to prioritize the isolation workflow towards undescribed molecules with potential anticancer properties, we acquired tandem UPLC-QToF-MS/MS (positive-ion mode) data on these six active fractions. The generated MS/MS (MS 2 ) data were uploaded to the publicly available Global Natural Product Social Molecular Networking (GNPS) platform (http://gnps.ucsd.edu) and analyzed following the molecular networking (MN) online workflow [22]. Software Cytoscape (Version 3.61) was used to visualize the resulting networks. The automated dereplication on GNPS platform did not annotate any pyrroloiminoquinone alkaloids. Hence, compound annotation was based on manual dereplication by comparing the predicted molecular formulae against multiple public or commercially available databases. After a comprehensive examination of the global MN of the SPE fractions, two clusters attracted our attention ( Figure 2). Cluster 1 contained five nodes, four of which were annotated as known molecules discorhabdin L [13], its analog, discorhabdin D [23], and 1-methoxydiscorhabdin D [19], leaving the node at m/z 368.0380 to be a putatively new derivative ( Figure 2). From this cluster, we were able to purify (−)-discorhabdin L (1), but failed to purify the potentially new discorhabdin analog (m/z 368.0380) due to its very minor quantity.
With 21 nodes, cluster 2 was the biggest in the generated MN, which can be further divided into three subclusters ( Figure 2). Based on the elemental composition analysis, MS/MS fragmentation patterns, and biological source, two brominated alkaloids discorhabdin A [24] and discorhabdin G [25] were identified in subcluster 3. However, only discorhabdin A (2) was isolated in sufficient amounts for NMR and other spectroscopic analyses. Discorhabdin H2 [26] was the only annotated compound in subcluster 2. Unfortunately, neither this compound nor the remaining nodes that represent potentially new discorhabdin analogs could be purified in sufficient quantity.  After a comprehensive examination of the global MN of the SPE fractions, two clusters attracted our attention ( Figure 2). Cluster 1 contained five nodes, four of which were annotated as known molecules discorhabdin L [13], its analog, discorhabdin D [23], and 1-methoxydiscorhabdin D [19], leaving the node at m/z 368.0380 to be a putatively new derivative ( Figure 2). From this cluster, we were able to purify (−)-discorhabdin L (1), but failed to purify the potentially new discorhabdin analog (m/z 368.0380) due to its very minor quantity.
With 21 nodes, cluster 2 was the biggest in the generated MN, which can be further divided into three subclusters ( Figure 2). Based on the elemental composition analysis, MS/MS fragmentation patterns, and biological source, two brominated alkaloids discorhabdin A [24] and discorhabdin G [25] were identified in subcluster 3. However, only discorhabdin A (2) was isolated in sufficient amounts for NMR and other spectroscopic analyses. Discorhabdin H2 [26] was the only annotated compound in subcluster 2. Unfortunately, neither this compound nor the remaining nodes that represent potentially new discorhabdin analogs could be purified in sufficient quantity. After a comprehensive examination of the global MN of the SPE fractions, two clusters attracted our attention ( Figure 2). Cluster 1 contained five nodes, four of which were annotated as known molecules discorhabdin L [13], its analog, discorhabdin D [23], and 1-methoxydiscorhabdin D [19], leaving the node at m/z 368.0380 to be a putatively new derivative ( Figure 2). From this cluster, we were able to purify (−)-discorhabdin L (1), but failed to purify the potentially new discorhabdin analog (m/z 368.0380) due to its very minor quantity.
With 21 nodes, cluster 2 was the biggest in the generated MN, which can be further divided into three subclusters ( Figure 2). Based on the elemental composition analysis, MS/MS fragmentation patterns, and biological source, two brominated alkaloids discorhabdin A [24] and discorhabdin G [25] were identified in subcluster 3. However, only discorhabdin A (2) was isolated in sufficient amounts for NMR and other spectroscopic analyses. Discorhabdin H2 [26] was the only annotated compound in subcluster 2. Unfortunately, neither this compound nor the remaining nodes that represent potentially new discorhabdin analogs could be purified in sufficient quantity.  The subcluster 1 (of cluster 2) contained four nonbrominated nodes at m/z 394.0560, 408.0710, 422.085, and 436.1010 connected with thick edges, indicating their high structural similarity. Elemental composition analysis revealed the difference of a CH 2 unit between these ions ( Figure 2, in red). In-depth analysis of their MS/MS spectra revealed the presence of the same MS fragment (m/z 352.0766, C 18 H 14 N 3 O 3 S) in all four compounds, suggesting these compounds to be discorhabdin alkaloids bearing an alkyl chain with varying lengths. From subcluster 1, we isolated the compound with m/z 394.0560 and identified it as a new metabolite, (−)-1-acetyl-discorhabdin L (5), as discussed later.
In addition, we purified three compounds, namely the known compound (+)-discorhabdin Q (3) as well as two new compounds, namely (−)-2-bromo-discorhabdin D (4) and (+)-1-octacosatrienoyl-discorhabdin L (6), which did not appear in the MN because of the low intensity of their MS fragments. The enantiopurity of all purified compounds was further checked individually by RP-DAD-HPLC on an analytical chiral column. The sharp single peak in the UV chromatograms confirmed the enantiopurity of 1-6.
In addition, we purified three compounds, namely the known compound (+)-discorhabdin Q (3) as well as two new compounds, namely (−)-2-bromo-discorhabdin D (4) and (+)-1octacosatrienoyl-discorhabdin L (6), which did not appear in the MN because of the low intensity of their MS fragments. The enantiopurity of all purified compounds was further checked individually by RP-DAD-HPLC on an analytical chiral column. The sharp single peak in the UV chromatograms confirmed the enantiopurity of 1-6.
On the basis of a strong NOE correlation between H-1 and H-7α ( Figure 6B), as well as the other NOE correlations shown in Figure 6B, the relative configuration of 6 was elucidated to be the same as compounds 1 and 5. The absolute configuration of 6 was established by comparing its experimental ECD spectrum with those of 1 and 5 (Supplementary Figure S8 with those of 1 and 5 (Tables 2 and 3) suggested 6 to be another analog of discorhabdin L esterified with a long chain fatty acid (δC 25-35 ppm; δH 1.2-1.4 ppm), which made this molecule very lipophilic. The discorhabdin L core structure that was evident from the 1D and 2D NMR data of 6 (Tables 2 and 3; Figure 6) accounted for 14 degrees of unsaturation.  Figure S36) accounted for the remaining 4 degrees of unsaturation. Hence, we concluded that the alkyl chain was an unbranched octacosa-triene-oic acid (28:3) (Tables  2 and 3). The HMBC correlation between H-1 (δH 5.79) and C-1' (δC 173.6) supported the attachment of the fatty acid at C-1 ( Figure 6A). The geometry of these three double bonds in the fatty acid portion was elucidated by analyzing the 13 C NMR chemical shifts of the six carbons neighboring the double bonds [29]. Carbon atoms adjacent to cis double bonds resonate around C 26.0-28.5, whereas those adjacent to trans double bonds appear at higher chemical shift values, namely C 29.5-38.0 [29]. The observed 13 C NMR shifts in (+)-6 at C 27.5, 28.1, 28.2 (x 2), 28.4 (x 2) confirmed the cis (Z) configuration of all three double bonds in the fatty acid part ( Table 3). The COSY correlations from C-2' to C-10' and HMBC correlations between H2-2'/C-1', C-4'; H2-3'/C-1', C-4', C-5'; H-6'/C-4', C-7', C-8'; H2-7'/C-9' (Figure 6A) allowed us to corroborate the position of two unsaturations at C-5' and C-9', and to assign the C-1' to C-10' portion of the fatty acid ( Figure 3). Comparison of the 1D NMR data of (+)-6 with the literature [30] also supports the presence of a ∆ 5,9 unsaturated fatty acid. This finding is not surprising since these cis,cis-5,9-dienoic lipids are common in sponges [31]. The chemical shift values of C-26', C-27' and C-28' (Table 3) were also assigned by comparison with the literature data [32]. Thus, compound (+)-6 was identified as C-1 octacosatrienoic acid (C28:3) ester of (−)-discorhabdin L (Figure 3). Due to availability of very minor amount of the compound (0.2 mg) and the failed attempts to improve the highly overlapped NMR signals through different solvents (e.g., MeOD, CDCl3, and DMSO-d6), we were unable to confirm the position of the third double bond. However, we believe that the lipid residue in 6 is related to the well-known 5Z,9Z demospongic acids bearing another double bond at C-17, or C-19, or C-23 [31].
On the basis of a strong NOE correlation between H-1 and H-7 ( Figure 6B), as well as the other NOE correlations shown in Figure 6B, the relative configuration of 6 was elucidated to be the same as compounds 1 and 5. The absolute configuration of 6 was established by comparing its experimental ECD spectrum with those of 1 and 5 (Supplementary Figure S8)

In Vitro Bioactivity Tests and Molecular Docking on Purified Compounds
Anticancer activity of the pyrroloiminoquinone-type alkaloids has been the main driving force for isolation of these intriguing structural types. Due to low quantities of the isolated compounds, we were only able to assess the in vitro anticancer activities of compounds 1, 5, and 6 against one cell line. We used HCT-116 colon cancer cells for testing, because of the observed high activity of the MeOH subextract and SPE fractions (Table 1), plus the availability of literature data for (−)-discorhabdin L (1) against this cell line (IC 50 value 6.2 µM) [17]. In the current study, compound 1 showed IC 50 value of 0.94 µM (equal to 0.33 µg/mL). The compound 5 displayed promising activity with an IC 50 value of 2.71 µM (= 1.1 µg/mL), while compound 6 was only modestly active against the same cell line (IC 50 value 34.0 µM, equal to 25.6 µg/mL). These results indicate that C-1 OH function is important for anti-colon cancer activity and the substitution of the C-1 OH group, especially with a long chain fatty acyl function is not favored.
Limited by the amounts of the isolated compounds, we performed a molecular modeling study (using Schrödinger software Maestro; www.schrodinger.com) on compounds 1-6 and (+)-discorhabdin L (1e), the (+) enantiomer of compound 1, against two known anticancer targets (topoisomerase I/II, indoleamine 2,3-dioxygenase IDO1) to estimate their potential anticancer activity and mechanism(s) of action. Where possible, based on suitable pdb structures, docking experiments were performed. We prepared available relevant pdb protein structures, removed the original ligands, and generated receptor grids. Small molecule 3D structures of the compounds containing a quaternary nitrogen were energetically minimized and possible tautomers/protonated states were evaluated (LigPrep, counter ion not specified). Next, we docked the optimized ligand structures into respective active sites (Glide SP). Calculated 3D binding modes were illustrated, or presented as 2D ligand-interaction diagrams, for clarity.
Docking of compounds 1-5 into the active site of topoisomerase I (pdb 1T8I) yielded plausible binding modes (Figure 7), while no binding pose could be calculated for compound 6, due to the sterically demanding side chain that did not fit into the tight binding pocket). Compared to the original ligand camptothecin, the flat, partly aromatic core of the discorhabdins 1-5 intercalated into the DNA part thereby forming aromatic π-π-stacking interactions while also addressing H-bonds towards residues of the topoisomerase I protein. Similar results were obtained by docking experiments with topoisomerase II (pdb 3QX3, original ligand etoposide) suggesting these proteins to be anticancer targets for the compounds 1-5. Molecular docking study performed (in analogy to the procedure described above) on compound 1e, which was reported to exhibit strong in vitro cytotoxicity [33], revealed a binding mode in the active site of topoisomerase I, too (Figure 7). Comparable to 1, the flat core forms a DNA-intercalating complex, but the ligand is distorted by 180 • . Thus, the binding modes of both compounds, 1 and 1e, suggest those ligands to be rather non-specific DNA intercalators.
We also performed docking experiments with another reported target for pyrroloiminoquinone alkaloids, namely the indoleamine 2,3-dioxygenase (IDO1) enzyme for which structural data including ligand-protein complexes are available. As cofactor to mediate physiological substrate oxidation, IDO1 contains a heme moiety and ligands typically form interactions by complexing the iron central atom. Examples for such IDO1 inhibitors and relevant interacting moieties include NLG919 derivative (imidazole-like nitrogen in pdb 5EK2) or ligand INCB14943 (hydroxylamidine moiety in pdb 5XE1). Since the compounds lack comparable nitrogen functionalities to able to interact with heme in a similar manner, docking of the compounds into these active sites revealed no plausible binding modes. However, recent reports demonstrated that another class of potent IDO1 inhibitors such as FXB-001116 (pdb 6AZW) and BMS-978587 (pdb 6AZV) bind to the IDO1 apo structure with high affinity, thereby displacing the heme moiety. Accordingly, we performed docking experiments of compounds 1-6 and 1e using the apo-protein. This approach suggested possible binding modes for compounds 1-4 and 1e in the apo active site of IDO1 (Figure 8), but not for the sterically more demanding compounds 5 and 6.  Inspecting the docking poses of 1 in the above-mentioned protein structures on a molecular level pointed towards a key role for the sterically defined OH-function in 1, as it mediates an Hbond towards Ala264. Another key H-bond interaction occurred between aromatic NH of 1 towards Ser167 (Figure 8). Compound 1 shows a rather planar core, anchoring the ligand with a strong shape-fit into the tight binding pocket. Relative to this core, the thioether bridge sits almost rectangular on top, occupying a lipophilic area within the binding site, flanked by residues Inspecting the docking poses of 1 in the above-mentioned protein structures on a molecular level pointed towards a key role for the sterically defined OH-function in 1, as it mediates an H-bond towards Ala264. Another key H-bond interaction occurred between aromatic NH of 1 towards Ser167 (Figure 8). Compound 1 shows a rather planar core, anchoring the ligand with a strong shape-fit into the tight binding pocket. Relative to this core, the thioether bridge sits almost rectangular on top, occupying a lipophilic area within the binding site, flanked by residues including Val350, Phe226, and Leu384. Within these small sets of compounds, docking of 1 reveals the optimal pose. We also docked (+)-discorhabdin L (1e), the enantiomer of 1, into the active site of IDO1 (in analogy to the procedure described above). Interestingly, the flat core was found to again fill the rather flat pocket. In comparison to 1, the core sits upside down in 1e, with the key H-bond towards Ala264 being maintained. Accordingly, the NH-bond towards Ser167 was lost, but the aromatic system formed a π-interaction to Tyr126 (Figure 8). These poses suggest the flat aromatic core of this type of compounds to be the main requisite to bind into the pocket. Furthermore, stereochemistry seems to play a minor role in binding. Thus, it can be assumed the compounds are rather non-specific IDO1-ligands.
The calculated binding mode of compound 4 yielded a shifted orientation of the core, with only one H-bond towards Ser167 (Figure 8). In contrast, compounds 5 and 6 gave no plausible docking solutions, again due to the sterically demanding ester moieties, thus also preventing H-bonding by OH towards Ala264.
In summary, the molecular modeling data suggested plausible binding modes for compounds 1-5 in the target structure of topoisomerase I, and for compounds 1-4 towards IDO1, respectively. However, it is well possible that this class of compounds bind further key anticancer targets, contributing to their cytotoxic potential.

Discussion
Since the discovery of discorhabdin C from a New Zealand Latrunculia sp. in 1986 [8], more than 40 discorhabdin analogs have been reported from different marine sponge genera [34]. Some discorhabdins contain bromination at C-2, C-4, or C-14 positions (e.g., discorhabdin A, discorhabdin C, 14-bromodihydrodiscorhabdin C) [8,23,24,35], some possess a sulfur bridge between C-5 and C-8 (e.g., discorhabdin B, discorhabdin Q) [24,36]. A few discorhabdins (e.g., discorhabdin L, discorhabdin D) are heptacyclic through the formation of an extra bridge between C-2 and N-18 [13,24]. Of the six compounds obtained in the current study, two of them (5 and 6) are (−)-discorhabdin L esters. Compound 6 is a triunsaturated C28 fatty acid ester of discorhabdin L. To our knowledge this is the first discorhabdin alkyl ester structure being reported from a marine sponge. Notably, Zou et al. (2013) reported atkamine, a new, large pyrroloiminoquinone scaffold containing a fused epoxybenzazepin and bromophenol groups connected with a cyclic sulfide ring [37]. Between the former and the latter rings, there is a substitution with a monosaturated C20 alkyl chain. The authors suggested this alkyl group to originate from (Z)-15-docosenoic acid, a fatty acid commonly found in sponge species that was possibly incorporated in the very early biosynthesis stages of atkamine [37]. Compounds 5 and 6 isolated in this study instead bear an esterification at C-1 position of the pyrroloiminoquinone ring system. Compound 5 is an acetyl ester of (−)-discorhabdin L, while 6 is an ester of discorhabdin L with an unbranched octacosatrienoic acid. Marine sponges, especially demosponges, are regarded as one of the richest sources of long-chain fatty acids (LCFAs; i.e., C23-34) [38,39]. The octacosatrienoic acid (C28:3) has been reported from marine sponges and corals [40][41][42], but not from any Latrunculia species. The only study that analyzed the FAs composition of Antarctic Latrunculia sponges in 2015 showed that Antarctic L. biformis contained diverse common and LCFAs (C16, C18), however longer chain unsaturated FAs were not found [43]. This is the first report of a discorhabdin-LCFA ester from nature, and the current study adds three new and intriguing analogs to the list of discorhabdin class alkaloids.
Discorhabdins have been repeatedly studied for their in vitro anticancer activity [10,13,14]. Limited by the strong cytotoxicity and the supply issue, no molecule from this chemical family has ever proceeded to further clinical studies. A few structure-activity relationship studies (SARs) associated with discorhabdins have been conducted, revealing that the ring closure by a bridge between C2 and N18 can significantly reduce the cytotoxicity, while a substitution at C-1 (i.e., OMe, NH 2 ) can enhance the anticancer activity [19,26]. The discovery of C-1 esters of discorhabdin L with good to moderate inhibitory activity against HCT-116 cell line here provides further insights for the SARs of discorhabdins.
Although many discorhabdins are associated with anticancer/cytotoxic activity, little is known on their exact mechanism(s) of action. Wada et al. (2011) evaluated (+)-discorhabdin A and its synthetic oxa analog for inhibition against a set of anticancer target enzymes, such as protein kinase, histone deacetylase, farnesyltransferase, telomerase, and proteasome [16]. (+)-Discorhabdin A and its synthetic oxa analog weakly inhibited the farnesyltransferase enzyme (IC 50 > 10µM) [16]. Geoy et al. recently tested the HIF-1α/p300 inhibition activity of several discorhabdins, including (−)-discorhabdin L (IC 50 value 0.73 µM) concluding them to be a novel class of HIF-1α/p300 inhibitors [17]. In the current study, an in silico molecular modeling study revealed the plausible binding modes of discorhabdins in two additional cancer target enzymes, topoisomerase I/II and IDO1.
In summary, guided by the anticancer activity and MN-based metabolomics, the Antarctic deep-sea sponge L. biformis led to the isolation and characterization of three known and three new discorhabdin alkaloids. Despite the small amounts of the extract and fractions that hampered the isolation of many further new discorhabdins, MN-based metabolomics proved useful for identification of chemical inventory of the sponge in early stages. Two compounds were a new type of discorhabdin esters that yielded meaningful SARs in comparison to the parent compound discorhabdin L. Mechanistic studies based on molecular modeling showed, for the first time, the potential binding of discorhabdins to additional anticancer targets that may be involved in their anticancer activity.
trawl at a depth of −291 m, and was fixated immediately after collection. Specimens were cleaned, pre-sorted, photographed, and transferred into buckets with cold seawater as soon as the catch was on deck. Subsamples were transferred into pure ethanol (96%) and the main parts were frozen at −20 • C. The sponges were transported to the Senckenberg Research Institute and Nature Museum in Frankfurt am Main, Germany. Tissue samples were taken and skeletal preparations were made for transmission light microscopy and SEM, according to standard protocols [44]. For taxonomic examination, the sponge spicules were mounted on microscope slides and studied by light microscopy and by SEM. Based on comparative morphology of skeletal characters, the sponge was identified as Latrunculia biformis, which is a common species in the Antarctic deeper shelf areas. For identification, the World Porifera Database [45] and relevant literature were used. A specimen (SMF 12109) is deposited in the Porifera collection of Senckenberg Research Institute and Nature Museum, electronically inventoried. The data are online available in the SESAM database.

Extraction and Isolation
The sponge material (43.615 g, frozen weight) was cut into small pieces and freeze-dried (Martin Christ, Osterode am Harz, Germany). The lyophilized biomass (5.809 g) was extracted at room temperature with water (3 × 200 mL) under agitation to yield the aqueous extract (1.892 g). The remaining sponge residue (3.572 g, dry weight) was extracted with MeOH (3 × 150 mL) and subsequently with DCM (3 × 150 mL) under the same conditions. Combined MeOH and DCM extracts were evaporated to dryness by a rotary evaporator to yield the crude organic extract (328 mg) that showed very strong anticancer activity against multiple cancer cell lines. This extract was partitioned between MeOH (100 mL) and n-hexane (100 mL) to yield MeOH (190 mg) and n-hexane (120 mg) subextracts. The MeOH-soluble portion, which exhibited strong anticancer activity was fractionated on a Chromabond SPE C18 cartridge. The elution with a step gradient MeOH:H 2 O mixture (0% to 100%)

Molecular Networking
The network was created using the UPLC-HRMS/MS data generated from the six active MeOH subfractions of L. biformis. All raw MS/MS data were converted from files (.raw) to mzXML file format using MSConvert (Version 3.6.10051, Vanderbilt University, Nashville, TN, USA). The converted data files were uploaded to the Global Natural Products Social molecular networking (http://gnps.ucsd.edu) platform using FileZilla (https://filezilla-project.org/) and a molecular network was created using the online workflow at GNPS [22]. The data were filtered by removing all MS/MS peaks within +/− 17 Da of the precursor m/z. MS/MS spectra were window filtered by choosing only the top 6 peaks in the +/− 50Da window throughout the spectrum. The data were then clustered with MS-Cluster with a parent mass tolerance of 0.1 Da and an MS/MS fragment ion tolerance of 0.05 Da to create consensus spectra. Further, concensus spectra that contained less than 2 spectra were discarded. A network was then created where edges were filtered to have a cosine score above 0.6 and more than 3 matched peaks. Further edges between two nodes were kept in the network if and only if each of the nodes appeared in each other's respective top 10 most similar nodes. The spectra in the network were then searched against GNPS' spectral libraries. The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 6 matched peaks. The output molecular networking data were analyzed and visualized using Cytoscape (ver. 3.61) [46].