Identification of chlorophyll a-b binding protein AB96 as a novel TGFβ1 neutralizing agent

The discovery of compounds and proteins from plants has greatly contributed to modern medicine. Vernonia amygdalina Del. (Compositae) is used by humans and primates for a variety of conditions including parasitic infection. This paper describes the serendipitous discovery that V. amygdalina extract was able to bind to, and functionally inhibit, active TGFβ1. The binding agent was isolated and identified as chlorophyll a-b binding protein AB96. Given that active TGFβ1 contributes to the pathology of many infectious diseases, inhibiting these processes may explain some of the benefits associated with the ingestion of this species. This is the first plant-derived cytokine-neutralizing protein to be described and paves the way for further such discoveries.


Results
The immunological effects of endotoxin-free aqueous extract of VA [VA(aq)], on human monocyte derived dendritic cells (mo-DCs) was studied. Mo-DCs were incubated with VA(aq) for 24 h after which they were assessed for changes in cell surface proteins and cytokine secretion. Although minor mo-DC activation was observed when assessing cell surface markers such as CD83 and CD86 by flow cytometry (data not shown), the most profound change was found to be in the supernatant where treated cells were shown to have significantly lower concentrations of total (active and latent) TGFβ1 (Fig. 1a). Our initial hypothesis was that this indicated a specific inhibition of TGFβ1 production, however when exogenous active TGFβ1 was added to the system, we saw the same reduction. This occurred in a similar manner to a known active TGFβ1 binder, heparin sulphate (HS) 23 (Fig. 1b), suggesting that an agent in the extract may be binding and 'removing' active TGFβ1 from the culture supernatant.
To assess if VA extract was binding to active TGFβ1 we developed a simple in vitro method whereby we plated out VA(aq) overnight, before washing and blocking. We then incubated the blocked VA(aq) with active TGFβ1 for 2 h. After incubation the supernatant was removed from the plate for testing (the 'unbound fraction'). The plate was then washed, acidified and neutralized, to elute the 'bound fraction' . Active TGFβ1 was measured in the unbound and bound fractions, with bound TGFβ1 seen in the VA(aq) and an anti-TGFβ1 antibody wells (Fig. 1c). We were next concerned that this binding may be non-specific and so tested 3 other cytokines, IL-10, IL-12p70 and, importantly, latent TGFβ1, and found that this phenomenon could only be observed with active TGFβ1 (Fig. 1d).
We then considered if sesquiterpene lactones, common efficacious agents in medicinal plants and present in VA, may be responsible for the observed binding. We assayed both artemisinin and parthenium finding that neither were able to bind active TGFβ1 under these conditions (Fig. 1e). Subsequently, we wanted to explore if VA(aq) was unique in having this TGFβ1 binding ability. We prepared extracts from 13 additional species, representing different branches of the plant kingdom and tested their ability to bind active TGFβ1. All species were able to bind, however there was a trend suggesting that members of the Asterales order, which contains many medicinal plants, contained more of the active TGFβ1 binding compound, whilst the Monocot clade contained less. (Fig. 1f). We next wished to assess if this binding rendered active TGFβ1 non-functional. Using a well-established reporter assay (where luciferase runs off the promotor of the TGFβ1 target SERPINE1 (PAI-1; plasminogen activator inhibitor-1) 24 ), we were able to show that VA(aq) was able to inhibit active TGFβ1 function (Fig. 1g).
We next used anion exchange chromatography to show a) that binding was occurring using a different assay and b) to suggest the isoelectric point of the agent for validation purposes. As can be seen in Fig. 1h, active TGFβ1 was again seen to bind to VA(aq) (with elution of active TGFβ1 only occurring with pre-incubation) and, in combination with the agent, active TGFβ1 eluted in fractions 46-52, which given the experimental conditions, gave the agent an isoelectric point of 4.5-5.5.
To isolate the active agent we performed a modified immunoprecipitation (IP), incubating VA(aq) with biotinylated active TGFβ1, or control protein (soybean trypsin inhibitor; STI) and streptavidin beads. The eluant was run on a gel and whole lanes were sent for mass spectrometric analysis. Two peptides were obtained, EVIHSRWAMLGALGCVFPELLSR and FGEAVWFK which are both present in the same protein; chlorophyll a-b binding protein (CabBP) AB96 (P04159). The initial hits were identified as being part of the sequence from P. sativum L. (Leguminosae), (Fig. 2a-c) but with the sequence for VA unavailable at this current time and the protein well-conserved, it is assumed that this sequence is also present in VA. These peptides were seen to have bound active TGFβ1 and not the controls (Fig. 2a,c). Using E.coli derived recombinant full-length folded CabBP AB96 from P. sativum L. (Leguminosae), we repeated the binding assays finding that CabBP AB96 was able to bind active TGFβ1 (Fig. 2d). Finally we repeated the functional assay using the luciferase reporter system, finding www.nature.com/scientificreports/ that CabBP AB96 was also able to inhibit the function of TGFβ1 in a significant manner when compared to either STI or CabBP buffer (Fig. 2e).

Discussion
This study reveals the first plant-derived cytokine binding protein. The central role that active TGFβ1 plays in many of the pathologies VA and other medicinal plants are ingested for, suggests that this protein, or a peptide derived from this protein, may help to explain some of the observed effects. One of our chief concerns with this work is the fact that monocots do not appear to be able to bind active TGFβ1 well, despite expressing CabBP. This is likely to be due the increased amount of vascular tissue in monocots compared to eudicots, meaning that, proportionally, less chlorophyll is present. Given our extract concentrations were based on total mass we believe it would be reasonable to conclude our results were due to these much lower amounts of CabBP in the starting material.
CabBP AB96 is membrane bound and therefore difficult to isolate without detergents and sonication. It is likely that the preparation methodology, which was robust and includes the use of detergents to remove endotoxin, allowed for its presence in the extracts. Indeed, with one of the peptides overlapping a transmembrane region (WAMLGALGCVFPELLSR; underlined region shows overlap) we believe that this must have occurred. www.nature.com/scientificreports/ Although we have shown that the whole recombinant E. coli produced protein can bind TGFβ1, we do not yet know if the whole protein is required for this binding. The peptides picked up in the mass spectrometry are next to each other, and may be suggestive of a functional region, however further studies are required to confirm whether or not this is the case.
The question remains as to why CabBP AB96 has the ability to bind active TGFβ1. We feel the answer may lie, in part, in relative charges; active TGFβ1 is more positively charged (isoelectric point; 7.73) and this charge is critical for its functionality in maintaining latency, whilst CabBP AB96 is more negatively charged (isoelectric point; 4.88) enabling it to bind magnesium ions (along with chlorophyll, lutein and neoxanthin). This could place CabBP AB96 as a heparin-mimetic, acting as a scaffold to cause the oligomerization of TGFβ1 23 . However, although charge may be important, it is unlikely that this is the sole determinant. Binding studies using more sophisticated techniques (e.g. isothermal titration calorimetry or surface plasmon resonance) are required to fully answer this question.
Whether CabBP AB96 can have a real impact on pathology, depends on the protein, or peptide's, functional survival after acidification, neutralization and digestion in the gastrointestinal tract, and entry into the circulation. Our mass spectrometry data show that the protein can be degraded by trypsin, however there is currently limited evidence of plant peptides entering and remaining in the circulation 25 . For intestinal helminth infections, there are fewer delivery concerns as CabBP AB96 peptides could act directly, or close to, the site of pathology without gaining entry into the circulation.
Regarding the reported efficacy of VA extract itself, the answer may lie in a combination of agents that act either directly on the pathogen (e.g. sesquiterpene lactones) or indirectly via immune polarization (e.g. β-glucans) or removal of immune evasion (CabBP AB96). We now know enough about pathogenic mutation under selection pressure to understand that a single agent is unlikely to be efficacious for the length of time VA appears to have been used. www.nature.com/scientificreports/ In conclusion, CabBP AB96 has been shown to bind to and inactivate active TGFβ1. The next step will be to identify whether this is a property of the whole protein, or merely a peptide fragment. In the case that this activity can be localized, then it may be possible to manufacture a mimetic or a peptide which carries the same activity, and this would have potential for therapeutic intervention.

Methods
Preparation of leaf extracts. Defined species were sourced from UK horticultural suppliers. These species are common, not endangered and therefore are not subject to the IUCN Policy Statement on Research involving species at risk of extinction, or the Convention on the trade in endangered species of wild fauna and flora. The species were validated by the suppliers. Fresh leaves were pulped with 50 ml distilled water using a pestle and mortar. The extracts were filtered (45 nm), before being boiled for 30 min and re-filtered (45 nm). Samples were then subjected to three endotoxin clean-ups [until the LAL test proved negative], using Triton-X114 (Sigma) as described 26 . The samples were then quantified either using a spectrophotometer or BCA assay (Pierce). Endotoxin tests (LAL; Lonza) were carried out in the presence of a β-glucan block (Lonza) [β-glucans present in plant cell walls cross react in the LAL assay] and samples were only used if levels were < 0.1EU/mg.

Generation of monocyte-derived dendritic cells. Blood packs were acquired from the National
Health Service Blood and Transplant Service (NHSBTS) an HTA licenced provider. Monocytes were isolated using CD14 microbeads (Miltenyi Biotec) and the MACS system (Mitenyi Biotec). After treatment with 1500U/ ml IL4 (Bio-Techne) and 400U/ml GMCSF (Bio-Techne), cells were cultured in AIM-V media (GIBCO) for 6 days. Cells were confirmed to be immature DCs using flow cytometry as described 27 . Extract was diluted in PBS to reach stock concentration for use on cells. Heparin sulphate (Sigma) was dissolved in PBS before use. Cell viability was assayed using Annexin-V/PI (BD biosciences) staining by flow cytometry. TGFβ1 reporter assay. MINK cells (generously provided by Professor Rifkin) were cultured and assays performed as described 28 . Antibodies, VA extract, CabBP, STI and CabBP buffer were diluted in PBS prior to use. Cell viability was assayed using trypan blue after the assay period.

Protein binding assay.
Anion exchange chromatography. Samples were prepared as follows: 10 ml 100 ng/ml active TGFβ1 (Bio-Techne) in PBS. 10 ml 50 mg/ml VA extract in PBS. 5 ml 100 ng/ml active TGFβ1 in PBS plus 5 ml 50 mg/ ml VA extract in PBS, mixed for 4 h on a rotator at RT. Samples were loaded using a 50 ml Superloop (Cytiva) and run on an ACTApurifier 10 system (Cytiva) using the equipment, reagents and programme described 29 . Fractions were assayed for TGFβ1 using an ELISA as described.
Modified immunoprecipitation. 500 µl 1 µg/ml biotinylated active TGFβ1 in PBS (Bio-Techne), 500 µl 1 µg/ml biotinylated Soybean Trypsin Inhibitor (Bio-Techne) and PBS were each mixed with 500 µl of 50 mg/ ml VA extract in PBS on a rotator for 4 h at RT. 50 µl magnetic streptavidin beads (Pierce) were washed in TBS + 0.1% Tween-20, added to each mixture, and mixing continued for a further 2 h. Beads were pulled out using a magnetic stand and samples were washed × 3 in TBS + 0.1% Tween-20. Samples were eluted from beads using SDS-PAGE reducing sample buffer (ThermoFisher) and run on SDS PAGE (ThermoFisher). Gels were stained and destained using a silver staining for mass spectrometry kit (Pierce). Whole lanes were cut out and analysed using mass spectrometry.