Viral infectious diseases continue to be a global threat to human health (WHO, 2020), as best illustrated by the ongoing SARS-CoV-2 pandemic. The development of preventive or therapeutic antiviral drugs allowed to effectively treat several viral diseases, but most approved antivirals target only a limited number of viral pathogens such as HIV-1, HCV, or herpes viruses. In comparison, the arsenal of drugs available to combat the more than 200 known human pathogenic viruses is very limited with no specific antiviral therapies against most of them (Nii-Trebi, 2017; WHO, 2022). Therefore, new drugs are needed, especially those that are less prone to resistance induction (Mahlapuu et al., 2016; Browne et al., 2020) and that have a broad activity which prepares for reemerging or emerging pandemic viruses (Kuroki et al., 2021). One unique class of therapeutic compounds are peptides which attained much attention over the course of the 20th century (Lau and Dunn, 2018). Peptides are recognized for their high selectivity and efficacy and are relatively safe and well tolerated (Mahlapuu et al., 2016; Wang et al., 2022). Peptide drugs now account for ∼7% of the total number of approved pharmaceuticals (Al Musaimi et al., 2021) and this number is expected to increase in the near future because of the tremendous progresses made in peptide synthesis and purification (Wang et al., 2022).

Antimicrobial peptides (AMPs) can also be found in humans, bacteria and plants (Huan et al., 2020). Plants have developed mechanisms to protect themselves against invading pathogens such as viruses, bacteria, fungi, nematodes and insects (Mott et al., 2014; Nishad et al., 2020). Since they only have the innate immune response, a wide variety of peptides is expressed (Mott et al., 2014; Mammari et al., 2021), many of which are rich in cysteines. Plant cysteine-rich peptides have been isolated from thousands of species (Liu et al., 2017); they are extremely diverse with hundreds of different peptides produced by a single plant species. These peptides are classified into different families mainly based on their cysteine spacing, disulfide bond network and three-dimensional fold (Silverstein et al., 2007; Tam et al., 2015; Hellinger and Gruber, 2019; Li et al., 2021). One of these families are cyclotides, which are circular, disulfide-rich peptides and typically consist of 28–37 amino acids (de Veer et al., 2019). They comprise a typical structural architecture, termed the cyclic cystine knot (CKK) motif. This is characterized by a head-to-tail cyclized peptide backbone and interlocking arrangement of three disulfide bonds (Figure 1). Due to this unique topology, they exhibit exceptional resistance to thermal, chemical and enzymatic degradation (Colgrave and Craik, 2004). Cyclotides are non-classical AMPs since they do not exhibit the typical AMP structure comprising a cationic amphipathic α-helical or β-sheet structure (Mahlapuu et al., 2016).

Cyclotides do not occur in all plants, but were isolated from flowering plants—including many medicinal plants with ethnopharmacological relevance—belonging to Violaceae, Rubiaceae, Cucurbitaceae, Fabaceae, and Solanaceae (Gruber et al., 2008; Poth et al., 2011, 2012; Nguyen et al., 2013; Burman et al., 2015; de Veer et al., 2019). Typically, cyclotides do not appear as singletons, but a single plant produces dozens of cyclotides (Hellinger et al., 2015; de Veer et al., 2019). Most plants have their own individual suite of cyclotides and most cyclotides are unique to a particular plant meaning that it is uncommon for one cyclotide to occur in multiple plant species (Gruber et al., 2008). However, variations in cyclotide content between tissues as well as over different seasons or in different geographic locations have been reported (de Veer et al., 2019). For example, in Viola tricolor (Figure 1C), a member of the Violaceae family, we previously isolated and characterized 164 different cyclotides (Hellinger et al., 2015). This plant is a common horticultural plant, and has been used in traditional medicine for heat-clearing, detoxification, and relieving coughs (Tang et al., 2010). Viola tricolor is also widely used in Russian official medicines, and a monograph on this plant is even included in the Russian Pharmacopoeia (Shikov et al., 2014, 2017, 2021). Since violaceous plants at a global scale may be the source to as many as 150,000 individual cyclotides, this commercially available medicinal herb may be a suitable starting point for bioactivity-guided screening studies (Hellinger et al., 2015). As a part of the plant host defense, cyclotides were shown to be insecticidal (Jennings et al., 2001), hemolytic (Wang et al., 2008), anti-bacterial (Strömstedt et al., 2017; Slazak et al., 2018), anti-fungal (Slazak et al., 2021), anti-cancer (Pinto et al., 2018), anti-fouling (Göransson et al., 2004), and antiviral (Daly et al., 2004; Ireland et al., 2008; Wang et al., 2008; Gerlach et al., 2013). It is hypothesized that the diversity of cyclotides in a single plant may reflect a strategy to overcome resistance development by pathogens or pests (de Veer et al., 2019).

Therefore, cyclotides represent a rich natural source of bioactive molecules with prospects for development as therapeutics against cancer and infectious diseases (Mammari et al., 2021). Making use of the natural evolution, a huge variety of already optimized peptides can be assessed (Lazzaro et al., 2020; Atanasov et al., 2021). Peptides have a good efficacy compared to small molecules (Wang et al., 2022) and a good safety profile because their degradation products are natural amino acids and they are generally less immunogenic than recombinant proteins or antibodies (Mahlapuu et al., 2016). In contrary, first-generation antivirals have severe side effects due to pour specificity (Mammari et al., 2021). Cyclotides are expected to be broadly active due to their direct mode of action destabilizing lipid membranes (Henriques and Craik, 2012; Henriques et al., 2012) which will most likely prevent resistance development (Mahlapuu et al., 2016; Browne et al., 2020; Huan et al., 2020). Finally, peptides can be chemically modified and structure-activity-relationship studies can help to improve therapeutic features such as bioavailability, specificity and bioactivity (Browne et al., 2020; Huan et al., 2020; Dijksteel et al., 2021; Wang et al., 2022). Typical AMPs have a low oral and systemic bioavailability due to enzymatic degradation (Mahlapuu et al., 2016), but cyclotides have an exceptional resistance to degradation because of their unique topology (Colgrave and Craik, 2004), making them a promising class of antivirals (Lau and Dunn, 2018). And indeed, anti-HIV activity was shown for the prototypic cyclotide kalata B1, but also other cyclotides like kalata B8, varv E and circulins A and B (Gustafson et al., 1994; Daly et al., 2004, 2006; Wang et al., 2008). Several classical AMPs are already on the market and further AMPs are currently evaluated in clinical trials (Mahlapuu et al., 2016). For example, the antiviral peptide T20 (enfuvirtide) is FDA-approved and commercially available (Huan et al., 2020). Other plant-derived peptides are also in clinical development. For example, griffithsin, a lectin isolated from the red alga Griffithsia sp., is investigated in clinical studies (NCT02875119; NCT04032717) as an anti-HIV microbicide for prevention of sexual transmission (Lee, 2019). Since it recognizes mannose, it has a broad-spectrum antiviral activity and was shown to inhibit SARS-CoV-2 recently (Cai et al., 2020). Additionally, the first cyclotide (called T20K) is currently in clinical trials for multiple sclerosis (Gründemann et al., 2019).

As nature-derived products are attractive candidates for translational application (Atanasov et al., 2021), we aimed to identify novel antiviral cyclotides. To this end, we established a library of plant extracts largely consisting of species with ethnopharmacological relevance that have previously been identified as a rich source of cyclotides or cyclotide-like knottin peptides. This encompasses ten plant species: Viola odorata, Viola tricolor, Psychotria solitudinum, Palicourea tomentosa, Carapichea ipecacuanha, Bryonia alba, Beta vulgaris, Sambucus nigra as well as Momordica charantia (He et al., 2013; Koehbach et al., 2013; Karpyuk et al., 2015; Fahradpour et al., 2017; Álvarez et al., 2018; de Veer et al., 2019; Retzl et al., 2020). Using solvent and solid-phase extraction yielded the plant extracts enriched in peptides. These extracts were analyzed for cytotoxic and anti-HIV-1 activity in cell culture-based infection assays. The resulting inhibitory concentration (IC₅₀) and cell toxic concentration (CC₅₀) were used to calculate selectivity indices and select the plants species for more detailed analysis. Using these criteria, the particular extract from Viola tricolor—a species used in traditional medicine and listed in the European Pharmacopoeia—had the most selective antiviral activity and was analyzed more thoroughly. To confirm the presence of cyclotides as bioactive molecule class in the extract, we applied (sub) fractionation by HPLC and mass spectrometry and identified several mixtures of cyclotides. These mixtures of cyclotides were subsequently analyzed again for IC₅₀, CC₅₀ and selectivity indices. Lastly, in an approach to determine the antiviral mechanism, we performed pre-treatment experiments of the virions with the cyclotides.

Plant defense peptides such as cyclotides are thought to be part of the innate plant host defense. The unique disulfide-knotted and head-to-tail cyclized topology may be one reason for the previously observed bioactivity of cyclotides (Daly et al., 2004; Ireland et al., 2008; Wang et al., 2008; Gerlach et al., 2019). To identify novel antiviral peptides, we established a library of plant extracts largely consisting of cyclotide-rich species with ethnopharmacological relevance and screened them against HIV-1, a representative and important member of the retrovirus family. Following the initial screen, the most active extract of Viola tricolor was selected to detailed analysis. Through two HPLC purification steps, four Viola tricolor-derived peptide subfractions were obtained which inhibited HIV-1 concentration-dependently with IC₅₀ values between 0.6 and 9.8 μg/ml. The antiviral fractions contained several cyclotides such as vigno 2/5, vitri 2, vigno 8/9, kalata S, vigno 3/4, kalata B1, and cycloviolacin O12/O22, belonging to the Möbius or the bracelet subfamily (Table 3, Supplementary Table S2). Cyclotides have been found in every member of the Violaceae family examined so far and they occur in all tissues of these plants, including roots, stems, leaves and flowers (Burman et al., 2015). These cyclotides have previously been identified in different Viola species (Ireland et al., 2006; Hashempour et al., 2013; Hellinger et al., 2015; Esmaeili et al., 2016), which is not surprising since more than 200 cyclotides are known within the Violaceae family (de Veer et al., 2019). However, only for some anti-HIV activity was shown to date (Gustafson et al., 1994; Daly et al., 2004, 2006; Wang et al., 2008).

All subfractions tested were more active when the virus was pre-incubated, suggesting that antiviral activity is not mediated indirectly via a cell factor, but by the direct action of cyclotides on viruses. When the compounds were washed out 3 hours after infection, the antiviral activities were in the same range as when treated for 2 days, which is further evidence that the antiviral activity is mediated by the actual effect on the virions. This is in line with previously published data showing that cyclotides are able to permeabilize model membranes (Henriques and Craik, 2012; Henriques et al., 2012). It has been published that cyclotides target membranes through specific interactions with phospholipids containing phosphatidylethanolamine (PE) headgroups and then insert through nonspecific hydrophobic peptide-lipid interactions (Henriques et al., 2011, 2012; Henriques and Craik, 2012; Cranfield et al., 2017). Subsequently, this promotes outward movement of PE phospholipids, exposing more PE in the outer leaflet. This self-promotes the binding of more cyclotides until a threshold concentration is achieved leading to self-aggregation, pore formation and eventual membrane disruption (Henriques and Craik, 2012). Whether the cyclotides identified in this work may exert similar effects needs to be further examined.

The analyzed cyclotide fractions also reduced metabolic activity, which is in line with previous studies (Ireland et al., 2008; Wang et al., 2008). This effect on the cells can probably be explained by the mechanism of action of the cyclotides acting on lipid membranes (Henriques and Craik, 2012). However, the activity of cyclotides is dependent on their affinity for lipid bilayers and the lipid composition. Although the total lipid in mammalian cells comprises ∼20% PE phospholipids (Patel and Witt, 2017), model membrane studies showed that kalata B1 targets HIV particles with high preference, since raft-like HIV membranes contain PE phospholipids and cholesterol/sphingomyelin domains, which enhances the interaction (Brügger et al., 2006; Henriques et al., 2011; Mücksch et al., 2019). The preference for HIV particles over eukaryotic cells is also supported by the calculated selectivity index (SI) of up to 8.1 ± 5.4 herein. Such therapeutic windows have also been observed by others (Ireland et al., 2008; Wang et al., 2008), e.g. Wang et al. determined a CC₅₀ of 5.7 µM and an IC₅₀ of 0.66 µM resulting in an SI of 8.6. The ideal compound should have a relatively high toxic but a very low active concentration. The higher the SI ratio, the theoretically more effective and safer a drug would be during in vivo treatment (Cushnie et al., 2020). Besides, the hemolytic activity of several cyclotides isolated from V. tricolor has been determined previously with 50% hemolytic concentrations (HD₅₀ values) ranging from 4 to 226 µM suggesting a similar or even higher SI for red blood cell lysis (Tang et al., 2010).

In the future, it would be interesting to investigate if cyclotides have preferences for specific vesicles sizes. The so-called curvature-sensing peptides such as the AH-peptide preferentially target smaller vesicles and are antiviral (Jackman et al., 2018; Park et al., 2019). Membrane bending is enlarged in smaller vesicles, leads to lipid packaging defects between neighboring lipid molecules and subsequently to insertion of the amphipathic α-helical peptides (Kawano et al., 2019; Park et al., 2019). It cannot be expected that cyclotides have the same mechanism since they have a different tertiary structure, however, stronger membrane bending in smaller vesicles leading to packaging defects may also facilitate hydrophobic peptide-lipid interactions, pore formation and membrane disruption. This could lead to the preferential destruction of viral over cellular membranes improving the therapeutic window.

It is noteworthy that the antiviral fractions contained several cyclotides, which raises the question of synergistic effects. A well-known advantage of plant substances is their complex composition, consisting of many compounds with multiple activities that together give a greater overall activity (Schmidt et al., 2008). Also, typical AMPs can have synergistic activity (Lazzaro et al., 2020). Therefore, mixtures of cyclotides could also have additive antiviral effects that reduce the tendency to develop resistance. Moreover, Gerlach et al. showed that the cyclotide cycloviolacin O2 increases the activity of the HIV protease inhibitors saquinavir and nelfinavir by pore-formation in HIV-infected cells and viral membranes at non-hemolytic concentrations (Gerlach et al., 2013, 2019). It is also suggested to use AMPs in combination therapies to have higher activity but also to reduce resistance development (Dijksteel et al., 2021). Hence, cyclotides may augment antiretroviral therapy efficacy and be useful for combination therapies.

Since cyclotides act on lipid membranes it is reasonable to assume that they could be broadly effective antiviral agents. Indeed, the antiviral activity of a cyclotide against influenza A virus has been reported (Liu et al., 2014) and it would be interesting to investigate the antiviral activity of the extracts or purified cyclotides against other viral pathogens such as SARS-CoV-2. In addition, cyclotides have been successfully used for molecular grafting, i.e. inserting small bioactive epitopes into the stable cyclotide scaffold (Wang et al., 2014). Due to the plasticity of cyclotides, this usually preserves their structural integrity and activity (Clark et al., 2006) and can enable the development of dual-function antiviral agents by conjugating the cyclotides with specific virus entry inhibitors.

In sum, cyclotides and extracts described herein may offer a promising starting point for innovative therapeutic antiviral agents: the peptides have been structurally optimized by evolution to serve particular biological functions, and can be optimized for activity and stability for therapeutic applications (Atanasov et al., 2021). Moreover, cyclotides have an exceptional resistance to thermal, chemical and enzymatic degradation (Colgrave and Craik, 2004) which might increase bioavailability, and their mode of action has limited propensity for resistance development (Mahlapuu et al., 2016; Browne et al., 2020; Huan et al., 2020). Above all, V. tricolor is a medicinal plant, listed in the European Pharmacopoeia (Hellinger et al., 2015; EMA/HMPC/131734/2009), and its herein described properties may be explored in the future for applications of herbal preparations with antiviral activity.