Here are some further details on toxicity of PAs which I hope are of interest. Subhuti Dharmanda’s article provides very good coverage but there are a number of points which need to be considered further, in particular relationships between chemical structure and toxicity and also the question of dosage. I’ve added a few other bits and pieces as well.

1. Structure-Toxicity Relationships

The relationships between chemical structure of PAs and their potential toxicity (including both acute and chronic toxicity) are reasonably well established. Firstly, it appears that there must be a double bond on the necine ring between C-1 and C-2. No toxicity has so far been demonstrated for PAs which are saturated at this position. Furthermore, there seems to be a definite ‘pecking order’ concerning patterns of subsitution whereby macrocyclic diesters are more toxic than acyclic diesters which in turn are more toxic than monoesters.

The table below shows those PAs included in the study on carcinogenicity in fruit flies1 mentioned in Dharmananda’s article, which is the most extensive of its kind so far conducted. They are listed according to their carcinogenic potential (senkirkine being the most toxic in this respect and supinine the least toxic). The pecking order mentioned above is easily observed. The macrocyclic diesters appear to be found mainly towards the top of the list, and the monoesters towards the bottom with the acyclic diesters intermediate. [NB. Some of the structures look identical - indicine, intermedine and lycopsamine. They are stereoisomers]

There are some additional details, not mentioned by Dharmananda, which are worth noting.

  1. The activity of senkirkine (top of the list) compared to supinine (bottom) is three orders of magnitude. Supinine has virtually no mutagenic activity.
  2. Hydroxylated derivatives of senecionine (a macrocyclic diester) such as retrorsine and jacoline exhibit an approximately five-fold reduction in toxicity. Mutagenic activity in general appears to decrease in each class with increasing hydroxylation.
  3. Acyclic diesters show a 5-10 times weaker activity than macrocyclic compounds.
  4. Monoesters in turn are between 10 and 100 times weaker than acyclic diesters with indicine N-oxide (and supinine) showing very weak activity.

Though it would be unwise to extrapolate directly from fruit flies to humans it is likely that qualitatively similar patterns of potency might be expected and it is perhaps significant that no cases of toxicity or carcinogenicity have been reported for species such as Eupatorium which contain predominantly monoesters.

One study2 demonstrates that the acute toxicity of PAs (as measured by LD50) seems to follow a similar pattern (in rats) though only 5 alkaloids were considered:

  1. Supinine and indicine (monoesters) - Rat LD50 > 1000 mg/kg.
  2. Echimidine (acyclic diester) - Rat LD50 = 200 mg/kg.
  3. Monocrotaline (macrocyclic diester) - Rat LD50 = 154 mg/kg.
  4. Retrorsine (macrocyclic diester) - Rat LD50 = 34 mg/kg.

2. Mechanism of toxicity

What is responsible for the toxicity of PAs? Pyrrole derivatives are formed as a result of microsomal oxidation of PAs in the liver. The initial step involves formation of dehydropyrrolizidines by loss of water, mainly catalysed by cytochrome P-450 monooxygenases, specifically CYP3A and CPY2B6. Hydrolysis of the ester subsituents to form the corresponding necine metabolites is mainly mediated by liver microsomal carboxylesterases. The resultant products react as alkylating agents with biological nucleophiles such as nucleic acids and proteins, though they have also been demonstrated to form glutathione conjugates which is presumably the major means of detoxification. A proposed scheme for the oxidation of an acyclic diester is shown below.


Though reactive, the resultant pyrroles can persist for long enough in aqueous media to react at sites remote from the site of formation. Thus they can pass from hepatocytes into adjacent sinusoids where they can react with endothelial cells, can affect red blood cells, and can reach the lungs and heart. The more persistant pyrrole derivatives may thus cause lung damage as well as being hepatotoxic.

The progression of liver damage appears to be fairly unpredictable and can occur suddenly months or even years after PA exposure3. The reversibility of damage is also uncertain though there is a high rate of complete recovery (50% or more).

As far as carcinogenicity is concerned, though tumours have been observed in rats there are apparently no known reports of cancer in domestic animals caused by exposure to PAs in their diet (grazing animals in particular may be exposed to high levels). In an analysis of reported outbreaks of human poisonings up until 1983, Culvenor4 reported that the dose range of alkaloids would have been in the range of 0.01-50 mg/kg/day. The recovery rate in humans from both acute and chronic PA poisoning appears to be high (>50%). Prakash at al.5 believe that the human liver repairs damage more efficiently after PA poisoning compared to lower animals. Although there don’t appear to be any proven cases of carcinogenicity in humans it is believed by some that primary liver tumours in natives of Central and South Africa may be due to long-term consumption of species of the genera Crotalaria, Cynoglossum, Heliotropium and Senecio6.

Even in rats it seems that it is only on chronic exposure to PAs at fairly high levels that damage is observed. Particularly relevant from the viewpoint of chinese herbs is a study by Hirono et al.7 who demonstrated that when a coltsfoot-containing diet was fed to rats for up to 600 days, there was a no-observable-effect level (NOEL) for the senkirkine in the plant of 300 ?g/kg/day. This is quite a large amount. Assuming a level of about 50 ppm in the dried plant (according to Bruneton8, who says that amounts in plants prior to blooming are larger - about 150 ppm) then this corresponds to 6 g dried plant/kg/day. Even if we apply an uncertainty factor of 100, as recommended in the technical report on PAs of the Australia & New Zealand Food Authority, and assume a tolerable intake of 3 ?g senkirkine/kg/day then this corresponds to something like 60 mg dried plant/kg/day. In an adult weighing 70 kg this would be equivalent to consumption of 4.2 g dried coltsfoot per day over a period of almost two years! This calls into question the very cautious recommendations of the German Commission E monographs (GCEM). More on this below.

3. PAs and Chinese Herbs

In a recent report on the supposed tumorigenic potential of PAs in chinese herbs, about 22 are reported as containing PAs9. Most of these are not likely to be used by TCM practitioners in the West. Of greatest relevance in countries such as the UK are Kuan Dong Hua (Tussilago farfara and Petasites japonicus), Zi Cao (Lithospermum erythrorhizon, Arnebia euchroma, A. guttata, and Onosma paniculatum) and Pei Lan (Eupatorium fortunei and E. japonicum). Following the information given in Tang’s paper on carcinogenic constituents of chinese herbs10, the table below shows the structures of those PAs most commonly associated with each species. Note that since there is very little information available on PAs in Eupatorium fortunei, those known to be present in the related E. cannabinum (hemp agrimony) and E. japonicum are given instead. Both of these herbs are apparently sometimes substituted for E. fortunei anyway. The alkaloids are arranged in terms of predicted toxicity, with the macrocyclic diesters at the top.


StructureToxicity classSpecies
SenecionineMacrocyclic diesterTussilago farfara, poss. Petasites japonicus
SenkirkineMacrocyclic diesterTussilago farfara
PetasitenineMacrocyclic diesterPetasites japonicus
NeopetasitenineMacrocyclic diesterPetasites japonicus
MyoscorpineAcyclic diesterLithospermum erythrorhizon
IntermedineMonoesterLithospermum erythrorhizon, Eupatorium cannabinum
AmabilineMonoesterEupatorium cannabinum, E. japonicum
EchinatineMonoesterEupatorium cannabinum
LycopsamineMonoesterEupatorium cannabinum
RinderineMonoesterEupatorium cannabinum
SupinineMonoesterEupatorium cannabinum
7-AngeloylretronecineMonoesterArnebia euchroma
TussilagineNontoxicTussilago farfara (poss. artifact)
IsotussilagineNontoxicTussilago farfara (poss. artifact)
PetasinineNontoxicPetasites japonicus
ViridiflorineNontoxicEupatorium japonicum
CynaustralineNontoxicEupatorium japonicum


Both Tussilago and Petasites (as well as Eupatorium japonicum) contain a number of nontoxic (1,2-saturated) PAs, though both tussilagine and isotussilagine may be artifacts produced by methanol extraction. Otherwise these two herbs are the only ones in this list to contain macrocylic diesters. If we were to assign toxicity simply on the basis of chemical structure we might expect the following order: Tussilago = Petasites > Lithospermum > Eupatorium = Arnebia. However, this ignores the question of PA concentrations.

There appears to be some disagreement on this between different sources. In the case of Tussilago, for instance, Dharmananda reports that PAs are essentially absent in chinese samples but that senkirkine may reach levels of 0.015% (150 ppm) in some samples (presumably european), while Bruneton, as already mentioned, gives figures of 150 ppm (total PAs) in capitula prior to blooming, decreasing to a level of 50 ppm for the whole plant on drying (this was for a specimen from North America). He also states that european plants are less rich in total alkaloids than oriental ones (contradicting Dharmananda) though he doesn’t cite any references. Assuming about 50 ppm then, as we have already seen, a single infusion made from 6 g of the dried plant would contain 300 ?g of PAs of which senkirkine, a macrocyclic diester, is the principle alkaloid . The GCEM specify that daily intake should not exceed 10 ?g in the case of infusions and that duration of use be limited to 4-6 weeks. In order to satisfy these requirements, therefore, no more than 200 mg of dried plant should be used.

This is a very low amount by the standards of chinese herbal formulae and mirrors the calculations made by Dharmananda for Lithospermum erythrorhizon, based again on the figure given above (10 ?g), where only 50 mg would be allowed (given an alkaloid concentration of 0.02% or 200 ppm). In the case of Lithospermum there are no macrocyclic diesters present however which calls into question the necessity of making the same recommendations for the two species. If Arnebia is substituted for Lithospermum (and we assume a concentration of 0.0006% or 6 ppm) then a daily ration of 1666 mg (1.666 g) of dried herb would be allowed. If the principle alkaloid is 7-angeloylretronecine, a monoester, then it is likely again that this recommendation is too stringent.

An important question, however, concerns why the authors of the GCEM should make these recommendations in the first place. Since they don’t give any reasons one is left to guess, and it seems quite likely that this figure comes from the available data on cases of veno-occlusive disease in humans which indicate a tentative NOEL of 10 ?g/kg/day7. Once again the Australia and New Zealand Food Authority recommend that an uncertainty factor of 10 should be applied to this figure to account for human variability, giving a provisional tolerable daily intake (PTDI) of 1 ?g/kg/day. Even with this figure the acceptable daily intake for an adult weighing 70 kg would therefore be about 70 ?g per day, somewhat higher than that proposed by the GCEM. In the case of Tussilago, for instance, an acceptable daily amount might be about 1.4 g in infusion. It should be stressed that, on the basis of existing information, this represents a very cautious assessment. It seems probable that up to 10 times that amount would be unlikely to cause harm particularly if the herb was not taken for more than 4-6 weeks.

4. Moderating factors

In the case of Eupatorium cannabinum at least there is some evidence that extracts of the whole herb exert choleretic and hepatoprotective effects on rats (at doses of 250 mg/kg). Experiments have also shown the cytotoxicity of the sequiterpene lactones from this species on several types of tumour cell in vitro8. It seems possible therefore that the whole plant may moderate potentially harmful effects arising from isolated constituents.

Another recent study showed that rats pretreated with glycyrrhizin and glycyrrhetinic acid (from licorice) resulted in a reduction of retrorsine-induced hepatotoxicity (retrorsine being a macrocyclic diester)11. Given the ubiquitous use of Gan Cao in chinese herbal medicine this might be a particularly useful herb to add to formulae containing PAs (it probably would be anyway), in particular Kuan Dong Hua, since retrorsine and senkirkine are structurally similar. It also has useful expectorant effects, thus complementing the antitussive properties of the latter.

English TCM-Source

  1. Frei HJ, Luethy J, Brauchli L, Zweifel U, Wuergler FE & Schlatter C, Chem. Biol. Interact., 83: 1 (1992)
  2. Mattocks AR, ‘Chemistry and Toxicology of Pyrrolizidine Alkaloids’, Academic Press, London (1986) (p.192).
  3. Bras G, Brooks SHE & Walter DC, J. Pathol. Bacteriol., 82: 503 (1961).
  4. Culvenor CCJ, J. Toxicol. Environm. Health, 11: 625 (1983).
  5. Prakash AS, Pereira TN Reilly EB & Seawright AA, Mutation Res., 443: 53 (1999).
  6. Röder E, Pharmazie, 55: 711 (2000).
  7. Australia New Zealand Food Authority, Tech. Report Series, no. 2, (2001).
  8. Bruneton J, ‘Pharmacognosy, Phytochemistry, Medicinal Plants’, 2nd. Ed., Intercept, pp. 833-847 (1999).
  9. Fu PF, Yang Y-C, Xia QS, Chou MW, Cui YY & Lin G, J. Food & Drug Analysis, 10: 198 (2002).
  10. Tang WC, Abstr. Chinese Medicine, 6: 227 (1995).
  11. Lin G, Nnane IP & Cheng TY, Toxicon, 37: 1259 (1999).