Malodorous impurities

Smelly impurities in APIs and intermediates can lead to multiple problems.  In addition to poor patient compliance and complaints all the way back to the supplier, those making and analyzing batches of APIs and intermediates may be forced to shower and change clothes before going home.  Some compounds can be identified at concentrations as low as 10^-1 ppt, as for the “grapefruit thiol” [(R)-2-(4-methylcyclohex-3-en-1-yl)propane-2-thiol] that imparts a pleasant aroma to grapefruit juice.  Such volatile compounds are usually detected and quantified by various chromatography / mass spectrometry techniques, and method development for extremely low levels of impurities can be challenging.  Chirality can also be important; the enantiomer of the grapefruit thiol is said to have an obnoxious thiol aroma.  Sotolon, found in oxidized dry white wines, has an intense curry odor, with the (S)-enantiomer about 100 times more potent than its enantiomer [1].  A compound can also be pleasant-smelling or malodorous depending on the concentration; the tertiary thiol 4-mercapto-4-methyl-2-pentanone contributes flavor to hops used in making beer and to the odor and flavor of wine from the Cabernet sauvignon grape, but at higher concentrations it has an off-odor similar to cat urine [2] [3].  The odor of phenylacetic acid has been described as honey-like, or similar to urine [4].  4-Phenylbutyric acid smells reminiscent of vomit, and a polymer coating masked its aroma in the drug product [5].  Because some smelly compounds can be readily sensed at very low concentrations, reducing them to undetectable levels can be challenging.

Often a means to reduce the level of an impurity can be posed once the compound has been identified.  For instance, the “catty ketone” is generated from the aldol reaction between two molecules of acetone, and aldol reactions can be catalyzed by base or acid; dehydration to enones such as mesityl oxide (an intermediate in preparation of MIBK) is favored under acidic conditions [6].  Under pH-neutral conditions less of both the aldol intermediate and the enone would be generated.  In juice from wine grapes the adduct from 1,4-addition of mesityl oxide with the cysteinyl thiol of glutathione has been identified [7]. Base-catalyzed beta-elimination of the glutathione adduct generates the tertiary thiol and the dehydro-alanine derivative; a similar sequence of thiol – ene reaction [8] followed by beta-elimination can provide the element of H₂S to generate this tertiary thiol from reaction with other mercaptans.  Overall, minimizing the amounts of base and acid should reduce the formation of this tertiary thiol derived from acetone and a sulfur nucleophile.  Experimental data did confirm this hypothesis; under less basic conditions the amount of this malodorous impurity was reduced in a Schotten-Baumann reaction in aqueous acetone [9]. But in general a better approach would be to avoid generating the tertiary thiol by not using or generating acetone in the reaction, or by not charging the compound that provides the elements of H₂S.

The generation of a smelly contaminant may not be related to the chemistry of generating the API.  2,4,6-Trichloroanisole is responsible for a moldy newspaper smell of “corked wines” and is attributed to applications of halophenols as fungicides and insecticides to trees used to make corks from the tree bark [10].  Johnson & Johnson recalled batches of Tylenol that had a musty odor, caused by 2,4,6-tribromoanisole, from application of 2,4,6-tribromophenol used to treat wooden shipping pallets associated with storage of the drug product [11]. Microbial methylation of those phenols generated the anisole derivatives [12].

The general approach to controlling impurities in an API is summarized by the Avoid, Control, and Expel approach (ACE).  Controlling the reaction pH, as mentioned earlier, reduced but did not eliminate the amount of a malodorous impurity.  Nor was this impurity completely purged (expelled) by crystallization and washing of the wet cake.  In that case substituting a solvent for acetone would avoid generating the tertiary thiol.  Since malodorous impurities can be detected at extremely low levels, and purging them to non-detectable levels may be difficult, the best approach is to identify these compounds and devise chemistry to not form them.

Citations

1. Pons, A.; Lavigne, V.; Landais, Y.; Darriet, P.; Dubourdieu, D. Agric. Food Chem. 2008, 56, 1606. DOI: 10.1021/jf072337r.
2. Kishimoto, T.; Kobayashi, M.; Yako, N.; Iida, A.; Wanikawa, A. Agric. Food Chem. 2008, 56, 1051. 10.1021/jf072173e
3. McGorrin, R. J. “Character-impact flavor compounds” Chapter 9 in Sensory-Directed Flavor Analysis; 1^st; Marsili, R., Ed.; CRC Press; 2006; p. 223.
4. Ruisinger, B.; Schieberle, P. Characterization of the Key Aroma Compounds in Rape Honey by Means of the Molecular Sensory Science Concept. Agric. Food Chem. 2012, 60, 4186. https://doi.org/10.1021/jf3004477.
5. Cederbaum, S. D.; Edwards, J.; Kellmeyer, T.; Peters, Y.; Steiner, R. D. Genet. Metab. 2023, 138, 107558. https://doi.org/10.1016/j.ymgme.2023.107558.
6. Smith, M. B.; March, J. March’s Advanced Organic Chemistry: 6^th; Wiley, 2007; p. 1349.
7. Fedrizzi, B.; Pardon, K. H.; Sefton, M. A.; Elsey, G. M.; Jeffery, D. W. First Identification of 4-S-Glutathionyl-4-methylpentan-2-one, a Potential Precursor of 4-Mercapto-4-methylpentan-2-one, in Sauvignon Blanc Juice. Agric. Food Chem. 2009, 57, 991.
8. Northrop, B. N.; Coffey, R. N. Thiol–Ene Click Chemistry: Computational and Kinetic Analysis of the Influence of Alkene Functionality. Am. Chem. Soc. 2012, 134, 13804.
9. Anderson, N. G.; Wang-Iverson, D., unpublished.
10. https://www.wineenthusiast.com/culture/wine/cork-taint-wine-fault-guide/
11. Eng. News 2010, 88(4), 18.
12. Bleiler, R. J.; Kuhrt, F.; Wright, D. Identification of Tribromoanisole (TBA) using Headspace Solid Phase Micro Extraction (SPME) Collection with Multi-Dimensional Gas Chromatography / Mass Spectrometry / Olfactometry (MDGC/MS/O). http://suremarketingintl.com/wp-content/uploads/2015/08/CTS_Article_11-4102.pdf (MOCON Inc.)

Smelly compounds occasionally contaminate final products, and can be difficult to remove because they can be detected at very low levels. Smell and taste are some organoleptic characteristics [1]. Many are small molecules with low molecular weight, and they may be quite volatile.

Examples: thiols, smoke (some may be released by hydrolysis: wine), trichloroanisole, tribromoanisole, more. p-Methoxybenzaldehyde, a component of cherry flavoring.

“What Is Organoleptic Testing?” https://www.reference.com/health/organoleptic-testing-55fb32bb6a9a5a8a

Bottom line: to minimize side products arising from isomerization, minimize decomposition of the metathesis catalyst.  This includes removing impurities that inhibit a catalyst or promote its decomposition.  Catalyst screening and conducting the metathesis at a lower temperature may also be helpful.

Overview In 2005 Yves Chauvin, Robert Grubbs, and Richard Schrock received a Nobel Prize for their work in metathesis.  Professors Grubbs, Schrock and others have expanded this chemistry from polymers to synthetic organic chemistry to pharma.  Metatheses in pharma have been recently reviewed [1, 2], and wider-ranging resources are available [3, 4, 5].  For pharmaceutical and specialty chemical applications, ring-closing metatheses (RCMs) [6] are more common than cross metatheses.

Metatheses are equilibrium reactions and need to be actively quenched on scale [7].  Differences in reactivity and stability of metathesis catalysts to O₂ and other species may be pronounced [1].  O₂ can decompose catalysts, so it may be best to degas (deoxygenate) the solvent before charging catalyst.  Instead of sparging to remove O₂, heating a solvent to reflux and cooling under N₂ is more effective [8], probably because boiling breaks the surface tension of the liquid.  By heating to reflux some solvent may be distilled off, potentially azeotroping off H₂O, which can decompose the catalyst.  Metatheses can be driven by boiling the solvent or by purging with N₂ to remove low-molecular weight olefins formed; ethylene has been shown to decompose the Grubbs I catalyst bis(tricyclohexylphosphine)benzylideneruthenium dichloride [9].  Gradually adding the catalyst while gradually adding the starting material helped drive a metathesis [10].   Vigorous agitation and not filling a reactor completely so that there is significant headspace above the reaction mixture may also be helpful.  Basic impurities such as morpholine [11] and DBU can deprotonate and thus decompose an active catalyst [12].  Impurities that might chelate Ru can slow metatheses, as GSK found with a urea and a beta-hydroxy secondary amide arising from opening the corresponding beta-hydroxy phthalimide with iPrOH.  Those researchers found that impurities can act synergistically [13].  Some tips on removing Ru species are discussed in an earlier Process Tip on this website.

Materia Inc. was founded in partnership with Prof. Grubbs and Caltech to expand the use of metathesis.  Users can now purchase metathesis (pre)catalysts from Materia to make commercial APIs without negotiating a license.  Materia also offers process guidance and other technical services [14].  Materia’s business model should facilitate pharma’s use of metathesis for preparing APIs.

Olefin isomerization and subsequent metathesis can generate impurities that are difficult to remove, such as 1) olefins from isomerization of the metathesis product; 2) olefins arising from loss of a CH₂, due to olefin isomerization away from the terminal position followed by metathesis; and 3) olefins including an additional CH₂, from isomerization of an internal olefin towards the end of the chain [6, 15].  Ruthenium hydride species, formed by decomposition of metathesis catalysts, have been shown to be responsible for isomerized side products [16, 17].  Additives such as AcOH or benzoquinone [18], 2,6-dichlorobenzoquinone [10], tricyclohexylphosphine oxide [19] and the monophenyl ester of phosphoric acid [20, 21] have been shown to reduce olefin isomerization, presumably by quenching the hydride species.  These additives may also slow or stop metatheses from proceeding to completion.  Lower reaction temperatures can also prevent isomerization [22].  The use of these and other reagents for minimizing olefin isomerization has been reviewed [23, 24].

The research group of Professor Fogg has investigated the mechanisms for catalyst decomposition.  They have shown that Lewis donors can promote decomposition of Grubbs catalysts, with the generation of MePCy₃Cl from the PCy₃ ligand; such Lewis donors include pyridine, MeOH, THF, and H₂O [25].  In an RCM with Grubbs II (IMes) reagent, only isomerized products were found in reactions run in dimethoxyethane, while reactions in 1,2-dichloroethane exhibited little isomerization [19]; the ability of DME to act as a Lewis donor may have promoted the decomposition of the catalyst.  The second-generation Hoveyda catalysts (PCy₃-free) can be decomposed by reacting with (basic) amines, generating Ru hydride species [26].  Ruthenium nanoparticles have also been shown to isomerize olefins, and these nanoparticles can be produced by decomposition of the Grubbs II reagent.  Removal of these nanoparticles did not halt isomerization, as further catalyst decomposition produced more nanoparticles [27].

In a recent development, a 2-isopropyl-6-methyl-CAAC [cyclic alkyl amino carbene] Ru complex has been shown to produce low levels of isomerization in metatheses.  For instance, an RCM at 150 ºC produced 42% of isomerized side products with a second-generation Hoveyda-Grubbs catalyst, but only 13% of isomerization using the CAAC catalyst.  Experimentally and by calculations this complex was shown to be resistant to the formation of Ru hydride species [28].  While this catalyst was quite effective in the presence of ethylene [29], more exploration is warranted on the potential impact of other impurities on this relatively new catalyst, which may not yet be commercially available.

Hence it is important to avoid decomposing the catalyst by controlling the impurities in the starting materials and the reaction mixtures.  Recrystallizing the starting materials and purifying the solvents may be necessary for rugged reactions [10, 13].  Minimizing side products from olefin isomerization in a metathesis is reminiscent of minimizing protodeboronation from a Suzuki reaction: to minimize the opportunity for side reactions to occur, the best approach may be to find conditions that lead to rapid conversion to the desired product.

I thank Professor Deryn Fogg and Dr. Philip Wheeler for helpful discussions.

1. Higman, C. S.; Lummiss, J. A. M.; Fogg, D. E. Chem. Int. Ed. 2016, 55, 3552.
2. Wheeler, P.; Phillips, J. H.; Pederson, R. L. Process Res. Dev. 2016, 20, 1182.
3. Virtual issue on metathesis: see Fogg, D. E. Organometallics, 2017, 36, 1881.
4. Olefin Metathesis: Theory and Practice; Grela, K., Ed.; Wiley 2014.
5. Handbook of Metathesis; Grubbs, R. H.; Wenzel, A. G.; O’Leary, D. J.; Khosravi, E., Eds.; Wiley-VCH, Weinheim, Germany; 2015.
6. van Lierop, B. J.; Lummiss, J. A. M.; Fogg, D. E., Chapter 3: “Ring-Closing Metathesis,” in Olefin Metathesis: Theory and Practice, First Edition; Grela, K., Ed.; John Wiley & Sons, Inc.; 2014; pp 85 – 152.
7. Yee, N. K.; Farina, V.; Houpis, I. N.; Haddad, N.; Frutos, R. P.; Gallou, F.; Wang, X.; Wei, X.; Simpson, R. D.; Feng, X.; Fuchs, V.; Xu, Y.; Tan, J.; Zhang, L.; Xu, J.; Smith-Keenan, L. L.; Vitous, J.; Ridges, M. D.; Spinelli, E. M.; Johnson, M. Org. Chem. 2006, 71, 7133.
8. Königsberger, K.; Chen, G.-P.; Wu, R. R.; Girgis, M. J.; Prasad, K.; Repic, O.; Blacklock, T. J. Process Res. Dev. 2003, 7, 733; Rassias, G.; Hermitage, S. A.; Sanganee, M. J.; Kincey, P. M.; Smith, N. M.; Andrews, I. P.; Borrett, G. T.; Slater, G. R. Org. Process Res. Dev. 2009, 13, 774.
9. Burdett, K. A.; Harris, L. D.; Margl, P.; Maughon, B. R.; Mokhtar-Zadeh, T.; Saucier, P. C.; Wasserman, E. P. Organometallics 2004, 23,
10. Kong, J.; Chen, C.; Balsells-Padros, J.; Cao, Y.; Dunn, R. F.; Dolman, S. J.; Janey, J.; Li, H.; Zacuto, M. J. Org. Chem. 2012, 77, 3820.
11. Less than 20 ppm of morpholine was found in the batch of toluene used as solvent for a metathesis. Under the dilute conditions for the RCM about one equivalent of morpholine was present relative to the catalyst.  Nicola, T.; Brenner, M.; Donsbach, K.; Kreye, P. Process Res. Dev. 2005, 9, 513.
12. Ireland, B. J.; Dobigny, B. T.; Fogg, D. E. ACS Catal. 2015, 5,
13. Wang, H.; Goodman, S. N.; Dai, Q.; Stockdalem, G. W.; Clark, W. M., Jr. Process Res. Dev. 2008, 12, 226; Wang, H., in Transition Metal-Catalyzed Couplings in Process Chemistry; Magano, J.; Dunetz, J., Eds.; Wiley-VCH; 2013; pp 233 – 251.
14. http://pharma.materia-inc.com/process-development
15. Reaction of a metallacyclobutane intermediate with a base can also generate an olefin impurity with an additional CH₂: see reference 11.
16. Schmidt, B. J. Org. Chem., 2004, 9, 1865.
17. Courchay, F. C.; Sworen, J. C.; Ghivirga, I.; Abboud, K. A.; Wagener, K. B. Organometallics 2006, 25,
18. Hong, S. K.; Sanders, D. P.; Lee, C. W.; Grubbs, R. H. Am. Chem. Soc., 2005, 127, 17160.
19. Bourgeois, D.; Pancrazi, A.; Nolan, S. P.; Prunet, J. Organomet. Chem. 2002, 643-644, 247.
20. Formentín, P.; Gimeno, N.; Steinke, J. H. G.; Vilar, R. Org. Chem. 2005, 70, 8235.
21. Firth, B. E.; Kirk, S. E. US 9481627 B2, 2016 (to Elevance Renewable Sciences, Inc.).
22. Pederson, R. L.; Fellows, I. M.; Ung, T. A.; Ishihara, H.; Hajela, S. P. Synth. Catal. 2002, 344, 728.
23. Stoianova, D., 2009:http://allthingsmetathesis.com/inhibiting-olefin-isomerization/
24. Discussion Addendum for: Ring-closing Metathesis Synthesis of N-Boc-3-pyrroline: O’Leary, D. J.; Pederson, R.; Grubbs, R. H. Synth. 2012, 89, 170; DOI: 10.15227/orgsyn.089.0170.
25. McClennan, W. L.; Rufh, S. A.; Lummiss, J. A. M.; Fogg, D. E. Amer. Chem. Soc. 2016, 138, 14668.
26. Bailey, G.; Lummiss, J.; Foscato, M.; Occhipinti, G.; McDonald, R.; Jensen, V.; Fogg, D. Am. Chem. Soc. XXXX, XXX, XXXX; DOI: 10.1021/jacs.7b08578.
27. Higman, C. S.; Lantera, A. E.; Marin, L. M.; Scaiano, J. C.; Fogg, D. E. ChemCatChem 2016, 8, 2446.
28. Butilkov, D.; Frenklah, A.; Rozenberg, I.; Kozuch, S.; Lemcoff, N. G. ACS Catal., 2017, 7, 7634; DOI: 1021/acscatal.7b02409.
29. Marx, V. M.; Sullivan, A. H.; Melaimi, M.; Virgil, S. C.; Keitz, B. K.; Weinberger, D. S.; Bertrand, G.; Grubbs, R. H. Chem. 2015, 127, 1939.

Removing residual ruthenium down to levels acceptable for pharma has been recently reviewed by Wheeler and coworkers at Materia [1], using both extractive and adsorptive methodology, and by the Fogg group [2].  For Ru the key consideration is to control the amount of residual Ru at a permissible daily exposure limit (PDE) in the drug substance of NMT 100 micrograms / day, if given orally [3].  In addition, it is necessary to control the amount of residual reagent in the product, so a convenient assay for any reagent used is desirable.  For operations on scale inexpensive and readily available reagents are desirable, and purification should be performed without preparative chromatography.

There are three main approaches to removing metal compounds [4].  These approaches include precipitating or crystallizing the metal impurities, solubilizing metal species away from the drug substance (including crystallizing the API or intermediate), or adsorbing metal species onto an insoluble reagent.  General approaches are to prepare an insoluble metal salt such as an oxalate and remove it by filtration [5] or to treat a reaction mixture with trimercaptotriazine to precipitate divalent metal species [6, 7].  Dithiocarbamates have also been found to effectively remove Ru by precipitation [8].  2-Mercaptonicotinic acid [9] and imidazole [10] have been used to quench metatheses and to extract the Ru species away from the products of metathesis; these reagents have chromophores that could permit ready quantitation of residues in the API by HPLC.  Treatment with an aqueous solution of cysteine has removed Ru from a process stream [11]; since cysteine is an essential amino acid, permissible levels of this contaminant in an API should be fairly high.  Ethylene diamine has been added to quench a RCM and remove Ru [12].  Controlling the pH may be critical for using any aqueous solution.  For the third approach, hydrogenation can remove Ru, probably by adsorption of the Ru hydride species [13] to the activated carbon provided with a Pd/C catalyst [14].  Treatment with activated carbon, such as Ecosorb® 908, has also been effective [15].  Polymeric scavenging reagents may sequester Ru species; a consideration with polymeric reagents is that over time the metal species can leach from the resin (“break through”). When optimizing processes it is helpful to remember that by minimizing the charge (wt/wt) of the metal-containing reagents the burden of removing related impurities can be reduced.

Tris(hydroxymethyl)phosphine has been used to remove Ru from intermediates for drug substances, but for pharma some caution is due.  In 1999 the Grubbs group applied this reagent for removing Ru species, using 10 or more equivalents of P(CH₂OH)₃ [16]; in other cases, 25 or more  equivalents have been required for efficient removal of Ru [17].  At a catalyst loading of 5 mol%, for instance, 1.25 equivalents of P(CH₂OH)₃ relative to starting material might be required, and even with the high solubility of P(CH₂OH)₃ in water complete removal of the reagent might be difficult.  Treatment of the less expensive tetrakis(hydroxymethyl)phosphonium chloride [18] with aqueous NaOH generated P(CH₂OH)₃, HCHO, and a N-methylol impurity of a secondary amide [11].  As an alkylating agent the latter could be flagged by regulatory authorities as a potential mutagenic impurity [19].  P(CH₂OH)₃ is a P-methylol and could also be flagged as a potential mutagenic impurity.  No mutagenicity or carcinogenicity data have been reported for P(CH₂OH)₃ [20].  The reactivity of P(CH₂OH)₃ towards amines has been documented: in 1999 P(CH₂OH)₃ was shown to react with glycine in H₂O at RT to form the tris glycine adduct P(CH₂NHCH₂CO₂H)₃ [21], and P(CH₂OH)₃ has been used to crosslink polypeptides in aqueous solution [22].  Since many drug candidates are amines, some impurities could be generated upon contact with P(CH₂OH)₃.  Because P(CH₂OH)₃ has no strong chromophore, detection and quantitation of residual P(CH₂OH)₃ may be difficult by HPLC with standard UV detectors, and detection by GC may require derivatization.  Hence researchers in pharma may prefer to avoid using P(CH₂OH)₃ for removing metal ions near the end of a route to prepare a drug substance.

As indicated in the paragraph above, many reagents that have been developed and employed for removing Ru are not amenable for preparing APIs on scale, but others may have promise.  For instance, the Paquette group published that addition of Pb(OAc)₄ (1.5 eq. relative to the Grubbs catalyst) removed colored byproducts from metathesis [23]; most researchers in pharma will not want to complicate their workups by adding Pb to the reaction streams.  Ethyl 2-isocyanoacetate and potassium 2-isocyanoacetate have been used to quench metatheses, but plug chromatography was required [24] and these reagents are relatively expensive .  Diethylene glycol monovinyl ether also efficiently quenches metatheses, but silica gel purification was required to remove the Ru species [25]; this reagent is relatively inexpensive, and might find adaptation for pharma.

I thank Dr. Philip Wheeler and Dr. David Snodin for helpful discussions.

References

1. Wheeler, P.; Phillips, J. H.; Pederson, R. L. Process Res. Dev. 2016, 20, 1182.
2. Higman, C. S.; Lummiss, J. A. M.; Fogg, D. E. Chem. Int. Ed. 2016, 55, 3552.
3. ICH Q3D Guideline for Elemental Impurities, December 2014: http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q3D/Q3D_Step_4.pdf
4. Anderson, N. G., Practical Process Research & Development – A Guide for Organic Chemists; 2^nd edition; Academic Press: San Diego; 2012; Chapter 11.
5. For example, Cu(II) can be removed by filtration as the oxalate: Sedergran, T. C.; Anderson, C. F. US 4,675,398, 1987 (to E. R. Squibb & Sons). Oxalate salts have also been effective for removing Pd.
6. Rosso, V. W.; Lust, D. A.; Bernot, P. J.; Grosso, J. A.; Modi, S. P.; Rusowicz, A.; Sedergran, T. C.; Simpson, J. H.; Srivastava, S. K.; Humora, M. J.; Anderson, N. G. Process Res. Dev. 1997, 1, 311.
7. Malmgren, H.; Bäckström, B.; Sølver, E.; Wennerberg, J. Process Res. Dev. 2008, 12, 1195.
8. Gallagher, W. P.; Vo, A. Process Res. Dev. 2015, 19, 1369.
9. Yee, N. K.; Farina, V.; Houpis, I. N.; Haddad, N.; Frutos, R. P.; Gallou, F.; Wang, X.; Wei, X.; Simpson, R. D.; Feng, X.; Fuchs, V.; Xu, Y.; Tan, J.; Zhang, L.; Xu, J.; Smith-Keenan, L. L.; Vitous, J.;Ridges, M. D.; Spinelli, E. M.; Johnson, M. J. Org. Chem. 2006, 71, 7133.

10. Brenner, M.; Meineck, S.; Wirth, T. U.S. Patent 7,183,374 B2, 2007 (to Boehringer Ingelheim).
11. Wang, H.; Matsuhashi, H.; Doan, B. D.; Goodman, S. N.; Ouyang, X.; Clark, W. M., Jr. Tetrahedron 2009, 65, 6291.
12. Puentener, K.; Scalone, M. US 7939668B2, 2011 (to Roche Palo Alto LLC).
13. Louie, J.; Bielawski, C. W.; Grubbs, R. H. Am. Chem. Soc. 2001, 123, 11312.
14. Wang, H., in Transition Metal-Catalyzed Couplings in Process Chemistry; Magano, J.; Dunetz, J. R., Eds.; Wiley – VCH; 2013; pp. 233-251; Wang, H.; Goodman, S. N.; Dai, Q.; Stockdale, G. W.; Clark, W. M., Jr. Process Res. Dev. 2008, 12, 226.
15. Welch, C. J.; Leonard, W. R.; Henderson, D. W.; Dorner, B.; Childers, K. G.; Chung, J. Y. L.; Hartner, F. W.; Albanese-Walker, J.; Sajonz, P. Process Res. Dev. 2008, 12, 81.
16. Maynard, H. D.; Grubbs, R. H. Tetrahedron Lett. 1999, 40,
17. Pederson, R. L.; Fellows, I. M.; Ung, T. A.; Ishihara, H.; Hajela, S. P. Synth. Catal. 2002, 344, 728.
18. Safety note: in water, decomposition of this phosphonium chloride may generate phosphine (PH₃) in addition to HCHO and HCl: https://pubchem.ncbi.nlm.nih.gov/compound/Tetrakis_hydroxymethyl_phosphonium_chloride
19. ICH M7 Guidance, May 2015: http://www.fda.gov/ucm/groups/fdagov-public/@fdagov-drugs-gen/documents/document/ucm347725.pdf
20. LD₅₀ Oral – rat – 178 mg/kg: http://img1.guidechem.com/msdspdf/2767-80-8.pdf
21. Berning, D. E.; Katti, K. V.; Barnes, C. L.; Volkert, W. A. J. Am. Chem. Soc. 1999, 121, 1658.
22. Lim, D. W.; Nettles, D. L.; Session, L. A.; Chilkoti, A. Biomacromolecules 2007, 8, 1463.
23. Paquette, L. A.; Schloss, J. D.; Efremov, I.; Fabris, F.; Gallou, F.; Méndez-Andino, J.; Yang, J. Lett. 2000, 2, 1259.
24. Galan, B. R.; Kalbarczyk, K. P.; Szczepankiewicz, S.; Keister, J. B.; Diver, S. T. Org. Lett. 2007, 9, 1203.
25. Liu, W.; Nichols, P. J.; Smith, N. Tetrahedron Lett. 2009, 50, 6103.

Benzene is a highly toxic compound, and long-term exposure to benzene can cause anemia, leukemia, and other medical conditions [1].  The ICH has classified benzene as a Class 1 solvent with a limit of NMT 2 ppm [2], and people in pharma and pharma-related industries are very concerned about controlling and minimizing the levels of benzene in their products.  Benzene and 50 other solvents of the four ICH classes have been screened by headspace GC to detect residual solvents as impurities in APIs [3].

Benzene has been used to make ethyl benzene, styrene, aniline, cyclohexane, cumene, phenol, chlorobenzene, and other simple feedstocks.  So perhaps we shouldn’t be surprised when benzene is found in our APIs, reagents, and solvents.  Although benzene could accidentally contaminate a drum of material, manufacturers catering to the pharma market are probably very alert to sources of benzene contamination in their products.  It is more likely that benzene is present in solvents or reagents, or is formed as an artifact.

Benzene in solvents Toluene is commercially available with less than 0.1 wt% of benzene [4].  Benzene can be separated from toluene by distillation; a high reflux ratio is key [5].

Acetone can contain low levels of benzene.  Acetone is primarily manufactured through alkylation of benzene with propylene to afford cumene, followed by generation of cumyl hydroperoxide, and scission through the Hock reaction to generate phenol and acetone [6].  Hence unreacted benzene could contaminate acetone manufactured from cumene.  A minor route to acetone is that from oxidation of propylene; in principle benzene could be generated as a minor byproduct under these conditions.  (Possible reaction: 2 propylene + 1.5 O₂ produces benzene + 3 H₂O.)  Due to similar volatility of benzene and acetone, separating acetone from benzene is difficult, and azeotropes may be employed [7].

Absolute ethanol can be produced by adding benzene, distilling off the EtOH – benzene – water azeotrope, and then distilling off the benzene – EtOH azeotrope to remove benzene; of course, not all the benzene is guaranteed to be removed [8].  Ethanol can be denatured with many compounds, including benzene.  EtOH denatured with benzene can be labeled as S.D.A. 2-B, S.D.A. 2-C, or S.D.A. 12-A. [9].  Denatured fuel ethanol can contain gasoline, toluene, xylene, and benzene [10].  Hence if denatured EtOH is used in manufacturing it is prudent to confirm that supplies are not contaminated by benzene.

Benzene in reagents Benzene was found in benzenesulfonic acid used to crystallize an NCE [11].

Benzene formed as an artifact Sometimes questioning the method of sample preparation / analysis can be informative.  For instance, benzene was formed from the preservative sodium benzoate when soft drinks containing ascorbic acid were heated at 100 ºC for 30 min [12].  Some control reactions may be needed to ensure that artifacts are not generated during sample preparation and analysis.

Bottom line We are well-advised to look at our reactions to anticipate if benzene could be present or generated in our process streams.  Only a small amount of benzene could derail the development or sale of a drug.  Nature doesn’t care if we think a “good reaction” is needed to generate an impurity.

1. http://www.atsdr.cdc.gov/PHS/PHS.asp?id=37&tid=14#bookmark05
2. http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q3C/Step4/Q3C_R5_Step4.pdf
3. Li Qin, L.; Chang-qin Hu, C.; Yin, L., “Establishment of a Knowledge Base for Prescreening Residual Solvents in Pharmaceuticals”, Chromatographia 2004, 59(7),
4. http://www.citgo.com/WebOther/CITGOforYourBusiness/Petrochemicals/Toluene.pdf, dated January 2016
5. http://seperationtechnology.com/benzene-distillation
6. Weissermel, K.; Arpe, H.-J. Industrial Organic Chemistry; 3d ed.; VCH: Weinheim; 1997; pp 276-7, 342-3, 353-5.
7. Berg, L., “Separation of Benzene from Acetone by Azeotropic Distillation,” US 4,931,145 (1990).
8. http://cool.conservation-us.org/waac/wn/wn28/wn28-3/wn28-304.pdf. “…smart chemists know to stay away from the lab punch made with absolute ethanol.”
9. https://www.law.cornell.edu/cfr/text/27/21.151
10. http://itecref.com/pdf/MSDS_Fuel_Alcohol_ITEC_Refining.pdf
11. Gross, T. D.; Schaab, K.; Ouellette, M.; Zook, S.; Reddy, J. P.; Shurtleff, A.; Sacaan, A. I.; Alebic-Kolbah, T.; Bozigian, H. Process Res. Dev. 2007, 11, 365.
12. Hileman, B. Chem. Eng. News 2006, 84(17), 10.

Residual metal ions can cause significant problems with operations. For instance, hydroxylamine has been found to react exothermically with Ti, Fe, Cu, Ni, Cr, and Mn, which may be components of metal reactors and fittings; to avoid contamination by metal ions a process involving hydroxylamine was developed using a glass microreactor (1). Contamination by 15 ppm of residual copper ions in new lines decomposed some of the performic acid charged in a process (2). MnO₂ and Fe(III) each react with H₂O₂ (3, 4). Pt is used for hydrogenation, but if residual Pd is present hydrogenolysis may also occur. Removing residual metal ions from reactors can be essential to ensure safe operating conditions and completion of reactions.

Many ions are more soluble under acidic conditions than under basic conditions, as for FeCl₃ vs. Fe₂O₆. Organic acids are often used to solubilize metal ions. ∝-Hydroxy acids are more acidic than the non-hydroxylated analogues (5); carboxylic acids have often been used to clean reactors. ∝-Hydroxy acids may also be good ligands because they can form stable 5-membered rings in octahedral coordination complexes (6).

Citric acid has been used to solubilize ions of Fe, Ca, and other metals. Citric acid forms soluble chelates of Fe, Cu, Mg, Mn and Zn (7). Citric acid has been used synthetically to solubilize Ti(IV)(8), Os(VIII) (9), and ruthenium (10). A citric acid (0.3 M) – oxalic acid (0.2 M) combination has been used to decontaminate nuclear reactors, and oxalic acid can be used to remove MnO₂ and rust from iron (11). Malic acid was used to solubilize Al(III) (12).

Under basic conditions (or at least conditions that are not strongly acidic), chelants such as EDTA and nitrilotriacetic acid (NTA) will complex with metals. Glyphosate (Roundup) forms a 5-membered ring with ions of Mn (13). EDTA and NTA strongly complex with Ni(II) and Cu(II), and can corrode reactors and fittings (14).

Matching the metal ion, the right oxidation state, the complexing reagent, and the right pH can all be important. For instance, a solution of 2,4,6-trimercaptotriazine sodium salt was ineffective in removing Cu(I), but after the process streams were sparged with air the Cu(II) generated was effectively removed by TMT treatment. Trivalent ions such as Al(III), Cr(III), and Fe(III) were not bound to TMT (15). (Since TMT precipitates metal ions it is not appropriate for removing trace metals from reactors.) Cu(II) oxalate precipitates from solutions made strongly acidic with HCl (16).

All such reagents can be corrosive. It is wise to test for the corrosivity of solutions to glass and metal reactors by soaking representative pieces of glass or metal coupons in the solutions that could be used for cleaning.

A general procedure for cleaning new equipment has been described (17). Cleaning the surface of reactors is called pickling, and then the reactors are passivated (oxidized) to prevent corrosion of the surface. With usage sediment can accumulate in the jackets of reactors, decreasing the efficiency of heat transfer.  Some companies have provided contract services to clean the jackets of reactors (18).

1)      Vörös, A.; Baán, Z.; Mizsey, P.; Finta, Z. Org. Process Res. Dev. 2012, 16, 1717.

2)      Thanks to Paul Jass for this personal communication.

3)      Decomposition of excess hydrogen peroxide to work up an epoxidation: Vaino, A. R. J. Org. Chem.2000, 65, 4210.

4)      Swaddle, T. W. Inorganic Chemistry: An Industrial and Environmental Perspective; Academic Press; 1997, p. 252.

5)      Some pKas of selected carboxylic acids: oxalic, 1.23, 4.19; citric, 3.08, 4.74; glyoxylic, 3.2; malic, 3.46; glycolic, 3.83; acetic, 4.75; peracetic, 8.2.

6)      Swaddle, T. W. ibid, p. 245.

7)      http://www.citric-acid.us/Citric_Acid_uses.html

8)      Working up a Kulinkovich reaction: Young, I. S.; Haley, M. W.; Tam, A.; Tymonko, S. A.; Xu, Z.; Hanson, R. L.; Goswami, A. Org. Process Res. Dev. 2014, XX, XXXX; DOI: 10.1021/op500135x.

9)      For achiral dihydroxylation: Dupau, P.; Epple, R.; Thomas, A. A.; Fokin, V. V.; Sharpless, K. B. Adv. Synth. Catal. 2002, 344(3+4), 421.

10)   Cleaning nuclear reactors: Row, T. H. Nucl. Sci. Abstr. 1967, 21, 27690; Reference 7 in Couturier, M.; Andresen, B. M.; Jorgensen, J. B.; Tucker, J. L.; Busch, F. R.; Brenek, S. J.; Dubé, P.; am Ende, D. J.; Negri, J. T. Org. Process Res. Dev. 2002, 6, 42. The Pfizer researchers used acetonitrile as solvent to contain volatile and toxic RuO₂ in solution.

11)   Kumar, V.; Goel, R.; Chawla, R.; Silambarasan, M.; Sharma, R. K. J. Pharm. Bioallied Sci. 2010 Jul-Sep; 2(3), 220–238; doi:  10.4103/0975-7406.68505

12)   Workup of a Friedel-Crafts alkylation: Wu, G.; Wong, Y.; Steinman, M.; Tormos, W.; Schumacher, D. P.; Love, G. M.; Shutts, B. Org. Process Res. Dev. 1997, 1, 359.

13)   http://cabecahealth.com/wp-content/uploads/2013/05/FoodPlaguePrimer.pdf

14)   Swaddle, T. W. ibid, p. 269.

15)   Malmgren, H.; Bäckström, B.; Sølver, E.; Wennerberg, J. Org.Process Res. Dev. 2008, 12, 1195.

16)   Sedergran, T. C.; Anderson, C. F. U.S. 4,675,398, 1987 (to E. R. Squibb & Sons).

17)   Mukherjee, S. “Preparations for Initial Startup of a Process,” Chemical Engineering 2005, 112(1), 36.

18)   For instance, the OptiTherm service from GE Water and Power Technologies: https://www.gewater.com/kcpguest/documents/Capability%20Profiles%20FAQ_Cust/Americas/English/cp102.pdf.

Statistical design of experiments (DoE; also termed factorial experimental design or FED) is a powerful tool for developing processes efficiently. DoE has been used in many phases of API development, ranging from initial process optimization to design space studies in validation to assay development. In the time spanning 2003 – 2013 there were over 100 manuscripts in Organic Process Research & Development that described using DoE for various stages of API optimization. The recent review below summarizes selected publications from that group of manuscripts to show how DoE findings have been applied for process scale-up and other aspects of API development.

Weissman, S. A.; Anderson, N. G. “Design of Experiments (DoE) and Process Optimization. A Review of Recent Publications” Org. Process Res. Dev. XXXX, XX, XXXX; DOI: 10.1021/op500169m

Salts of 2-ethylhexanoic acid, inexpensive carriers of the sodium or potassium ion, have been used to make salts of carboxylic acids, such as fosinopril sodium and sodium clavulanate. Limits should be set for residual impurities such as 2-ethylhexanoic acid in APIs. The toxicity of 2-ethylhexanoic acid must be reviewed in order to set realistic limits, especially since developing highly sensitive assays can consume considerable effort [1].

CAUTION: 2-ETHYLHEXANOIC ACID IS A TERATOGEN AND REPRODUCTIVE TOXIN. WOMEN ESPECIALLY MUST TAKE SUITABLE PRECAUTIONS TO MINIMIZE CONTACT WITH THIS COMPOUND AND ITS SALTS. This SAFETY information should be made available to everyone, including chemists, QA, QC, and operators, who handles 2-ethylhexanoic acid and its salts.

Some people have avoided using 2-ethylhexanoic acid because of a reputation for being highly toxic. The question is: How Much? How toxic is it? BG Chemie has manufactured 2-ethylhexanoic acid, and stated in 2000 that it has “low acute oral toxicity (LD₅₀ rat oral 2043 to 3640 mg/kg body weight)” [2]. An EPA report in 2001 [3] described toxicity to rat embryos and fetuses with high dosing, referencing a 1987 report. The latter stated that valproic acid (2-propylpentanoic acid, isomeric to 2-ethylhexanoic acid) was about twice as potent to rat embryos and fetuses as 2-ethylhexanoic acid [4]. (Valproic acid is marketed generically as an anticonvulsant and mood stabilizer. The starting dosage for sodium valproate is 600 mg/day, increasing up to 1000 – 2000 mg/day. Contraindications are listed for women taking this drug [5].) The EPA report also cited a 1993 report that mentioned that effects on reproduction and post-natal development in rats were seen with high doses [6]. 2-Ethylhexanoic acid was examined on a cellular level in 2005 and found to increase the incidence of sister chromatid exchange [7]. In a 1998 study 2-ethylhexanoic acid was shown to be a weak embryotoxic compound in male rats and mice, and the NOEL (no observable effects limit) was 60 mg/kg [8]. A 2005 study showed that 2-ethylhexanoic acid was toxic to pregnant rabbits, and the NOEL was determined to be 25 mg/kg [9]. 2-Ethylhexanoic acid has no functional groups alerting for mutagenicity, and tested negative in vitro for mutagenicity [9]. Female mice developed some liver cancers when exposed to 2-ethylhexanol, which can be oxidized in vivo to 2- ethylhexanoic acid, and the TD₅₀ was calculated at 1650 mg/kg/day [10]. In a study with rats in which 2-ethylhexanoic acid was administered orally, 7 – 12% of ¹⁴C-labelled 2-ethylhexanoic acid was detected in the feces, indicating that about 90% of 2-ethylhexanoic acid was orally absorbed [11].

Several approaches could be taken to set permissible daily exposure (PDE) limits for residual 2-ethylhexanoic acid in drug candidates. One could propose that the PDE limit should be the same as for residual 2-ethylhexanoic acid in clavulanate, thus minimizing discussions with regulatory authorities. Another, sound approach is to calculate a limit for 2-ethylhexanoic acid based on toxicology data and applying the calculations discussed in ICH Q3C (R5) [12]. Using those calculations, the PDE limits for a 50 kg patient are calculated at 12 mg/day from the study in rats and mice (F1, F2, F3, F4, F5 = 5, 10, 5, 1, 1), and 50 mg/day from the study in rabbits (F1, F2, F3, F4, F5 = 2.5, 10, 1, 1, 1). The more conservative approach would be to propose the lower limit, a PDE of 12 mg/day for a 50 kg patient, for a drug substance administered orally. For other means of administration a limit could be proposed at 90% of the oral dosing, or 10.8 mg/day.

The toxicity of residual 2-ethylhexanoic acid and its salts should of course be acknowledged in reviews with the FDA and other regulatory authorities. Dionex has developed a procedure to quantitate residual 2-ethylhexanoic acid, which may be better than the USP assay for 2-ethylhexanoic acid in clavulanate [13].

Many thanks to Dr David Snodin for his helpful comments.

1. Snodin, D. J. Org. Process Res. Dev. 2014, 18, XXXX (doi.org/10.1021/op5000368)
2. http://www.bgrci.de/fileadmin/BGRCI/Downloads/DL_Praevention/Fachwissen/Gefahrstoffe/TOXIKOLOGISCHE_BEWERTUNGEN/Bewertungen/ToxBew275-E.pdf (2000)
3. http://www.epa.gov/hpv/pubs/summaries/metalcarb/c14172rr12.pdf
4. E. J. Ritter, E. J.; W. J. Scott, W. L., Jr., Randall, J. L.; Ritter, J. M. Teratology 35(1), 46 (February 1987).
5. http://www.medicines.org.uk/emcmobile/medicine/26115/spc
6. Pennanen, S.; Tuovinen, K.; Huuskonen, H.; Kosma, V. M.; Komulainen, H. Fundam Appl Toxicol. 1993 Aug;21(2), 204.
7. Song, J.-Y. ; Cho, Y.-H. ; Kim, Y.-J. ; Chung, H.-W. Environmental Mutagens and Carcinogens 2005, 25(3), 110.
8. Juberg, D. R., et al., 1998; see http://www.contactalimentaire.com/fileadmin/ImageFichier_Archive/contact_alimentaire/Fichiers_Documents/Avis_de_AESA/cef_ej959-963-966-968_22list_op_en.pdf
9. http://www.vkm.no/dav/c067aa526e.pdf (Norwegian Scientific Committee for Food Safety)
10. http://toxnet.nlm.nih.gov/cpdb/chempages/2-ETHYLHEXANOL.html
11. http://toxnet.nlm.nih.gov/cgi-bin/sis/search/r?dbs+hsdb:@term+@DOCNO+5649
12. http://www.gmp-compliance.org/guidemgr/files/Q3C_R5_STEP4.PDF
13. http://www.dionex.com/en-us/webdocs/88443-AN262-IC-2EthylhexanoicAcid-Clavulanate-01Oct2010-LPN2608-R2.pdf

Carbodiimides are widely available and relatively inexpensive reagents that are often used to generate amides from the corresponding amines and carboxylic acids.  Such carbodiimides include dicyclohexylcarbodiimide (DCC), diisopropylcarbodiimide (DIC), and N-ethyl –N’-(3-dimethyaminopropyl)-carbodiimide (available as free base or HCl salt; a.k.a. EDC, EDCI, EDAC, water-soluble carbodiimide, and WSC).  EDC can couple amines and carboxylic acids in predominantly aqueous solutions,[1] and is also widely used as a crosslinking reagent in biochemistry.[2]

Carbodiimides are known to be relatively toxic.  For instance, EDC is irritating to the skin and may cause severe eye damage.[3]  DCC is a known skin sensitizer.[4]  Exposure to DIC led to temporary blindness to a worker who cleaned up a spill of 1 L while wearing a respirator and gloves.[5]  On a molecular level EDC has been shown to crosslink double-stranded DNA, which could be responsible for cell death;[6] little has been published on the genotoxicity of carbodiimides.  People should undertake precautions to avoid contact with carbodiimides, and to quench reactions containing carbodiimides as part of workups.

Treatment with carboxylic acids or aqueous acids can decompose carbodiimides, thus affording effective quenches.  Oxalic acid, acetic acid, and phosphoric acid have been used to quench DCC.[7]^,[8]  The decomposition of EDC in aqueous solutions has been studied quantitatively: decomposition occurred in minutes at pH 4.0, and was said to be instantaneous at pH 2.8 (1% aq. AcOH) and pH 2.2 (1% aq. HCO₂H).[9]  Using the kinetic data in the latter paper it may be possible to develop a QbD argument for the decomposition of EDC.