Natural Products
Separation and purification

SummaryTen core conclusions

Most failures in natural product separation arise before the column rather than in the chromatographic conditions. Raw material contains water, fats, pigments and tannins, and the target is often below one part per thousand. This part deals with three things: judging the properties of the target, choosing the extraction method, and getting the extract into a state fit to load.

1Five property assessments before separation

The choice of separation method is determined by the physicochemical properties of the target. Published textbooks generally list five, and these five together determine the extraction solvent, the pretreatment route and the chromatographic mechanism.

2Extraction methods and how to choose between them

2.1 Solvent extraction

Based on the solvent's ability to dissolve the target, this is the most broadly applicable class. The five basic operations trade off temperature, time and solvent volume differently.

2.2 Other extraction techniques

3Pre-column work: partition cuts and impurity removal

3.1 Gradient partition cuts

Loading the whole extract directly is generally uneconomical. The usual practice is a solvent gradient partition first, cutting the system into several fractions by polarity, and then running chromatography only on the fraction containing the target. The usual sequence for a plant extract is petroleum ether (or n-hexane) → dichloromethane (or chloroform) → ethyl acetate → n-butanol → water, corresponding respectively to fats and low-polarity terpenoids, low- to medium-polarity aglycones, medium-polarity phenolic acids and flavonoid aglycones, saponins and flavonoid glycosides, and sugars and inorganic salts.

For components with ionizable groups, pH gradient partitioning provides an extra dimension of selectivity: adjusting the pH of the aqueous phase switches the target between its free and salt forms, transferring it directionally between the two phases. This is commonly used for coarse separation of alkaloids and organic acids.

3.2 Removing common impurities

This part is the main body of the text. Six compound classes are set out for each of the four sources, 24 in all, with the representative compounds, material sequence and practical difficulties for each. Materials are given by class; the mapping to specific grades and specifications is in Chapter 22. Colour key:SilicaAluminaPolyamideBonded silicaMacroporous resinIon exchangeGelMembrane separation

The structures in sections 4.1 / 5.1 / 6.1 / 7.1 are rendered precisely by RDKit from SMILES, with molecular formula and ring counts verified individually; they show only the skeleton and key functional groups, and stereochemistry follows the original literature.

4Plant sources: six compound classes

Origin and extraction entry point

Origin　Roots, rhizomes, bark, leaves, flowers, fruits and seeds of angiosperms and gymnosperms.
Extraction entry point　Maceration, percolation, decoction, reflux and Soxhlet extraction; volatile components go through steam distillation or supercritical CO₂ extraction.
Shared pretreatment　A petroleum ether–ethyl acetate–n-butanol–water gradient partition is usually run before the column; material containing tannins and chlorophyll needs these removed first.

4.1 Representative skeletons

The structures below determine how each compound class behaves in chromatography: the number and position of hydroxyls set the strength of hydrogen bonding adsorption, ionizable groups determine whether ion exchange and pH partitioning are available, conjugation and aromatic rings determine the selectivity of a phenyl phase, and the density of sugar and phenolic hydroxyls determines how much it tails on silica.

Menthol · monoterpene

A single secondary hydroxyl and no conjugation. Low polarity and highly volatile, retained on normal-phase silica; losses at the concentration and distillation steps exceed those in chromatography.

Artemisinin · sesquiterpene lactone

Contains a peroxide bridge (—O—O—) and a lactone ring, both of which may open on exposure to base or high temperature. Avoid basic alumina and prolonged heating throughout the separation.

Diosgenin · spirostane type

A steroid nucleus with a spiroketal side chain and only one hydroxyl at position 3. The aglycone itself is low in polarity and runs on silica; once glycosylated the polarity jumps and it moves to macroporous resin and reversed phase.

Quercetin · flavonol

Five phenolic hydroxyls, of which 3′ and 4′ form a catechol. Polyamide orders compounds by the number of phenolic hydroxyls and their ortho relationships, a dimension reversed phase cannot reproduce.

Berberine · quaternary alkaloid

The quaternary nitrogen carries a permanent positive charge, so retention in reversed phase is weak and tailing on bare silica is marked. Enrich by cation exchange first and add a basic modifier in reversed phase.

Gallic acid · the hydrolysis unit of gallotannin

A pyrogallol with a carboxyl. Polyhydroxyphenols are the main source of irreversible adsorption on silica and must be removed beforehand by polyamide or precipitation.

5Marine sources: six compound classes

Origin and extraction entry point

Origin　Sponges, gorgonians and soft corals, ascidians, echinoderms, macroalgae and their symbiotic and epiphytic microorganisms.
Extraction entry point　Fresh material is homogenized or freeze-dried and then extracted repeatedly with methanol and dichloromethane.
Shared difficulty　A high-salt matrix. Residual salt weakens reversed-phase retention, suppresses ion exchange capacity and interferes with mass spectrometry, so desalting generally comes before the fine separation step.

5.1 Representative skeletons

The difficulties with marine material cluster at two extremes: at one end a high-salt matrix and strongly polar nitrogen- and sulfate-containing groups, barely retained in reversed phase; at the other long conjugated polyenes and halogenated aromatics, sensitive to light and oxygen.

Astaxanthin · carotenoid

A C40 long conjugated polyene with a hydroxyl and a ketone at each end. Sensitive to light and oxygen and prone to oxidative discoloration on silica, so work in the dark, control the temperature and keep contact time short.

2,4-dibromophenol · bromophenols

Bromine substitution raises hydrophobicity and molecular weight, so reversed-phase retention is stronger than for a phenol of the same carbon number; the bromine also makes the isotope pattern easy to identify by MS.

L-fucose-4-sulfate

The repeating unit of fucoidan. The sulfate group makes the polysaccharide strongly negatively charged, which is the direct reason for using DEAE and strongly basic anion exchange.

Guanidino · a strongly polar nitrogen group

A common fragment in marine guanidine alkaloids. Very strongly basic and fully protonated at physiological pH, so essentially unretained in reversed phase; ion exchange or HILIC is needed instead.

6Animal sources: six compound classes

Origin and extraction entry point

Origin　Amphibian secretions, reptile venoms, insects and arthropods, molluscs and annelids, mammalian glands and bile.
Extraction entry point　Fresh tissue is homogenized or freeze-dried and extracted with dilute acid in ethanol; bound forms can be released enzymatically.
Shared requirement　Keep everything cold and control the pH; peptides and proteins need protease inhibitors to prevent autolysis and degradation.

6.1 Representative skeletons

Separation of animal-derived material often has to meet purity and activity targets at the same time, and the acid–base groups, lactone rings and sulfate groups in the structures set the usable and unusable condition windows.

Ursodeoxycholic acid · free bile acid

A steroid nucleus with a C24 carboxyl. Free and taurine- or glycine-conjugated forms coexist with a wide polarity difference; where necessary hydrolyse first to a single form before separating.

Bufogenin · cardiac steroid

A six-membered α-pyranone lactone ring at C17, which opens on exposure to base. Avoid basic alumina and alkaline mobile phases, and control pH throughout.

Octopamine · polar phenolamine

A primary amine with a phenolic and an alcoholic hydroxyl: very polar and weakly retained in reversed phase. Enrich by cation exchange first, then finish in reversed phase or HILIC.

N-acetylgalactosamine-4-sulfate

The sugar unit of chondroitin sulfate. The degree of sulfation sets the charge density, and with it the elution salt concentration and the fractionation order in anion exchange.

7Fungal and microbial sources: six compound classes

Origin and extraction entry point

Origin　Basidiomycete fruiting bodies (Ganoderma, Trametes, shiitake), filamentous fungi (Penicillium, Aspergillus), actinomycetes.
Extraction entry point　Fruiting bodies are extracted with ethanol or hot water; fermentation systems are clarified first, then run on two parallel routes — intracellular (mycelium) and extracellular (fermentation filtrate).
Shared difficulty　Fermentation broth is large in volume and heavy in impurities. Using in-line resin adsorption immediately after clarification reduces the volume to be concentrated and the risk of emulsification.

7.1 Representative skeletons

Fungal and microbial material is dominated by three skeleton types — polyketides, steroids and phenol/quinones — and homologues often differ by only one methyl or one hydroxyl, so selectivity is obtained mainly by changing the separation mechanism rather than by adjusting the gradient.

Lovastatin · polyketide

A hexahydronaphthalene skeleton with a δ-lactone ring and a branched ester. The lactone and open hydroxy acid forms interconvert, so the form must be confirmed and standardized before chromatography.

Ergosterol · fungal sterol

A steroid nucleus with a conjugated diene in ring B and a C22 double bond, and only one hydroxyl at position 3. Low in polarity with characteristic UV absorbance, so normal-phase silica with UV detection locates it readily.

Emodin · hydroxyanthraquinone

An anthraquinone core with three phenolic hydroxyls. The number and position of the phenolic hydroxyls determine the retention order on polyamide, which does not match the reversed-phase order.

8Master matrix of material sequences for 24 compound classes

Setting the four source routes side by side shows the pattern in how materials are used: bonded reversed-phase silica appears at the end of most of the 24 classes and is the general finishing method; gels have a place both in fine separation of small and medium molecules and in fractionating macromolecules; membrane separation and ion exchange concentrate on the three macromolecular classes — polysaccharides, glycosaminoglycans and peptides/proteins; and polyamide appears only in the phenolic classes — flavonoids, polyphenols and quinones — making it a fairly specialized material.

● means the material carries a main separation step in the usual route for that compound class; ○ means optional or used under particular conditions; — means not used in the usual route. The table summarizes the material sequences of Chapters 4–7 at route level and does not exclude other combinations used in individual papers.Industry consensus

Published textbooks generally list seven stationary phase types — silica, bonded silica, alumina, polyamide, macroporous adsorption resin, polyacrylamide and gels — grouped into two mechanisms, adsorption and size exclusion. This part adds ion exchangers and membrane separation to that basis and lists modified silica separately, making ten classes in all, each with its mechanism, scope, elution systems and practical points.

9Silica and modified silica

9.1 Silica: activity grade set by water content

Silica separates by hydrogen bonding adsorption between surface silanols and the solute, which is normal phase. Adsorption strength falls as adsorbed surface water increases, and published textbooks divide activity into five grades by water content — one of the more commonly used ways of adjusting a silica column.

Particle size and operating pressure　Silica for column chromatography is commonly 100–200 mesh for atmospheric columns and 200–300 or 300–400 mesh for pressurized (flash) columns. A higher mesh number means both higher efficiency and higher resistance, and must be matched to the pressure available.

9.2 Modified silica: introducing a second interaction

Loading a metal salt or another reagent onto the silica surface gives it a specific selectivity in addition to hydrogen bonding adsorption. The common types are:

Modified silica generally has a shorter column life than conventional silica, and the loaded reagent may bleed with the eluent and contaminate fractions, so residues must be assessed in preparative work and in subsequent testing.Industry consensus

10Bonded silica: reversed phase and polar bonded

Silanization introduces alkyl or polar functional groups onto the silica surface, changing the mechanism from adsorption to partitioning (reversed phase) or altering the adsorption selectivity (polar bonded normal phase). In natural product separation, bonded silica carries most of the final polishing work.

11Alumina and polyamide

11.1 Alumina: three pH grades

Alumina adsorbs more strongly than silica and is made in basic, neutral and acidic grades according to the pH at manufacture, each suited to different targets.

Alumina that is too active readily causes irreversible adsorption and catalytic decomposition, and generally needs adjusting to grade III–IV by water content before use.Industry consensus

11.2 Polyamide: ordering by phenolic hydroxyls

Polyamide forms hydrogen bonds between its amide groups and the solute's phenolic hydroxyls, so the elution order relates directly to the number and position of those hydroxyls. This selectivity dimension is hard to reproduce in reversed-phase chromatography, which is why polyamide has kept its place in flavonoid and polyphenol separation.

12Macroporous adsorption resins and ion exchangers

12.1 Macroporous adsorption resin

Porous microspheres built on a crosslinked polystyrene or polyacrylate skeleton, adsorbing organics from the aqueous phase through van der Waals forces, hydrogen bonding and a molecular sieving effect. In natural product processes it mainly performs enrichment and desalting, and is the usual starting point at kilogram scale.

12.2 Ion exchangers

Working by reversible ion exchange, these are divided by exchange group into cationic (strongly acidic sulfonic, weakly acidic carboxymethyl CM) and anionic (strongly basic quaternary ammonium, weakly basic diethylaminoethyl DEAE) types. The skeleton may be polystyrene, cellulose, dextran or agarose; polystyrene types are mostly used for small molecules, and cellulose, dextran and agarose types for macromolecules, to avoid exclusion limits.

13Gels, polyacrylamide and membrane separation

13.1 Dextran gels

The Sephadex G series is built on crosslinked dextran and works mainly by size exclusion. The naming corresponds to swelling volume: G-15 needs about 1.5 mL of solvent per gram of dry gel and G-100 about 10 mL, with an overall separation range of roughly 10–100,000 Da.

Sephadex LH-20 is the hydroxypropylated form of G-25 and swells in organic solvents, applying to a molecular weight range of about 100–4000 Da. Its behaviour is not pure size exclusion but also involves adsorption and partitioning, which gives it two fixed positions in natural product separation: removing chlorophyll and polymers beforehand, and final polishing to reduce losses with a small sample.

13.2 Polyacrylamide gels

The Bio-Gel P series is an uncharged, low-hydrophobicity size exclusion medium with a particle size of about 45–180 μm, graded by molecular weight range (P-10 covers about 1,500–20,000 Da, for example). It suits carbohydrates, peptides, tannins and other components prone to non-specific adsorption on other media. These gels swell considerably in water, and in practice are mostly used with alcohol–water systems above 20% rather than with pure water.

13.3 Membrane separation and molecular distillation

Membrane processes involve no phase change, operate at low temperature and use relatively little energy, which suits concentration and desalting of heat-sensitive components. The main limitations are membrane fouling and flux decline, which need a cleaning and regeneration scheme. Molecular distillation separates by short-path evaporation under high vacuum and suits high-boiling, heat-sensitive fat-soluble components, with applications in essential oil refining and lipid separation.

14Quick reference to the ten material classes

Chromatography is not the whole of separation work. In most natural product processes, the extraction, precipitation and crystallization carried out before chromatography achieve most of the volume reduction and impurity removal, and chromatography only comes in where selectivity is insufficient. This part follows the order of operations, covering the classical separation methods, thin layer chromatography, the scale ladder in column chromatography, and preparative HPLC.

15Before chromatography: extraction, fractionation, precipitation and crystallization

16Thin layer and preparative thin layer chromatography

Thin layer chromatography (TLC) plays two roles in natural product work: rapid screening of column chromatography conditions, and direct preparative work on small samples.

The choice of visualization reagent relates directly to the compound type (vanillin–sulfuric acid for terpenoids and steroids, aluminium trichloride for flavonoids, Dragendorff's reagent for alkaloids, anisaldehyde–sulfuric acid for glycosides). Establishing a detection system matched to the target is a prerequisite for judging fractions.Industry consensus

17The particle size – pressure – resolution ladder

The three scale bands in column chromatography are set by particle size, which simultaneously determines the pressure needed, the resolution obtainable and the time per batch. The three cannot all be optimized at once.

Resolution runs HPLC > MPLC > LPLC and time per batch runs LPLC > MPLC > HPLC. The figures are from published textbooks and vary with column length, diameter and system.Industry consensus

18Preparative HPLC and scale-up

18.1 The order in which a method is established

A preparative method is generally converted from an analytical one, in the order: establish selectivity on an analytical column → establish loading and the collection window on a semi-preparative column → scale up on a preparative column. The first step, selectivity, decides the outcome; optimizing loading and flow rate cannot make up for insufficient selectivity.

18.2 Engineering constraints on scale-up

As the scale rises, the mechanical strength of the medium, bed stability and packing method become the main variables. A preparative column can be packed by dynamic axial compression (DAC), keeping the bed consolidated by continuous axial pressure and reducing bed collapse and loss of efficiency over long use. Particle size band, sphericity and the width of the size distribution all affect packing and column pressure.

Packing diagnostics and acceptance criteria are covered at DAC Packing SupportandDAC Packing Problem Diagnosis; analytical column packing is atAnalytical Column Packing。

Once the materials and methods are fixed, what remains is sequencing them. The logic of sequencing in natural product separation comes down to one rhythm and four general rules; this part gives both, with six published routes for comparison.

19The four separation stages

20Four general selection rules and orthogonal combinations

21Six published separation routes reviewed

The six routes below are taken from published reports, to show how the general rules above appear in practice.

All six routes are taken from the experimental sections of published reports; only the stationary phase and solvent system are quoted here, and the gradients, column specifications and yields follow the original papers.Literature example

The first five parts describe materials by class. Moving to actual purchasing and process development, the class has to be resolved into pore size, particle size, carbon load and scale band. This part gives the correspondence between the ten material classes and the product lines on this site, a one-page reference, and what this text does not cover.

22From material class to Microwants products

The table maps the ten material classes of Part Three onto the product lines available on this site. The correspondence is drawn on separation mechanism and does not amount to a conclusion about suitability for a specific sample; actual selection must still be confirmed against the properties of the target and the scale.

23One-page reference: target compound → material sequence

24Data basis and scope

The empirical ranges in this text are taken from published textbooks and vary with sample complexity, medium lot and operating conditions; replace them with measured values before use.

AppendixPrincipal source categories and evidence grading