I’m often asked to help with the development of column ‘screening’ platforms and automated development systems.  Whilst this covers a large amount of analytical science there are some common elements to this type of approach, perhaps the most important of which is column selection.  Unsurprising given that ‘selectivity’ is the most powerful tool we have in chromatography and we all know that the best way to optimise selectivity is to choose the most appropriate stationary phase.

Whilst I’ve written about column classification before: Practical HPLC Method Development Screening - as has my fellow blogger Kevin Schug: The LCGC Blog: Automated Method Development in Liquid Chromatography (August 2017) - who wrote an excellent consideration of his own experiences with column screening, I wanted to expand upon this important topic with some more pragmatic reflections and advice which I hope will be helpful to those who seek to develop a more rapid and effective approach to column selection and HPLC method development.

Whilst we all might seek a ‘single’ solution approach which will lead to a satisfactory method for every analyte encountered, in my experience this does not exist.  Think of the type and variety of analyte molecular structure, relative hydrophobicity (solubility), functional group chemistry, physico-chemistry, stereo chemistry etc. which you might encounter and this becomes obvious, unless you work with a particularly narrow and very stable range of compounds.  Further, the intended outcomes and attributes of the methods under development may also differ vastly and the requirements of a method for identification and/or quantification of impurities and degradants, will differ widely from a potency assay.  These methods may also differ in the type of detector which is available or required, in order to obtain fit for purpose data and which may also impact on the HPLC method being designed.

That being said, there are approaches which can improve the effectiveness of method development for ‘broad’ categories of method types or analytes and it is worthwhile highlighting some considerations which will dictate approaches to the implementation of screening and automated development protocols.

Method Attributes

It is necessary to define the scope and attributes of the methods being developed in order to select appropriate screening methods and crucially to know when to stop optimisation.  Most screening approaches follow the broad workflow;

  • Screen a number of stationary phases based upon the analyte types or application
  • During screening – investigate the column selectivity with different organic modifiers, buffers and perhaps pH values
  • Choose the most likely one or two combinations and further optimise the separation conditions (often with the assistance of optimisation software or calculated estimates of optimum gradient conditions / pH etc.)

Stage 3 of this process is often the most time consuming and suffers from the greatest degree of ‘overwork’.  In properly defining method attributes and having discipline in the approaches used for optimisation, a great deal of time can be saved in chasing lost causes before reverting to an alternative stationary phase.  It’s that age old problem - if you don’t know where you are going, how do you know when you have arrived?’

Typical method attributes, defined at the outset, may include;

  • Favoured modifiers, eluent additives or pH windows
  • Favoured Column Manufacturers*
  • Column Dimensions
  • Stationary phase particle size, pore size and morphology
  • Maximum run time
  • Maximum or minimum gradient time and limits of gradient slope (%B/min)
  • Minimum resolution (including resolution of peaks on API compound tails
  • Minimum peak efficiency (or maximum dispersion with UHPLC methods)
  • Peak asymmetry
  • Peak Capacity (usually estimated using simple peak width measurements)
  • Back pressure
  • Repeatability / Reproducibility

Many factors need to be considered in order to set a ‘first intent’ method specification as defined by some of the attributes above and might include;

  • Equipment available within the developing lab and in the labs where the method will be implemented
  • Availability of UHPLC equipment
  • Requirement for mass spectrometric detection in-use
  • Required throughput
  • Likelihood for the method to require extremes of pH
  • Complexity of the method – assay versus stability indicating versus impurity identification for example
  • Equipment performance (gradient repeatability, extra column volume, pressure limits, detector characteristics etc.)
  • Complexity of the resulting chromatogram and the requirement for good quantitative performance
  • The need to validate the method and the legislative recommendations for validation criteria

*  The preference for particular column manufacturers deserves a special mention.  If the method is to be used in a wide range of locations or at geographically remote sites, one may want to choose columns from manufacturers with a wide ranging distribution network and who have proven reputations for quality of production and perhaps a commitment to developing new and improved phases.  These are very important considerations, but should be tempered when considering the selectivity space being explored by the screening method, more of which later.

Analyte and Sample Properties

Several general approaches to column / method screening platforms have emerged in the past 5 years or so, and many of these use the analyte and sample properties to influence the decision on which type of platform to implement.  The properties typically considered include;

  • Range of polarity (LogP/LogD)
  • Structural similarity of the analytes
  • Structural motifs which indicate a particular column type – highly aromatic / fluorinated / halogenated etc.
  • Size / molecular weight range of analytes
  • Ionogenicity of the analytes and range of pKa values involved
  • Presence of conformers / isomers
  • Chirality

The subject of chiral phase screening is not considered here as the topic is at least one blog entry on its own!

Popular screening approaches

As a result of the analyte survey, one might choose from one of several popular approaches including;

Maximised selectivity space screen (where log P/D and analyte chemistry are widely varying or the analyte characteristics are not known in advance)

4 – 6 columns which may typically include;

  • C18
  • C8
  • PFP
  • Phenyl (Phenyl Hexyl)
  • Polar Embedded Phase
  • Aq*
  • Cyano

* Phase capable of use in 100% aqueous mobile phase and typically containing a polar endcapping or polar surface treatment)

Typically, column characterisation databases might be employed which allow the user to compare and contrast the various characteristics of each phase employed in order to maximise the selectivity space covered, although the list of ‘preferred’ column suppliers may somewhat limit the size of the available selectivity space.

There are many factors which influence the selectivity of a stationary phase and some of these are explored in various phase characterisation databases which I’ve previously written about in this free eBook:

Relating Analyte Properties to HPLC Column Selectivity – On the Road to Nirvana.

One important factor worthy of further comment here is that not all columns are created equal, and it is important (perhaps even essential) that when choosing columns for screening it should be remembered that there can be huge variation in the selectivity of phases which are nominally of the same phase type. 

Figure 1 shows the example of a 4 column wide selectivity screen where, for exactly the same nominal phases, the selectivity space (as indicated by the magnitude on the axes of the spider diagram as well as the general shape of the plots) is much broader when phases from different manufacturers are explored rather than choosing several phases from the same manufacturer.  The reasons for these differences are many fold and arise from the different silica types, bonding chemistries, surface and deactivation treatments etc. and these variations should be bourn in mind when selecting columns for screening.  This is often taken into account during secondary columns screening (see below) where variants of the same nominal phase chemistry are explored in order to further investigate the selectivity options of a column which has shown promise in a wide selectivity screen.  In this instance, column selection for the secondary screen can be greatly influenced and helped by using the screening databases in order to select columns which show marked variation in the important selectivity characteristics.

Figure 1: Spider diagram plot to indicate the relative selectivity space achieved when screening columns of nominally the same chemistry from the same manufacturer (Top) and different manufacturers (Bottom).

Specialist / Secondary screen (either after promising columns have been identified using the maximised selectivity screen or where analyte properties indicate a preference for a particular phase type)

3 – 6 columns which may include;

  • Variants of C18 columns (different silica types / mono or bidentate / surface treatments / ligand density etc.)
  • Variants of Phenyl Phases (diphenyl / biphenyl / phenyl hexyl / activated phenyl (containing oxygen atom in the alkyl spacer etc.)
  • Phases for conformer / isomer separations (high density C18 phases, PFP phases and C30 phases)

And a host of other phase variant combinations

Polar screen (reversed phase) (where poor retention is seen in the maximised selectivity screen or where analyte structure and physical-chemical data indicate the requirement for more polar interaction)

3 – 6 columns which may include;

  • Aq type phases
  • Polar Embedded phases
  • Mixed mode phases (acidic, basic or zwitterionic functionality)
  • Diol phases
  • Amide phases

Polar Screen (HILIC or Aqueous Normal Phase Mode) (where poor retention is seen in the maximised selectivity screen or where analyte structure and physical-chemical data indicate the requirement for more polar interaction)

3 – 6 columns from the following broad classifications;

  • Diol
  • Polyol
  • Amido
  • Amino
  • Ionic / zwitterionic
  • Cyano
  • Carbamate

Each of these screens may use different mobile phase combinations to fully explore the selectivity space with each stationary phase.  There are obviously a wide range of mobile phase variants which can be used to investigate the selectivity variations however here a few of the more popular combinations;

  • Screen using both methanol and acetonitrile – they selectivity differences can be very large
  • Screen over a wide range of modifier concentrations to ensure all analytes are eluted
  • 0.1% TFA v/v is a popular choice for low pH work (pH ~ 2.1)
  • 0.1% Formic Acid (FA) v/v gives a pH of around 2.7
  • Note that both of the above have virtually no buffering capacity
  • Ammonium Formate (10mM) pH 3.8 and Ammonium Acetate (10mM) pH 4.8 are useful buffers for exploring lower pH range is a useful mobile phase depending upon the analyte chemistry (if known) especially where mass spectral detection is used
  • To explore the mid pH range – 10mM Ammoniuim Bicarbonate works well
  • Ammonia (Ammonium Hydroxide) at pH 9 is useful for exploring the separation at high pH (note the column should be suitable for high pH work or column dissolution will quickly occur!)

It’s impractical to screen using all of the stationary phases with all of the eluent combinations above, therefore the choice should be based on your typical applications or a screen across a wide pH range if the analyte chemistry is unknown.

Computer aided optimisation software can be used to identify the most promising combinations of stationary phase with mobile phase composition or, if necessity dictates, one might eyeball the separations to pick out the candidate combinations for further optimisation.

From this point, the method will be further optimised and the best way to do this is to again use computer aided optimisation software or to use simple calculation to estimate the optimum gradient time, starting and ending composition. 

For the purposes of brevity I refer you to a previous blog which can be found at: Practical HPLC Method Development Screening.

For pH optimisation, if you have no access to computer optimisation software, then a series of experiments at varying mobile phase pH can be undertaken to select the optimum value to obtain the best separation selectivity.  In practical terms, the pH ‘window’ explored can be as little as 3 pH units depending upon the results from the initial screen.

As mentioned above, the key to automated or rapid screening and development approaches is knowing when to stop exploring a particular mobile phase / stationary phase combination.  Your protocol should be clear when, if no satisfactory separation has been obtained, to switch to the next most promising stationary phase and mobile phase combination or indeed to switch to a different screen altogether.