1. RP HPLC Purification of Small Molecules
Lou Cheng
Good Morning Everyone!!!
First I like to thank Paul, Kerri, Marvin, and everybody else to give me a chance to present here. Today, I like to talk about “RP HPLC purification of small molecules”. This work was part of my work at Astrazeneca or AZ in the past year, and it was the part I have permission to present out of AZ.
Astrazeneca R&D Boston

2. Outline Reversed-Phase HPLC – Isocratic and Gradient
HPLC Basics: Classification, k, α, N, Rs
Reversed-Phase HPLC – Isocratic and Gradient
Analytical Method Development – Screening & Optimization
Analytical Scale-up of Optimized HPLC Method
From Analytical Scale-up to PrepLC Purification
PrepLC Purification – Fraction Collection & Recovery
Summary
This is also a chance for me to share you my understanding and experience on this topic. I would appreciate any questions from you at any time during this presentation.
First I like spend one minute or two talking about some HPLC basics, so that you can better understand my presentation. After that I will use some real Astrazeneca examples to explain principles of both isocratic and gradient RP HPLC. Then I will show you how I apply these basics and principles in the screening and optimization of the analytical HPLC method, as well as in analytical scale-up. After that I will discuss how to go from analytical to preparative and the fraction collection and recovery in prepLC separations.

3. HPLC Basics - Classification
Increasing polarity
Water-insoluble
Water-soluble
Nonpolar
Ionic
Nonionic polar
Molecular Weight
102
103
104
Liquid Chromatograph can be classified into four categories according to the separation mechanism: Adsorption, Partition, ion exchange, and size exclusion. This classification does not include chiral and affinity chromatography. Normally, molecules with less than 1000 Dalton molecular weigh are considered as small molecules.
105
106

4. HPLC Basics - Classification
Increasing polarity
Water-insoluble
Water-soluble
Nonpolar
Ionic
Nonionic polar
Molecular Weight
102
103
104
My experience focused on the Gel Permeation Chromatography for macromolecules and adsorption/partition for small molecules. Reversed-Phase High-performance Liquid Chromatography, or RP HPLC is the topic of this talk.
105
106

5. HPLC Basics – k, α, N, Rs tR1 tR2 t0 w1 w2 tRi - t0 k = t0 k2 α = k1
Minutes 10
mVolts
The retention factor or capacity factor, describes how long the molecule was retained on a column. Theoretically, it is the ratio of the number of molecules in the mobile phase to the number of molecules in the stationary phase. Optimum Retention Factor k for isocratic HPLC is 2 < k < 20 , corresponding to retention time between 1.2 min and 13.2 min for 5 cm HPLC columns.
The selectivity or separation factor evaluates how well two peaks are separated. α larger than 1.1 is normally required for analytical purpose. For an analytical method developed for PrepLC separation, we desire selectivity as big as possible.
Efficiency/Plate number is a measure of how narrow peaks are in relation to how long the compound is retained. Plate number is a leading indicator of how much the column could be overloaded. Plate number less than two thousand is usually unacceptable according to FDA guideline. For an analytical method developed for PrepLC separation, we desire plate number as big as possible.
Resolution is a comprehensive measure of separation by considering all retention factor, selectivity, and column efficiency. The purpose of our method optimization is to maximize resolution for peaks of interest.
Retention factor
tRi - t0
k = t0 k2
α = k1
Selectivity
N = 16 2 t Wi
Efficiency
Rs = 0.25 (-1)N0.5 k
k + 1
Resolution

6. Reversed-Phase HPLC – Isocratic
log k = log kw – Sφ
Eq. 1*
k: retention factor
kw: retention factor by 100% water as mobile phase (φ = 0)
S: a constant for a given sample compound
φ : organic fraction (volume) in binary mobile phase
Kw: Hydrophobicity
RP HPLC could be operated under isocratic or gradient conditions. In isocratic, the composition of the mobile phases remains constant during separation. The most often used mobile phase combinations are water-ACN or water-methanol. The retention in isocratic is defined by this equation over some practical range.
Kw and S are sample’s HPLC properties. They can be used to predict retention under both isocratic and gradient conditions.
As a chemist, I tend to relate the Kw to the hydrophobicity of the sample, and S to the retention sensitivity of the sample to the change of mobile phase. Higher the Kw, more hydrophobic the sample is. High the S, more sensitive the sample is to the change of mobile phase.
S: Sensitivity to change of mobile phase strength
*Ref: Practical HPLC Method Development, Lloyd Snyder, etc. Wiley, New York, 1997.

7. Isocratic RP HPLC - AZ Example 1
Conditions: HPChem 10, XBridge C8, 4.6 × 50 mm,
10 mM NH4Form/AcN, 1.0 mL /min
Rule of Adjusting Isocratic Retention by φ: 5%  Ξ 100 % k
Log k = log kw – Sφ
ID:
(BCL2)
(MW = 426.5)
(MW = 454.3)
Code:
(CoaD)
Here are two real AZ examples. In this figure, triangles are for the logarithm of retention factor, and circles are for the retention factor. As you can see, both cases fit this linear relationship over the range studied. Judged from Kw and S, blue compound is more hydrophobic and more sensitive to the change of mobile phase than the red one. Besides, in both case, 5% increase in organic percentage would cause ~100% decrease in retention.

8. Isocratic RP HPLC - AZ Example 2
log k = log kw – Sφ
, 10 g (HE-TMK)
Kw = 28
S = 3.35
Kw = 62, S = 3.67
Kw = 72, S = 3.72
Mw = 163
Mw = 158 +
In the case of isomers, the Kw and S of the isomers are almost identical, as red and green samples indicated here. The starting material had a very different Kw and S and its prep separation by RP HPLC from its isomeric product was very easy. However, the prep separation of isomers was proven undoable, and their final separation was done by normal phase using pure silica column.
Conditions: XBridge C18, 4.6 × 50 mm, 0.1% NH4OH/MeOH
Flow Rate: 1.0 ml /min, room temperature.
The Kw and S of the two isomers are too close to be separable by RP gradient runs.

9. Reversed-Phase HPLC – Gradient
b = SVm/tGF log k0 = log kw - S0
tR = (t0/b) log [2.3k0b(ts/t0) + 1] + ts + tD
Eq. 2*
T0: column dead volumn; b: gradient steepness; k0: k at the beginning of the gradient;
ts: value of tR for a nonretained solute; tD: dwell time for gradient elution;
: change of organic percentage in the mobile phase; S: system constant;
Vm: column dead volumn; tG: gradient time; F: flow rate.
For Linear Solvent Strength (LSS) Gradient:
*Ref: Practical HPLC Method Development, Lloyd Snyder, etc. Wiley, New York, 1997.
*By assuming S = 4 for all small molecules, one gradient run is sufficient to resolve kw
Two gradient runs can solve kw and S, by assuming log k/φ linear relationship
Automatic method development
Retention Prediction (Drylabs, Chromsword)
In gradient mode, the organic percentage changes normally linearly with time, the retention is given by this equation:
The three important parameters in the equation are b, gradient steepness, k0, retention factor at the beginning of the gradient which is determined by the starting organic percentage in the mobile phase, and delta phi, the total change of organic percentage in the mobile phase during the run. These three can be used in gradient method optimization.
By assuming S = 4 for all small molecules, one gradient run is sufficient to resolve kw; Two gradient runs can solve Kw and S. These two assumptions are protocols adopted by Drylabs or Chromsord software for automatic method development. Here I like to show you two real examples.

10. Retention Prediction of Gradient RP HPLC - AZ Examples
tR = (t0/b) log [2.3k0b(ts/t0) + 1] + ts + tD
Sample b  tG
tR, predicted
tR, Exp
Error (%)
(%/min)
(%)
(min.)
(BCL2)
(Kw = 933, S = 4.53)
6.0
35-65 5
2.55
2.56
-0.3
12.0
35-95
2.28
2.35
-3.0
(CoaD)
(Kw = 2290, S = 7.85)
9.0
50-95
3.00
2.82
+6.4
40-70
4.41
4.42
-0.2
Here the value Kw and S for sample 197 and sample could be from two gradient runs or isocratic runs. The predicted retention for gradient runs agrees well with the experiment one, and the prediction error is acceptable.
Conditions: HPChem 10, XBridge C8, 4.6 × 50 mm, 10 mM NH4Form/AcN, 1.0 mL /min, room temperature.
t0 = ts = 0.6 min, tD = 1.2 min.

11. Reversed-Phase HPLC – Gradient Over Isocratic
Why Gradient?
Flexibility (b, , 0) to optimize separation with minimal effects on efficiency (N)
Samples with a wide k range, sometimes containing late-eluting interferences that can either kill the column or carryover to subsequent runs
More precise, robust, and automatable
Dilute solutions of sample dissolved in a weak solvent
We routinely start our method development with gradient RP HPLC. Why? One reason I mentioned before is that we have more tools available to optimize separation in gradient that in isocratic. Besides, optimization by organic percentage in isocratic has more adverse effects on column efficiency.
Samples we received normally have a wide k range, containing very hydrophilic and very hydrophobic components, and the later sometimes are late-eluting interferences that could kill the column or carryover to subsequent runs. No isocratic conditions can give acceptable k for all of these components. While under gradient mode, these components can elute under a full gradient, and thus provides protection for the columns and for the sequent runs.
Also, isocratic retention is more sensitive to temperature, solvent mixing by the pump, and its precision, robustness, and automatability under prep conditions are not as good as gradient ones.
An extrra advantage for gradient HPLC is that you can prepare your samples in a weaker solvent than the starting gradient such as 100% water.
So, gradient RP run is always the best starting point for method development, even if the final method was isocratic or normal phased.
Gradient RP run is the best starting point for method development

12. RP HPLC Analytical Method Development – Screening
Solvents
Column Examples
B: ACN
C: MeOH
D: THF/H2O (9/1)
1: XBridge C18
2: Gemini C6-Phenyl
3: Atlantic dC18
4: YMC ODS AQ
5: Synergi Hydro-RP
6: YMC Carotenoid (C30)
●●●●●●
E: 0.1% TFA
F: 0.1% Formic Acid
G: 10 mM NH4Ac pH8
H: 0.1% TEA
J: 0.1% NH4Form
K: 0.1% NH4OH
L: 10 mM NH4Ac/HOAc pH5
M: 20 mM NH4Ac
N: 10 mM (NH4)2CO3
So gradient RP screening is normally the first step after we receive purification request. The screening Agilent 1100 is equipped with a 12 solvent valve and a 6 column selector, and MS detector, ELSD, and DAD for peaking tracking purpose. Selection of solvents and columns depends on sample’s charge, hydrophobicity, solubility, and molecular weight. Examples of aqueous solvents with volatile buffers are. Among them, E, F, L are acidic, G, J, M, N are neutral, and H,K are basic. The organic solvents are normally AcN, MeOH, and THF. The gradient is normally from 5-95 %.
Column examples are. Column 1 and 2 can be used under all acidic, neutral and basic conditions. Column 3 to 6 can only be used under acidic and neutral conditions.
By this setting, 162 conditions can be screened for one sample if necessary. Sometimes 2nd or 3rd screening is needed for the best results.
Valves: G1159A 6-ColSelector, G1160A, 12/13 Selvalves
Detectors: Agilent MSD and Sedex 75 ESDL Detector, Finnigan AQA mass spectrometer by closed contact
Waters MicroMass Massspectrometer
NP, chiral, SFC have similar settings
162 conditions can be screened for one sample if necessary

13. Criteria for Evaluating and Optimizing HPLC Methods
General:
Low k, low tailing factor, high N
High α, high Rs
Client-specific:
●●●●●●
MPS/library (universal applicability)
Fraction collection for one component, multiple components, or all components?
Purity/Recovery?
pH stability of the desired components?
Amount of the sample (high loading)
How to evaluate the screening results? How to optimize HPLC methods after screening?
Generally, we look for low k, low tailing factor, and high N for peaks of interest. For critical band pairs, we look for high α, and high resolution.
Besides, there are some client-specific criteria. If the method is developed for QC, resolution of all peaks with sufficient symmetry and efficiency is required. For reaction monitoring, baseline resolution of reactants and products is required. In this talk, the purpose of our method development is purification. So are we requested to collect one component or multiple components? Request of one component collection would cost less, compared to multiple component collection.
What is the purity/recovery requested? The answer would determine how to collect fractions and how to combine fractions.
Some clients may tell us their samples are unstable under acidic or basic conditions. In that case, we would avoid methods using these conditions.
If the sample is from a library, then the method developed should have the potential to fit all library samples.
If the amount of sample is more than 1.0 gram, we would focus on the method with the highest loading capacities.
Some times these chemist-specific criteria keep changing. So, clear and timely communication with chemists is the prerequisite to a successful method development.
Clear communication with clients is a prerequisite to successful method development

14. AZ Example of Screening Sequence (2252-026)
Here is a real example of intelligent screening sequence for sample It is intelligent because the first standard injection is used as a control sample to check system suitability. The sequence would stop automatically if it fails system suitability test.

15. From Screening To PrepLC
(QuinFF)
0.8 mg/ml, XBridge C18 (4.6 × 50 mm)
5-95% MeOH/10 mM HCOONH4, 5 minutes, 1.0 ml/min, 240 nm.
Tracked by MSD * Rs
1.8
2.2
Optimization
30-80% MeOH Rs
3.6
4.8
Anal. Scale-up
100 mg/ml, 4.6 × 100 mm XBridge C18
30-80% MeOH, 10 min, 1 ml/min
PrepLC
19 × 100 mm XBridge C18
30-80% MeOH, 10 min
20 ml/min, 100 mg/ml
Baseline Resolution
For this sample, this method, 10 mM AmForm/MeOH with XBridge C18, stood out, because it gave the best resolution and separation factor for this critical band pairs. Besides, the product peak, as tracked by MSD detector, was symmetrical and has the highest plate number.
The client wanted to collect the product peak only. So by fine-tuning the starting gradient, gradient steepness, and gradient range, we were able to maximize the resolution of these two critical band pairs from 1.8 to 3.6 and 2.2 to 4.8 respectively.
We then further scaled up this separation on analytical HPLC and finally PrepLC with baseline resolution.

16. From Screening To PrepLC
Scale-up/loading
100 mg/ml, 4.6 × 100 mm XBridge C18
30-80% MeOH, 10 min, 1 ml/min
Screening
(QuinFF)
0.8 mg/ml, XBridge C18 (4.6 × 50 mm)
5-95% MeOH/10 mM HCOONH4, 5 minutes, 1.0 ml/min, 240 nm.
Tracked by MSD * Rs
1.8
2.2
Optimization
30-80% MeOH
3.6
4.8
PrepLC
19 × 100 mm XBridge C18
30-80% MeOH, 10 min
20 ml/min, 100 mg/ml
Baseline Resolution
For this sample, this method, 10 mM AmForm/MeOH with XBridge C18, stood out, because it gave the best resolution and separation factor for this critical band pairs. Besides, the product peak, as tracked by MSD detector, was symmetrical and has the highest plate number.
The client wanted to collect the product peak only. So by changing the starting gradient, gradient steepness, and gradient range, we were able to double the resolution of these two critical band pairs from 1.8 to 3.6 and 2.2 to 4.8 respectively.
We then further scaled up this separation on analytical HPLC and finally purified the desired product on Gilson prep system with baseline separation.

17. Analytical Scale-up of Optimized HPLC Method
Goal: 1) Is the optimized analytical HPLC method good for preparative one?
2) If it is, what is the maximum loading for touching-band separation?
Optimization EN
Synergi Hydro-RP (4.6 × 50 mm)
30-60% MeOH/TFA, 5 min,
1.0 ml/min, 240 nm, 0.6 mg/ml.
Product (MW =471.5)
Anal. Scale-up
Synergi Hydro-RP (4.6 × 100 mm)
30-60% MeOH/TFA, 10 min,
1.0 ml/min, 240 nm, 12.5 ul, 160 mg/ml.
Product
The screening and optimization were always done under analytical concentrations of the sample, which is normally below 1.0 mg/ml. We want to know if the optimized method could be used under preparative conditions. This could be done through analytical scale-up experiments. In analytical scale-up, the sample is normally hundred times concentrated than analytical one. Because of this, we have to replace analytical flow cell by prep flow cell. Besides, we always try to use columns with the same length as the prepLC columns. These changes make analytical scale-up conditions very close to prep conditions.
The analytical scale-ups can answer following two questions: 1) is the optimized analytical HPLC method good for preparative one? 2) if it is, what is the maximum loading for touching-band separation?
Sometimes, the optimized analytical is good for the prep separation, as indicated in the previous slide. But very often, the optimized analytical is not that good for prep. Here comes an example. In this optimized method, the resolution from the best screened method were already maximized. However in its scale-up, the efficiency and resolution were not optimized, and the touching-band loading was only 12.5 ul. We have 5 grams in 30 ml solvents, and it means it would take 125 injections for the prep separation. Clearly, this method is not practical for 5g separation.
This method is not practical for separation of 5 gram samples! (125 prep injections for
19 × 100mm column!)

18. Scale-up of Optimized HPLC Method – Search for the Best
Optimization
Anal. Scale-up
Gemini C6-Phenyl (4.6 × 50 mm)
50-70% MeOH/NH4OH, 5 min,
1.0 ml/min, 240 nm, 0.6 mg/ml.
Gemini C6-Phenyl (4.6 × 100 mm)
50-70% MeOH/NH4OH, 10 min,
1.0 ml/min, 240 nm, 12.5 ul, 160 mg/ml.
XBridge C18 (4.6 × 100 mm)
30-60% CH3OH/NH4OH, 10 min,
1.0 ml/min, 240 nm, 12.5 ul, 160 mg/ml.
XBridge C18 (4.6 × 50 mm)
30-60% CH3CN/NH4OH, 5 min,
1.0 ml/min, 240 nm,
5ul inj, 0.6 mg/ml
I then have to continue optimization and scale-up experiments first with Gemini C6-phenyl column, and resolution is still not satisfactory. I further tested XBridge C18, and it seemed resolution and efficiency are much better than previous two, although the screening results by XBridge was not the best.

19. Gilson PrepLC via UV-triggered Fraction CollectionEN
XBridge C18 (50 × 250 mm)
30-60% CH3CN/NH4OH, 25 min,
100 ml/min, 240 nm, 6.0 ml, 160 mg/ml.
Only 5 injections!
The final prep separation was done on a 50 X 250 mm XBridge C18 column, with only 5 injections.

20. Scale-up of Optimized HPLC Method – Theoretical AspectsW TR
W2 = W02 + Wth Eq. 3
= 16 N-1 t02 ( k) t02 k2 w ws-1
As we mentioned before, in analytical scale-up experiments, the sample is hundred-times concentrated than the analytical one, and the column is often under overloading conditions. In the ideal case of mass overloading and partition mechanism, the bandwidth W of the overloading peaks can be expressed by this equation:
In other words, the maximum loading was determined by retention factor and plate number before overloading, and the maximum fronting distance after overloading due to excessive sample weight. When the front of this desired peak first reached the base of this impurity peak, we call it touching-band separation. This equation allows the prediction of maximum loading for touching-band separation under PrepLC conditions.
Just a reminder here: Overloading in real life is complicated, may include both mass overloading and volume overloading, and other phenomena, under which Eq. 3 may not be applicable.
(column effect)
(sample-weight effect)
Ws: column saturation capacity ( ≈ 0.4 surface area)
Overloading in reality may include mass overloading,
volume overloading,
and others.

21. Loading of Analytical Scale-up - Theoretical vs. Practical
Optimized Analytical *
Product
(BCL2)
Xbridge C8 (4.6 × 50mm)
20-50% CH3CN/NH4Ac (pH 8)
5min, 1 ml/min, 254 nm, ~1 mg/ml
5 μl injection Rs
3.5 *
Scale Up
(BCL2)
Xbridge C8, 4.6 × 100mm
20-50% CH3CN/NH4Ac (pH 8)
10 min, 1 ml/min, 254 nm, ~70 mg/ml
Inj Vol: 20 μl (Vth = 90 μl)
For example, here optimized method gave the Rs of 3.5 for this critical band pair. Theoretical prediction by Eq.3 said we could inject 90 ul and still see separations of this critical band-pair, or touching-band separation. But in our experiment, 20 ul injection already resulted the disappearance of this critical band pair.
So, what is the maximum loading of touching-band separation for this sample? This question can only be answered by analytical scale-up experiments.
What is the maximum loading of touching-band separation?

22. Analytical Scale-up of Optimized HPLC Method - Mass Loading Studies
(BCL2)
Xbridge C8, 4.6 × 100mm
20-50% CH3CN/NH4Ac (pH 8)
10 min, 1 ml/min, 254 nm, ~70 mg/ml
Inj Vol: 5 μl, 10 μl, 15 μl, 20 μl
5 μl
10 μl
15 μl
20 μl Rt *
From this loading studies, we found the touching-band loading for this sample is only 15 ul. More than that would cause excessive fronting of the product peak and the merge of this impurity peak. * *
10 μl: Baseline separation
15 μl: Touching-band separation
20 μl: No separation

23. From Analytical Scale-up to PrepLC
F = (dprep/danal)2 x Lprep/Lanal
Constant:
Column chemistry
Particle size
Sample concentration
To scale:
Flow rate
Injection volume
4.6 × 100 mm
19 × 100 mm
50 × 100 mm
VTB, 1 ml/min
17 VTB, 17 ml/min
118 VTB, 118 ml/min
× Fsemi
× Flarge
(BCL2)
XBridge C8,
20-50% CH3CN/NH4Ac (pH 8),
10 min, 20 ml/min,
~70 mg/ml, GilsonTM LC
Baseline Separation
200 (255) μl, 20 (17) ml/min, 19 × 100 mm
15 μl, 1 ml/min, 4.6 × 100 mm
XBridge C8
20-50% CH3CN/NH4Ac (pH 8)
10 min, 1.0 ml/min
~70 mg/ml, Agilent 1100
If your analytical scale-up did not encounter any problem, your PrepLC won’t have any problem either. The scale-up factor for flow rate and injection volume for PrepLC can be calculated by this equation.
For example, here we already found the touching-band (TB) injection volume for this sample is 15 ul. The scale-up factor from 4.6 by 100 columns to 19 by 100 columns is 17. So the calculated prep injection volume on Gilson would be 255 ul, and flow rate 17 ml/min. The final prep separation was done with 200 ul injection at 20 ml/min, and we were still able to see baseline separation in this case.

24. PrepLC Purification - UV-triggered Fraction Collection
XBridge C18 (50 × 250 mm)
30-60% CH3CN/NH4OH, 25 min,
100 ml/min, 240 nm, 6.0 ml, 160 mg/ml.
There are two ways to trigger the fraction collection in PrepLC. One is UV based and the other is MS based. This is an example of UV-triggered fraction collection. We also analyzed each fractions of this peak by injecting each of them back into analytical HPLC.

25. PrepLC Purification - Fraction Analysis & Recovery
As you can see from this figure, all fractions were pure except the first one had a minor shoulder. From this calibration curve, the recovery is 97% if combining all fractions and 91 % without first fraction. The molecular ions of all fractions were 472.6, thus confirmed the specificity of the fractions.
97 % Recovery (91 % without first fraction)

26. PrepLC Purification - MS-triggered Fraction Collection
In mass-triggered fraction collection, the fraction collection is triggered by desired molecular weight. In this case, the fraction collection was triggered by appearance of molecular ions of

27. PrepLC Purification - MS-triggered Fraction Collection
This is the zoom-in part of the previous fraction collection region. As you can see, one advantage of the MS-triggered fraction collection is that it can collect the desired mass part only, while neglecting the co-eluted part in the same peak. The final recovery is only 80%, which is lower that UV-triggered fraction collection.
Recovery 80 %

28. Anal. Scale Up
Summary
Gradient RP analytical run is the best starting point for developing PrepLC method
Screening, optimization, and scale-up are effective steps toward PrepLC method development┐
The best analytical methods are not always the best PrepLC methods, and scale-up experiments are imperative to validate the performance and loading of the analytical method under PrepLC conditions
UV-triggered fraction collection has high recovery and lower purity than MS-triggered fraction collection.
Screening
Optimization
Anal. Scale-up
Prep LC
Here comes the summary.

29. Acknowledgment
Members of Analytical Group: Tatyana, Camil, Nancy, Mark, Sharon,
Milena, Ziling.
Randstad USA: Yushen Chang, Vincent Cianciaruso.

30. Thank You for Your Time and Attention!

31. Analytical Method Screening - Results
XBridge C18 (4.6 × 50 mm)
5-95% NH4OH/AcN, 5 min,
1.0 ml/min, 240 nm.
Gemini C6-Phenyl (4.6 × 50 mm)
5-95% NH4OH/MeOH, 5 min,
1.0 ml/min, 240 nm.
Atlantis dC18 (4.6 × 50 mm)
5-95% NH4Fm/MeOH, 5 min,
1.0 ml/min, 240 nm.
Synergi Hydro-RP (4.6 × 50 mm)
5-95% TFA/MeOH, 5 min,
1.0 ml/min, 240 nm.
Luna C6-Phenyl (4.6 × 50 mm)
5-95% AcONH4/MeOH, 5 min,
1.0 ml/min, 240 nm.
Curosil PFP (4.6 × 50 mm)
5-95% HCOOH/CH3CN, 5 min,
1.0 ml/min, 240 nm.

32. Analytical Method Optimization
XBridge C18 (4.6 × 50 mm)
30-60% NH4OH/AcN, 5 min,
1.0 ml/min, 240 nm,
5ul inj, 0.5 mg/ml
Gemini C6-Phenyl (4.6 × 50 mm)
50-70% NH4OH/MeOH, 5 min,
1.0 ml/min, 240 nm.
Atlantis dC18 (4.6 × 50 mm)
50-95% NH4Fm/MeOH, 5 min,
1.0 ml/min, 240 nm.
Synergi Hydro-RP (4.6 × 50 mm)
30-60% TFA/MeOH, 5 min,
1.0 ml/min, 240 nm.

33. Analytical Scale-up of Optimized Analytical Methods
XBridge C18 (4.6 × 100 mm)
30-60% NH4OH/AcN, 10 min,
1.0 ml/min, 240 nm, 12.5 ul, 160 mg/ml.
Gemini C6-Phenyl (4.6 × 100 mm)
50-70% NH4OH/MeOH, 10 min,
1.0 ml/min, 240 nm, 12.5 ul, 160 mg/ml.
Synergi Hydro-RP (4.6 × 100 mm)
30-60% TFA/MeOH, 10 min,
1.0 ml/min, 240 nm, 12.5 ul, 160 mg/ml.

34. Analytical Scale-up of Optimized HPLC Method - Volume Loading
EN c
Gemini C6-Phenyl (4.6X100mm)
40-50% CH3CN/0.1% HCOOH
14 min, 1 ml/min, 254 nm, ~30 mg/ml
Inj Vol: 3 μl, 6 μl, 12.5 μl, 25 μl Rt
From this loading studies, we found the touching-band loading for this sample is only 15 ul. More than that would cause excessive fronting of the product peak and the merge of this impurity peak.

35. Effects of Buffer Concentration on Preparative Loadability
AG-166
Gemini C6-Phenyl (4.6X100mm), 0-20% CH3CN, 20 min,
1 ml/min, 254 nm, 12 ul, ~100 mg/ml
0% HCOOH
0.05% HCOOH
0.1% HCOOH
0.2% HCOOH

36. Difficult Purifications: Example 2
Screening
(CoaD)
1JB05955
1JB25955
1JB50955
Optimization
1JB60605
1JB50505
1JB40405
Scale-up/loading
4.6 × 100 mm XBridge C18
40% AcN, 60% NH4Form,
30 min, 1 ml/min, 200 mg/ml
4.6 × 50 mm XBridge C18
10mM NH4Form, 5min,
1 ml/min, 200 mg/mL
4.6 × 50 mm XBridge C18
10mM NH4Form, 5min,
1 ml/min, 200 mg/mL
Purification
19 × 250 mm XBridge C18
40 % AcN, 100 ml/min
100 min, 1.0 ml (200mg/ml)
Here is another sample which could be purified by gradient conditions but was purified under isocratic conditions.

37. Valves: G1159A 6-ColSelector, G1160A, 12/13 Selvalves
Detectors: Agilent MSD and Sedex 75 ESDL Detector
Finnigan AQA mass spectrometer by closed contact
NP Chiral HPLC
Solvent: HX, MeOH/EtOH(1/1), IPA, 0.1% Diethylamine
Detectors: Advanced Laser Polarimeter, PDR_Chiral Inc
Columns: Chiralpak AD Chiralpak OD, Chiralpak AS, Chiralpak IA, Chiralpak IB, Chiralcel OD, Chiralcel OJ,
Regis Pirkle covalent (S,S) whelk O2 10/100 FEC, Regis Pirkle covalent (S,S) whelk O1 5/100, Regis (S,S) ULMO 5/100,
Regis (S,S) DACH DNB 5/100, Phenomenex Chirex ®-PGLY and DNB,
Large 5cm X 50cm Prep Columns: Chiralpak AD Chiralpak OD, Chiralpak AS, Chiralcel OJ
NP Columns:
Luna silica 10/100, YMC-PVA-sil 5/120, YMC-Pak Diol 5/60, YMC-Pak CN 5/120, Luna NH2 5/100,
PrincetonSFC Pyridine 5/60
SFC Columns:
Berger silica, Diol, CN, Pyridine, Chiralpak AD-H

38. Method Optimization Summary for 02154-137
We run the gradient screening, and under most gradient conditions we did not any separation for the isomers. We start to see separations with narrow gradients. For example, we saw Rs of 1.8 for 20-40% gradient.
We also tried isocratic and it seemed isocratic conditions was a better choice in term of resolution. As you can see from this table, with the decrease of MeOH percentage, the isomer resolution increased. But decrease MeOH further after 20% did not improve the resolution significantly, although you see significant increase of retention.

39. Gradient and Isocratic Runs for 02154-137
Gradient Runs
5-40 % MeOH, Rs = 1.6
20-40 % MeOH, Rs = 1.8
XBridge C18, 4.6 × 50 mm, 0.1%
NH4OH, 5 min, 1.0 ml /min.
Isocratic Runs
40 % MeOH, Rs = 1.15
20 % MeOH, Rs = 2.29
10 % MeOH, Rs = 2.30
XBridge C18, 4.6 × 50 mm,
0.1% NH4OH, 1.0 ml /min.
20 % MeOH
10 %
40 %
Here are some examples of gradient chromatograms and isocratic ones, and finally we decided to proceed with 20% isocratic to the scale-up experiment.

40. From Scale-up to Gilson Separation for 02154-137
(HE-TMK)
XBridge C18, 4.6 × 100mm
20 % MeOH/0.1% NH4OH
20 min, 1 ml/min, 254 nm, 100 mg/ml
Inj Vol: 8 μl, 12.5 μl, 25 μl
Agilent HP 1100
(HE-TMK)
XBridge C18, 50 × 250mm
20 % MeOH/0.1% NH4OH
50 min, 100 ml/min, 254 nm,
Inj Vol: 1.5 (1.6) ml,100 mg/ml)
GilsonTM LS System
Here we found the touching-band separation was 8 ul. The scale-up factor from 4.6 by 100 to 19 by 250 column is about This means we inject 1600 ul in the prep column. Here is the the Gilson separation by injecting 1500 ul, which could be still considered as touching-band separation.
But we were not satisfied by the touching-band loading. Only 150 mg, and we had 10 gram to separate. Besides, retention is too long, and total separation would take 400 L mobile phase.
So we are still working on the possibility of normal phase separation.
Touching-band loading still low
Retention still too large
NP HPLC in progress!