Purification process speed-and efficiency-Hayward-etal

Utilizing LC & SFC separations with
UV, ELSD & MS detection for
purification in drug discovery:

Driving toward capacity, quality,
efficiency, and rapid turnaround

Xu Zhang, David P. Budac, Qing
Ping Han, and Mark J. Hayward
Lundbeck Research USA
Paramus, NJ

The need for process
efficiency and quality
Med chemists, on average, spend ≥50% of their time on
purification, even when state of the art tools are provided.
20 medicinal chemists means ≥10 FTE effort in purification.
If 1-2 experts could do all the purification, a huge human
efficiency increase in medicinal chemistry can be realized.
Time is of the essence in drug discovery (LO cycle time).
So, turnaround time must be fast (high need: hand crafted cmpds).
Losing compounds costs a lot of Med chemist time. Thus,
quality (minimized losses) has crucial impact on efficiency.
Med Chem compounds cost ≈2k$ ea. (total FTE cost /
average # of compounds).
You can’t afford not to make every effort not to lose
them (Aim for highest success rate or at least 6 sigma).
Don’t be penny-wise (on instrument, solvent, or salary)
and then pound foolish on the big $ = compounds.
Quality and speed are crucial in every respect!
1
Lundbeck Research USA Chemistry – Analysis and Purification

Purification – needs and goals
Early Drug Discovery Full coverage (Med Chem definition)
Hit to Lead Parallel Synthesis – 40+ mg, approximately 20-100
compound/batch, every few days
Lead Optimization – hand crafted for in vivo – 100 - 500 mg, some
requiring ultra high purity, approximately 20+ compound/batch daily
Development Candidate Candidates – Up to 50 g with at least 10 g at
ultra high purity, 2 day turnaround on 20 g, 20 - 25 compounds/yr

Variety of sample quantities – 10s of mg to 10s of grams
Variety of sample qualities
Sample purities range from 5 to 95% prior to purification.
Impurities may or may not be baseline resolved according to OA-LCMS
(levers needed).*
Dissolution remains an ongoing challenge.*

Above needs define scale or capacity!
Success requires following capabilities:
100 mg per injection
20 – 120 compounds/day *Experts are crucial
≥120 injections & collections/day in addressing these
≥0.5 g/hr in full gradient mode challenges
2

Given the needs and goals there
are some significant challenges
• Operational philosophy must be simple and streamlined yet
diverse enough to cover most chemical space and keep
quality high
– Integrate Med Chem, analysis, purification, drying, and Compound
Management into the way of working
– Trusted partnership throughout the process is crucial for efficiency
• Mass per injection is 5 fold higher than norm (LC/MS based)
– Adapt off the shelf components to increase capacity
• Cycle times are 3 fold faster than norm (LC/MS based)
– Apply fast LC techniques
• FTE load must be low to achieve desired gains in efficiency
– Automate everywhere possible for tasks and transparency of data
• Med Chemists already have their own instrumentation. Why
should they come to you?
– Must be able to do it much better / faster than Med Chemists
– Must gain the trust of the Med Chemists
3

Purification Operational Philosophy
Tried and true technique: get a quality separation at
analytical scale, then scale up (50 fold in volume).
Required to achieve high success rate.
Nevertheless, must be able to adapt at prep scale
2 x 2 x 2 matrix of gradient fast analytical LC methods
adapted to the preparative scale (needs more levers):
2 columns – C18 and C8.
2 gradients – C18 gradients favor moderate LogP to polar
compounds and C8 moderate to high LogP
2 pHs – 4 and 6.5 (extremes rarely needed).
Crucial efficiency component: align with OA-LC/MS and
achieve high Med Chemist competence (simplicity)
Routine 100 mg per injection & Gaussian peaks.
Success rate as close to 100% as possible (measure
and collect waste, if needed).
Automated with optimized conditions built into
predictable, calibrated methods.
4

Operational Philosophy:
Why use analytical data and scale up?

In many cases,
only data from the HNJ_19813-127-002_03
18:38:1305-Jun-2006
2487 TUV

actual sample 100
Waste stream
3.38 An2
1.99e6

allows accurate after fraction
threshold collector when 1.86

collection works 2.16
prediction due to
%

perfectly 4.18 4.31
earlier eluting 0.58 1.47
2.64 2.95

peaks and
0
baseline rise from 0.50
HNJ_19813-127-002_03
1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
2: Diode Array
low level 1.0e-1
Collected peak
3.38
229.9
254
Range: 1.007e-1

impurities. 8.0e-2
Stream prior
6.0e-2
to fraction 1.85
Losing collector
AU

238.9 2.47
4.0e-2 2.15 228.9

compounds is not 227.9
4.18 4.32
2.0e-2 2.93 228.9 228.9
a viable option for 209.9

achieving 0.0
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
Time
5.00

increased speed. Background levels at ~15% peak height (UV & MS)
5

8 LC/MS method choices: 2 x 2 x 2 polarity
matrix (3 binary choices = simplicity)
Neutral-pH~7
Most likely:
pH choice mod LogP &
polarity,
(Ionization/Polarity) neutral pH
(for CNS cmpds)
Acid-pH~4
C8
Low polarity
Chemical space: & high LogP
More typical
pH dimension (polarity) is truly
Column choice orthogonal to column/ gradient
(Polarity) (see Thu talk)
Column/gradient choices allow
Highly polar access to extremes & optimization
(H2O soluble) in mid region
not as common
some intermediates C18
Low Organic Moderate Organic High Organic
C18 only C8 & C18 C8 only
Gradient choice (Polarity)
6

2x2x2 approach: Optimize pH & polarity
Simple technique to achieve optimal separation:
•Send molecules through column un-ionized (adjust pH of sample and mobile phase).
•Select column/gradient combo in order to achieve near 1.0 minute retention time.

Goal: sharp peaks in the middle
60% of the chromatogram
(0.4min < RT < 1.6min)
Broad Peak(s)
Solution:
Elutes too early 1. Change pH, then Elutes too late
Solutions: re-evaluate RT / polarity Solutions:
1. C8 C18 Column 1. C18 C8 Column
2. Less organic gradient 2. More organic gradient
3. Change pH 3. Change pH

7

Operational philosophy: Med Chemists
perform the pre-analysis used for purification
• Why make Med Chemists do all that work?
– Our Med Chemists perform OA-LC/MS 11 times per day (avg)
• Following reaction progress (>90%) and final products (<10%)
– They want to do it. They already do it = efficiency
• How will Med Chemists know which method to use?
– We train them
– Med Chemists can become highly competent in the use of OA-
LC/MSs choosing column, gradient, & pH (2 x 2 x 2)
– Binary choices often sufficient to allow expert results without being
expert
• Med Chemist entry point for purification
– OA-LC/MS data where one can say “I want that peak or those peaks
in bottle(s)” and structure of compound
– Highly interactive process with purification expert
– Trusted partnership throughout the process is crucial for efficiency
– Making excellent progress toward same approach with SFC
8

Challenges of 100 mg/injection
Compared to analytical scale, injection mass
increased 105 fold but mobile phase volume for
separation can increase only about 50 fold
Maintaining speed and resolution requires compensation in the
chromatographic system.
Injection process must be adapted for high load.
Adsorption and buffering capacity must be adjusted for high load.
Collection volume and separation time can limit
number of compounds collected
High separation efficiency (analytical like) must be routine to keep
collection volumes reasonable.
High separation efficiency (fast analytical like) must be routine to
keep cycle times reasonable.
Additional processes must be automated
Adapted injection process must be automated.
Extra adsorption / buffering capacity must be on-line.
Automatic column switching & regeneration.
Solvent and waste handling must be streamlined
9

Challenges of apparent mass-overload
Ionized Un-ionized
form form
100 mg
Doxylamine
(pKa 8.7): 25-
100 mg 50 mg
injection.
N
Mobile phase
temperature: 25 mg O
45°C.
°
N
Buffer: 0.2%
formate.
Mobile phase
pH = 6.5
Conventional Wisdom:
Higher mass loading kills chromatographic performance!
Thus one injects lower mass (10-20 mg: common LC/MS scale).
10

Achieve 100 mg/injection
Injection via “at column dilution”
Deliver fully-dissolved sample to column – improve mass and
volume loading
Completely separate sub-system (pump, valves), enabling
sample-dependent injection method selection from software
High separation capacity
Larger diameter column (30 mm) provides enough stationary
phase surface area to retain compounds
Higher temperature improves adsorption kinetics
Higher buffer concentration enhances buffer capacity
Completely separate buffer mixing sub-system enables buffer
capacity selection within the separation methods software
High separation efficiency
Higher temperature allows for lower back pressure and faster,
higher velocity separations
Use all other known techniques, i.e. minimize extra-column
volume, small particles (3 µm)
On-line back-flush maintains column condition (>>2000
injections) and eliminates delay time for column re-equilibrium
11

RP-LC Purification System Schematic
All components under full software control (MassLynx V4.1) [except automatic heaters]

Back-flush heater SunFire
C18
regenerating pump
XBridge
Back-flush solvent selections Column
C18
(ACN, 5% acetic acid and DMF) Makeup pump water bath
Inertsil
C8
Photo diode Inertsil
splitter C18
array
6-pos. column selectors
choose up to 6 column
Dilution solvent selections (ACN,
chemistries
50/50 ACN water mixture, etc..)
MS ELSD heater At column
dilution pump
Fraction Injection
collector port

Waste UV
heater
Concentrated buffers at 1-4 M: NH4COOH,
NH4COOCH3, CH3COOH, NH3, H2O etc…
Waste level sensor
and auto switcher
Concentrated
buffer pump
mixer

MilliQ Gradient water
B A purification and auto-
Binary pump delivery system
Waste barrels
Degassers
Acetonitrile
Lundbeck Research USA Chemistry – Analysis and Purification 12

NP-SFC Purification System Schematic
Most components under full software control (MassLynx V4.1)
Back –flush/regeneration heater AD
P50 CO2 + P50 Modifier
(awaiting software) OD Columns 5-6:
515 515 Ethyl pyridine – 5
AS Variety of columns - 6

heater OJ
Splitter
heater 6-pos. column selectors Dilution solvent selections
Photo diode choose up to 6 column (alcohols, 50/50 alcohols and CO2
Array-2996 chemistries mixture, etc..)
MS-ZQ ELSD-2420 heater At column dilution pump
(Thar analytical FDM)
Fraction Injection
collector port

SIII CO2 vent
Waste UV (2487) G
Make up PR 40 PSI Diethylamine, triethylamine, isopropylamine,
L ammonium formate, formic acid etc in alcohols
pump S
heater Concentrated buffer
Waste level sensor pump (515)
and auto switcher mixer (front panel control)

P-200 P-50
CO2 modifier
G700 with Bulk Tank
Degassers Alcohols
Waste barrels
Much of our SFC design philosophy comes from our established approach
toward RP-LC/MS based purification 13

LC System Photo: overall picture

14

LC System Photo: pumps (5) picture

15

LC System Photo: injector/collector (combo),
columns, UV detectors (2) picture

16

LC System Photo: columns & detectors
[MS, ELSD & UV (2)] picture

17

LC/MS & SFC/MS System Photos:
note on vacuum pump ergonomics

There are noise abatement solutions that work.
(required in a Danish lab)
18

SFC Instrument photo (viewed from right)

19

Instrument photo (viewed from front)

20

“At Column Dilution”
Approach for Sample Injection
Nature of injection instantaneous >10x dilution!
Choice of dilution solvent can have big impact on keeping samples in
solution as injection mass increases and can help chromatographic
performance
Goal: Deliver sample to stationary phase as individual
molecules in solution (best way: separate pump)
Flow parameters are important
Flow rate must be sufficiently high to deliver sample without introducing
band-broadening (7.5 mL/min into 100 mL/min).
Diverting at column dilution solvent after injection process can be helpful
to eliminate the effects of injection solvent on the separation (divert at 0.3-
0.5 min).
Dilution solvent composition is an untapped resource
for scaling up injection mass
100% B not always universal best choice
50/50 clearly much better in about half the cases
Binary choice (50%/50% A/B & 100%B) covers small scale (100mg inj)
Further refinement of %B and buffering can be worthwhile for larger scales
where ≥200mg injections are desired (it’s all about solubility)
21

At Column Dilution: RP-LC
Flow Rate & Divert
Elevated flow rate to
avoid wasting time and
8 mL/min 2 mL/min
minimize band-broadening
before column.
Target: injector sweep
time 15 s max, 8 s typical. 1 mL/min
Elevated flow rate also
helps prevent sample loss.
Diverting “at column
dilution” flow also can 50:50:H2O/ACN 1% acetic Acid
improve separation by
allowing dissimilar
injection and separation N

solvents. N
ACN
Improved solubility by
lowering pH eliminates
injection precipitation
Example: 200 mg imipramine
22

At column dilution: all the same applies to SFC
Example: challenging chiral resolution 100 mg injection
(all aspects mirror RP-LC, i.e. goal is solubility)
60742-048-010-boc-t79 2: Diode Array
2.24 250
Range: 1.379
1.25
100% B injection
1.0

7.5e-1
AU

2.67 Classic sign of precipitation
5.0e-1 (in addition to pressure spike):
2.5e-1 Second peak for same analyte
0.0
1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
60742-048-010-boc-t94 2: Diode Array
2.54 250
1.25 Range: 1.315

1.0
50/50 CO2/MeOH injection
7.5e-1 (best choice in about half of cases)
AU

5.0e-1 3.62

2.5e-1

0.0 Time
1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00

Already purified by RP-LC: we know that only 2 enantiomers present
23

Separation capacity and efficiency (LC & SFC)
Column particle size and diameter:
Smaller particles (3-5 µm) enhance surface area, adsorption capacity,
separation efficiency, and speed.
We have tested 3 µm particles at prep scale and they work as well as
they do in analytical scale
However, we have not found suppliers that pack 3 cm columns with 3 µm
particles consistently (but we would like to)
Thus, we use 5 µm particles for all prep scale work
30 mm diameter consistently provides enough capacity for 100 mg
injections under “infinite diameter” conditions (1 mL injections typical, 2
mL max).
Column length:
Column length is an expensive and slow way to gain resolution
Column cost approximately proportional to length
Separation time approximately proportional to length
Stationary phase (SFC) and eluent choices (RP-LC & SFC) are the
most time effective way to achieve resolution
We find 50 mm length optimum for RP-LC
Adjusting eluent conditions can be done much faster to achieve resolution at this
length (total time starts to increase at shorter lengths) More in buffering section
We find 100 mm length optimum for SFC
We haven’t found shorter chiral columns to be available (but we would
like to)
24

Separation capacity and efficiency
Temperature:
A crucial parameter that affects adsorption / desorption
rate, and thus must be properly controlled:
Four independent heaters used:
Mobile phase heater (up to 400 Watts applied, J-KEM Sci.)
Column heater (water bath kept at temp of mobile phase)
Dilution solvent heater (up to 20 Watts applied, Sererity).
Back-flush solvent heater (up to 80 Watts applied, J-KEM Sci)
Benefits of temperature control:
Improved peak shapes due to faster adsorption kinetics
fronting = missed adsorption opportunities and tailing = delayed
desorption
Significant selectivity changes also possible for SFC.
Maintain highly concentrated samples in solution. This can be
especially crucial during the injection process.
Reduces back pressure allows higher flow rates and faster runs
combination of solvent choice / temp = speed for RP-LC!
25

RP-LC Temperature Effect on Peak Shape
Reserpine
55oC
More heat! 45oC o O

= more mass transfer 35 C
25oC N N
= less fronting H

O
O O
O
O O
O
O

0.5

26

RP-LC: Temperature Effect on Speed
• To achieve resolution, one generally must have retention
– No retention = no separation
• elutes in void volume
– Must sweep multiple column volumes (k’)
• k’ must be greater than 2
• 5 < k’ < 15 very often optimal & ∆%B / k’ should be <5 (LSS)
• To get resolution fast one must sweep column volumes quickly
(k’/min ≥ 3 whereas typically k’/min ≤ 1 prep scale)
– Maximize/increase velocity (flow) while maintaining mod ∆%B / k’
– Tune temperature to match velocity
– Make good choices (required to achieve first 2):
• Column type – polymer vs. silica/BEH
• Mode of operation – isocratic (static) vs. gradient (linear sweep from A
to B)
• Solvent (B) – acetonitrile (ACN) vs. methanol (MeOH)

27

RP-LC: Velocity vs. temperature for:
elution mode and column choice
Optimal Velocity vs. Temperature
160 Fit gradient ACN - silica and BEH
Observed gradient ACN - silica and BEH
BEH or
135 Isocratic ACN silica - Guiochon etal
polymer
Separation Temperature (C)

only
(particle Iso (& grad) MeOH silica - Guiochon etal
110 stability)
Isocratic (& grad) ACN polymer - Carr etal

85 MeOH Gradient ACN
Limited stability for silica
60 High
Velocity
BEH or silica with
Range where both reduced H2O
35
analyte and silica content only That’s
stability are well (not polymer) Why
established
10 ACN!
0 5 10 15 20 25 30
Optim um Eluent Velocity (m m /s)
28

RP-LC: Scientific case for
gradient ACN operation
• Temperature offers much greater ability than other
techniques to achieve higher velocities
• Gradients with acetonitrile are truly fast!
– Optimum velocity with gradient ACN is much more
responsive to temperature elevation than any other mode
of operation
– Isocratic operation is at least 3 fold slower than gradient
(ACN)
– Other solvents (alcohols) behave like isocratic operation
even in gradient mode (still >3 fold slower)
– ACN gradients appear to be uniquely crucial to
achieving high productivity!
29

RP-LC: Acetonitrile (ACN) usage – Why?
• We use gradient ACN for purification
Let’s compare methanol use with ACN
– One third velocity or 3 fold more time (35 vs. 100mL/min @50°C)
– Half the solubility (loading) or double the number of
injections
– Total 6 fold loss in productivity! (add instruments and FTEs!)
– Double injections = double solvent = 12k$/mo
(vs. 20k$/mo for ACN) (also doubles waste volume)
– Lower organic strength = lower column lifetime (more
column cost) & less reliability
– Overnight runs = lower reliability
– Lower reliability = lower quality & lost compounds
– Must also change OA-LC/MS to MeOH (Med Chem disruptive)
• Conclusion: the expense of 6 fold loss in
productivity and lower quality would seem to
thoroughly outweigh the potential 8k$/mo in
savings on solvent 30

SFC: Effect of separation temperature
on carbamazpine peaks
carbamazepine-4
oC
40 oC oC
2: Diode Array
30 2.25
50 230
Range: 5.515
5.0

4.0

3.0
AU

2.0

1.0

Time
2.10 2.15 2.20 2.25 2.30 2.35 2.40 2.45 2.50

Peak shape doesn’t change much with increasing temperature compared
with RP HPLC condition (note opposite effect on retention). However,
temperature can still be very helpful with selectivity (see next).
31

SFC: Temperature tuning the separation

A
Temperature can
dramatically
enhance B
selectivity,
sometimes in
unexpected ways C

D

°
Mixture of endo/exo isomers/enantiomers at (A) 30, (B) 40, (C) 50, and (D) 60°C

32

RP-LC Mobile Phase Buffer Capacity
Background:
Goal: send desired compound through column un-ionized.
Buffer must be more concentrated than analytical scale because sample
is more concentrated.
Don’t over rely on aqueous pKa and Henderson–Hasselbach equation to
know ionization state – Use peak shape! (see talk on Thu)
KEY MESSAGE: must view un-ionized state loosely, as pKa shift or
anion complex with buffer can be sufficiently un-ionized for good
separation peak shape for basic drugs where pH < pKa (pKa - pH = 2-4 is OK!).
i.e. more buffer goes a long way toward reducing analyte charge.
Practical Approach:
Mix buffer on-line like “at column dilution:”
Flow rate proportional to buffer concentration.
Valve makes it easy to have 6 buffers on-line & method selectable.
Target high pH and buffer concentration for high pKa compounds:
pH 4 and 6.5 can cover a very full range of drug-like compounds.
Lower buffer concentration (0.2%) usually works well for bases when
pH > pKa. Also, more quickly removed during drying process.
Higher buffer concentration crucial for high mass loading.
Higher buffer concentration (1%) at pH 7.5 usually works well with
high loading of stronger base intermediates when pH ≤ pKa. 33

Effect on peak shape & peak capacity
Doxylamine
[formate] = 260 mM (1%)
pKa = 8.7
100 mg
injection.
Mobile phase
pH = 7.5
[formate] = 130 mM (0.5%)
Mobile phase
temperature
45°C
° N

O [formate] = 52 mM (0.2%)
Buffer needs
to reach 10x N

peak
concentration
to correct
peak shape
34

Direct effect on loading and relative to pH shift
Amitriptyline @ pH = 7.5 (pKa = 9.2) Buffer capacity
(buffer concentration)
Peak width @

0.7
base (min)

100 mg has a large impact on
0.4
50 mg column loading (peak
25 mg
width).
0.1
0.0 0.5 1.0
Buffer concentration (%)

100 mg Amitriptyline (pKa = 9.2)
1.0
The effect on loading
Peak width @
base (min)

0.7
pH 6.5 from buffer
pH 7.5
concentration can be
0.4
considerably larger
0.1 than that of the pH
0.0 0.5
Buffer concentration (%)
1.0
effect.
35

RP-LC peak shape / loading with pH: 3 distinct states
3.15
6.0e-1
AU
4.0e-1
pH = 7.5

1 st pH range with distinct chromatographic behavior:
2.0e-1

Basic compound exists as free base despite buffer
0.0 Time
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00

pH = 6.5 Small
3.12
6.0e-1

transition pH being at or below the aqueous pKa. Buffer
4.0e-1
from a big concentration does not affect chromatographic
AU

2.0e-1
change in pH. behavior (linear Langmuir behavior). Acetonitrile
0.0
0.50 1.00 1.50 2.00 2.50 3.00 3.50
Time
4.00 likely mitigates protonation. Peak shape and loading
still suggest linear Langmuir behavior.
3.07
6.0e-1 pH = 5.5
4.0e-1
AU

3 rd pH range with distinct chromatographic behavior:
2.0e-1

0.0
0.50 1.00 1.50 2.00 2.50 3.00 3.50
Time
4.00 Distinctive peak shape, loading pattern and analyte
state. Retention increases with loading (anti-
6.0e-1
pH = 5.0
Langmuir behavior). Analyte is likely to be at least
4.0e-1 2.78
AU

2.0e-1 partially protonated, but buffer anions may form
0.0 Time neutrally charged complex when present in
sufficient concentration. Rapid equilibrium seems to
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
6.0e-1
pH = 4.5
4.0e-1
2.20
result in behavior like non-ionized complex (not
AU

2.0e-1 protonated base) even at the effective pKa ≈ 4.5 in
0.0
0.50 1.00 1.50 2.00 2.50 3.00 3.50
Time
4.00
eluent.
6.0e-1
pH = 4.0 Big transition
4.0e-1
1.53
from a small 2 nd pH range with distinct chromatographic behavior:
AU

q
e
F
u
n
r
l
t
y
o
p
i
v
d
s

i
t
l
c
u
v
y
e
x
m
s
a

2.0e-1
change in pH. Basic compound primarily exists as protonated
0.0 Time base, buffer anions provide charge shielding
(nonlinear Langmuir behavior).
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00

6.0e-1
pH = 3.5 1.12
4.0e-1
AU

2.0e-1

0.0 Time
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Noscapine at 25 and 100 mg loading at
20% ACN 60% ACN different pH conditions. (Aqueous pKa = 7.8)

Another practical aspect of getting the right
buffer concentration
FEB2007_317 2: Diode Array
0.55 Un-retained desired compound 254
Range: 5.028e-1
4.0e-1

3.0e-1
AU

2.0e-1
1.53

1.0e-1 2.12
(a)
1#1,1:10 1.78
1
0.0
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
FEB2007_321 2: Diode Array
1.52 254
Range: 5.28e-1

4.0e-1

3.0e-1
AU

2.0e-1
1.22
1.0e-1 1.82
2.17
(b)
0.98
1#1,2:2
1
0.0 Time
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00

Chromatograms of a compound synthesized in-house. (a) 60 mg loading
with 48 mM of formate (b) 80 mg loading with 96 mM of formate.
37

Another practical aspect of getting the right
buffer concentration (selectivity)
1.2

1.0

8.0e-1

2.38
6.0e-1

4.0e-1

1.53 2.72 3.62
2.0e-1
0.75
(a)
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00

2.23
1.2

1.0

8.0e-1
Impurity resolved
6.0e-1

4.0e-1
1.57

2.75 3.63
2.0e-1
0.87 2.43 (b)
Time
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00

Chromatograms of a compound synthesized in-house. (a) 40 mg loading with 48 mM of formate (b) 55
mg loading with 96 mM of formate. Mobile phase pH: 6.5
38

While we don’t understand the chemistry
as well as we do RP-LC (yet), buffering
also helps SFC
1 0 0 m g , 5 % to 2 0 % M E O H + 0 .2 % D E A , 1 0 0 G /M IN , b p 1 2 0 , s p 3 0 0 ,4 0 o C
im p r a m in e - 8 2 : D io d e A r r a y
2 .9 1 320
R a n g e : 3 .7 3 6 e -1
No additive
2 .7 5 e -1

2 .5 e -1
Imipramine
2 .2 5 e -1

2 .0 e -1
0.2% DEA 100 mg injection
1 .7 5 e -1
A U

1 .5 e -1

1 .2 5 e -1

1 .0 e -1

7 .5 e -2

5 .0 e -2

2 .5 e -2

0 .0 T im e
1 .6 0 1 .8 0 2 .0 0 2 .2 0 2 .4 0 2 .6 0 2 .8 0 3 .0 0 3 .2 0 3 .4 0 3 .6 0 3 .8 0 4 .0 0

Buffering can help a lot with peak shape under high loading conditions
SFC peak shape becomes much better with adding 0.2%DEA in MeOH.

39

Other Automation Features:
Column Back-flush Regeneration
– a simple but crucial component for success
Without back-flushing, columns show increased
peak width in as few as 50 injections
Benefits of back-flushing column
Prolongs lifetime of columns; >4000 injections (> 400 g)
without loss of performance (increased peak width).
Prevents carryover and pressure gain.
Much more consistent performance.
Allows for re-equilibration of column prior to starting next cycle,
i.e. no time lost.
Back-flushing technique
Gradient back-flush repeated 3 times over duration of run
Flow rate of 20 mL/min sufficient (1/5th of prep flow)
Acidic buffer (5% acetic acid in water removes bases well) and
organic (ACN) removes lipophilic compounds.
Resolution difference between columns (in pair) may be ideal
way to evaluate condition (significant difference = dead column)
DMF is a quick way to dislodge nitrogen containing tar

Other Automation Features (LC & SFC):
ELSD collected mass estimation
• ELSD Characteristics
Mass based
detection (not
concentration)
Fairly analyte
independent
+/- 20% accuracy
readily achievable
Automated
inclusion in
FractionLynx report
41

Other Automation Features (LC & SFC):
Immediate access to data by Med Chemists
• FractionLynx (FL)
Reports captured by
NuGenesis SDMS
Our pipeline the the Med
Chemist ELN
Automated by printing FL
browser reports to SDMS.
Report includes waste UV
chromatogram to show
compound was collected (not
lost). This (and ELSD mass)
builds data driven trust with
Med Chemists
Convincing nature of data
presentation minimizes need
for post purification QC for
ordinary compounds (95%
purity threshold cmpds).
42

RP-LC Other Automation Features:
Water plumbed directly to point source
• Benefits of making
buffers on-line
Greater selection (6)
Far less labor: people
handle only small
volumes of
concentrated buffer
Software select buffer
concentration
Achieves best water
quality direct from
Millipore Gradient
Water circulates in
ceiling & loop is tapped
at point of use
43

SFC has analogous set up of “A” solvent
CO2 source photos

44

LC - Other Automation Features:
Waste collected in drums w/ automatic switching
• Waste set up:
30 gal. drum (110 liter)
Keeping drums in ventilated
cabinet achieves best safety
& aesthetics.
Simple industrial level sensor
detects full & switches to
stand-by drum.
Simple industrial level sensor
detects full & switches on
blue light to indicate need to
replace drum.
Entire waste handling
process can be maintained
by anyone.
SFC process is same but
waste package volume
reduced to 20 liters and is
easily placed under lab
bench (smaller sensor used)
45

LC - Other Automation Features:
Waste collected in drums w/ automatic switching
• Waste set up:
110 liter (30 gal) drum is best
balance between capacity &
move-ability (55 gal. drums
would work in same set up).
30 gal. capacity allows a full
(24 hr) day of operation
before drum must be
replaced (2nd drum full).
Thus, waste management
workflow is decoupled from
LC/MS workflow.
Use of DOT approved
containers & labels allowed
us to shift drum removal &
replacement to night time
cleaning staff.

46

LC & SFC drying fractions
• Med Chemists perform drying and downstream
aspects through delivery to compound
management
• Med Chemists and Analytical own the process
• We have used lean 6σ approach to streamline
our way of working
• We have found a simple, time efficient way to
remove buffers and water without extra heating
• We have 3 collection packages and workflows
depending on mass purified that rapidly move
compounds to transfer to compound
management (CM)
• Automated SD file generation for registration

47

RP-LC: Drying fractions - volatile buffer removal
• Can be washed, but that is a manual approach =
laborious
• Can use more heat and time (roto-vap, Biotage V-10, or
Genevac), but that may not be good for compound and
we want faster not slower (8-12 hr Genevac)
• Alternative: dry down to viscous goo at 35°C achieving
approximately 95% volume reduction (2-3 hr in
Genevac), then re-dissolve in pure acetonitrile
– High organic content drives off buffer and water first
• Dry again (4-5 hr in Genevac)
– Extra step for Genevac but less total time
– Result: no formate, acetate, or water (by NMR) and 1-2%
residual acetonitrile
– Easily achieved / automated with V-10 using acetonitrile as
wash solvent or by adding pure acetonitrile tube at end of batch
*volatile buffers are easily removed with a single pass dry for SFC fractions because there is no water

48

LC & SFC collection packaging: operating
with downstream process in mind
• Libraries (≤50 mg): collect
directly into pre-tared, bar-
coded 16 x 100 mm tubes
accepted by CM in
Genevac racks (Genevac
to dry = done)
• Singletons (≤2 g approx):
collect into EPA tubes w/
one tube for each 100-300
mg injection V-10 xfer &
dry to done into one or two
pre-tared, bar-coded 4 mL
tubes accepted by CM
• Multi-gram: (≥10 g) collect
into 500 mL jars
rotovap

LC & SFC drying tools: 3 needed
• Genevac for libraries
– 100% next day
• Biotage V-10 for
singletons (≤1 g)
– Like Genevac with
automatic pipetting and
serial work flow
– Acetonitrile tube added to
drive off buffer / H2O at
low temp (35°C)
– Multiple tubes transferred
to one or two 4 mL pre-
tared CM tube
– >75% same day
• Roto-vap for multi-gram
– Mostly next day
50

RP-LC purification example I:
Crude Synthesized Product

Analytical result
submitted sample

Purification performed Prep LC
using a neutral pH method purification
and C18 column
Temperatures: 40 o C

Analytical result
purified sample

51

RP-LC purification example II:
Closely Eluting Species with High Background

11:47:0623-Aug-2006
MAJG_42700-018-001_03 2487 TUV
2.88 An2
100
2.28e6
2.45

Waste stream
Purities of after fraction
>95% are
%
collector 3.84

2.27

routinely
achieved 0
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
MAJG_42700-018-001_03
Collected peak 2: Diode Array

for samples 2.88
250.9
254
Range: 7.557e-1

such as 6.0e-1
Stream prior
to fraction
these. 4.0e-1
AU

collector
2.68
2.0e-1 2.48 209.9
209.9

0.0 Time
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00

Background levels at >20% peak height (UV & MS)

52

LC purification example III:
200 mg Injection of Development Candidate

Full Scale Same Data
2% Scale
Concentrations
of fractions
10-20mg/mL.
On cooling
crystals fall
out of solution.

Typical injection for purification of gram quantities of material. In this case 8
grams were purified in 4 hours for ultra high purity (no visible impurities for
toxicology study). Level of recovery was >90%.
53

SFC example 1: Chiral purification comparison of
enantiomeric mixture by prep SFC/MS and NP-LC
(in house compound)
NP-HPLC: cycle time 23min SFC/MS: cycle time 5 min
40 mg/injection 60 mg/injection.
IA column - 2x25 cm, 5 um AD-H column - 3x15 cm, 5um
UV detection – must fish out UV & MS detection – one peak
relevant tubes desired, one tube collection
Resulting ee is 90% Resulting ee is 100%
2.00

1.80

1.60

1.40

1.20

1.00
AU

0.80

0.60

0.40

0.20

0.00

2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 20.00

Comparison: SFC is 7 fold more productive, far less laborious, and delivers
higher quality!
54

SFC example 2: Chiral resolution of flurbiprofen
(well known chiral example, comparison of isocratic with gradient)
100 m g, 5 to 30 % M E O H , 100 G /M IN , bp120 sp 280,40oC
flurbiprofen-t7 2: D iode A rray
2.71 3.36 250
R ange: 5.515
4.0
AU

2.0

0.0
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00
2.76 3.46 250
R ange: 5.512
4.0
AU

2.0

0.0
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00
1.97 2.57 250
R ange: 5.341
4.0
AU

2.0

0.0 T im e
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00

Top and middle: 100 mg and 50 mg/injection with 5 to 40% MeOH in 5
min gradient; Bottom: 50mg/injection with 15% MeOH isocratic
100g/min and 30x150 mm AD-H column, BP 120 bar
Gradient often gives better separation / loading and does not cost time!
Message: USE GRADIENT!
55

SFC example 3: Purification comparison of reaction
mixture by prep LC/MS and SFC/MS (in house compound)
Prep LC/MS: desired Prep SFC/MS fully separates
product and starting desired product (m/z = 364)
material are partially from starting material (m/z =
co-eluted 288)

m/z = 364
m/z = 364

m/z = 288
m/z = 288

We couldn’t find LC separation. SFC/MS was straight forward.
56

SFC example 4: Purification of achiral product isomers
by prep SFC/MS with chiral column (in house compound)
AD-H, 3x15cm, 30:70 IPA/CO2, 100g/min, 280 nm
A mix of isomers (meta/para 55:45) separated by SFC – 250 mg loading
Complete co-elution with RP HPLC
15:44:12 04-Mar-2010
AF28962-1c Sm (Mn, 3x4) 2: Diode Array
230
Range: 5.078e-1
Time Height Area Area%
3.24 503336 72270.71 55.27
4.5e-1 6.49 204468 58490.75 44.73

4.0e-1

3.5e-1

3.0e-1
AU

2.5e-1

2.0e-1

1.5e-1

1.0e-1

5.0e-2

0.0 Time
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00

We couldn’t find RP-LC separation. SFC/MS was straight forward.
57

RP-LC summary
Achieved analytical quality for 100+ mg injections
Good RT / threshold correlation with OA-LC/MS.
Gaussian peaks routinely.
High throughput
1-3 g/hr full gradient routine purification rate OR >20 g day (single cmpd).
High velocity separations – 4 mm/s (UPLC = 5 mm/s).
5 min run time, k’ = 20 separations, 6 min cycle time (early terminate library gradients).
Column switching eliminates need for column wash or equilibration time.
Theoretical >200 compounds purified / day (actual peak demand 400 / week).
Versatile
Range of buffers and columns selected (C8 and C18) cover a wide range of
compound purification applications.
Equipped with additional column selection to allow purification at high pH, and
if desired, HILIC or reverse phase chiral purification.
High Reliability/Success Rate
Back-flushing prolongs column life (>4000 injections per or >400
grams purified on each column).
2 full systems means “always on.” (2 work flows, no down time).
>99.99% success rates with >99% same day turnaround.
Quality = Human Efficiency = Saves You $
One expert purifies > 95% of all compounds (>150g / month).
>10 to 1 increase in human efficiency (Med Chemist time reduction).
Even-though all out quality would seem to cost ≈20% more, it
saves a lot of Med Chem effort worth many, many fold more. 58

SFC summary and overall thoughts
We are doing essentially the same thing with SFC that we do
with RP-LC (100 mL/min, 3 cm columns, most hardware same)
and we achieve essentially the same results in the same time
1-3 g/hr full gradient routine purification rate
Reliability and success rates are the same
Same day turnaround on achiral separations
However, there are some exceptions
Chiral column / solvent screening takes longer (turnaround 1-3 days chiral)
Sometimes chiral strategy requires resolution of intermediates
We don’t understand buffering as well, so we can’t yet exercise the same
degree of control on resolution using this lever
RP-LC came first, so we tend to go there first
Med Chemists just starting to learn OA-SFC/MS
It is important not to be religious about one technique
Many want to force fit into one or the other there is no such thing as a
universal technique
Instead, one should play to the strengths of multiple techniques
RP-LC is best suited for separating based on the sum of the lipophilic
parts of molecules
NP-SFC is best suited for separating based on the specific polar
functional groups and shapes of the molecules in specific regions
There is considerable overlap between RP-LC & NP-SFC in the kinds of
molecules that can be separated Use this to enhance capacity! 59

RP-LC Purification Parameters
Flow Rates: 100 mL/min total with 1- 5 mL/min buffer and 7.5 mL/min from
the dilution pump (for the first 0.3-0.5 minutes if dissimilar to eluent).
Temperatures: 45-55°C for mobile phase, columns (water bath), and back-
flushing. 45-75°C for the dilution heater.
Work horse columns: C18 Inertsil ODS-3, 30x50mm, 3 µm particles.
C8 Inertsil C8-3, 30x50mm, 3 µm particles.
Tubing ID: 0.03” prior to column, 0.02” after column.
Splitter: 1/10000 split with 1 mL/min MeOH / 0.1% formic acid as makeup
solvent.
Mobile Phase: (A) Water purified by Millipore Milli-Q Gradient system
(B) ACN UV grade from B&J (important).
Buffers: Neutral, 0.2-1.0% ammonium formate (high purity);
Acidic, 0.2-1.0 % acetic acid in ACN/water (high purity).
Dilution Solvents: Varies but predominantly 1:1 ACN/water or 100% ACN.
Back-flush Solvent: 3 gradient sweeps (A) 5% acetic 1% ACN buffer
(B) ACN (sometimes DMF plug).
Pumps / Injector / Detectors / Collectors / Software: Waters
Heaters: J-KEM Scientific, LAUDA and Selerity Technologies 60

Highlight of new application in SFC:
Open Access (OA) SFC/UV/ELSD/MS
• To gain efficiency, complementary capabilities, and greater
capacity, we have deployed OA-SFC/UV/ELSD/MS
– True orthogonal separation option for Med Chem support (TLC with
awesome detectors)
– Still has broad overlap with RP-LC/UV/ELSD/MS for Med Chem
support, thereby providing added capacity for routine reaction
monitoring
– Also opens up chiral method development and ee measurement to
“everyone”
– 3 achiral column choices & 7 for chiral (6 modifier / buffer options)
• Using new detector interfacing techniques and recent
software releases, SFC/UV/ELSD/MS is ready for prime
time in providing immediate gratification in the above
applications
61

Orthogonal SFC separations can be highly
complementary to the frequently used RP-LC
60744-024-003 1: Scan ES+ • Truly orthogonal
100
0.61
251.0 MH=251 LC/MS 251
2.67e6 SFC approach
%

can separate
0
0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 starting material
60744-024-003 1: Scan ES+
100
0.69 213 and products
213.0 MH=213 9.67e6
that RP-LC can’t
%

0
0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80
Time
2.00
• These SFC
methods also
60744-024-003, diol
60744-024-003b 1: Scan ES+ are aligned with
2.21
SFC/MS
251
6.89e7 preparative
MH=251 SFC/MS scale methods
%

allowing
0
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
immediate
60744-024-003b
0.82
1: Scan ES+
213
purification
6.13e7
• MS used in this
MH=213 application due
%

0 Time
to lack of
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 chromaphore
Normal phase separation gives TLC-
like outcome for polar intermediates
62

OA-SFC/UV/ELSD/MS can provide similar
information as OA-LC/UV/ELSD/MS
0.65
UV • Chromatograms
0.41
LC/MS showing starting
%

28
1.24
material and
0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
product (reaction
0.69
ELSD 1.66 1.76 1.85
1.93 2.00
2.03
progress)
0.39
%

0.07

0.11 0.20 0.33
0.44
0.57 0.66 • Essentially same
92
0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
Time data with either
approach except
reverse elution
1.0 order (TLC-like)
SFC/MS UV
AU

5.0e-1
• Note the
0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00 improved quality
of ELSD with
200.000 ELSD SFC!
LSU

0.000 Time
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00

63

Chiral screening of many methods on a single
sample login (MassLynx / OpenLynx SCN 798)

Method
set for
achiral
analysis
First set of methods for chiral
column screening

Second set of methods for
chiral column screening if
first set doesn’t work

New software makes method screening easy!
64

Purification process speed-and efficiency-Hayward-etal

Purification process speed-and efficiency-Hayward-etal

Recommended

Recommended

More Related Content

Viewers also liked

Viewers also liked (11)

Similar to Purification process speed-and efficiency-Hayward-etal

Similar to Purification process speed-and efficiency-Hayward-etal (20)

More from Mark Hayward

More from Mark Hayward (13)

Recently uploaded

Recently uploaded (20)

Purification process speed-and efficiency-Hayward-etal