Mitigation of analyte loss on metal surfaces in liquid chromatography

Journal of Chromatography A 1650 (2021) 462247

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Mitigation of analyte loss on metal surfaces in liquid chromatography
Martin Gilar∗, Mathew DeLano, Fabrice Gritti
Waters Corporation, 34 Maple Street, Milford, MA 01757, USA

a r t i c l e

i n f o

Article history:
Received 16 February 2021
Revised 21 April 2021
Accepted 7 May 2021
Available online 19 May 2021
Keywords:
Adsorption
Stainless steel
Titanium
Hybrid surface technology
Peak tailing
MISER

a b s t r a c t
The adsorptive loss of acidic analytes in liquid chromatography was investigated using metal frits. Repetitive injections of acidic small molecules or an oligonucleotide were made on individual 2.1 or 4.6 mm
i.d. column frits. Losses were observed for adenosine 5 -(α ,β -methylene) diphosphate, 2-pyridinol 1oxide and the 25-mer phosphorothioate oligonucleotide Trecovirsen (GEM91) on stainless steel and titanium frits. Analyte adsorption was greatest at acidic pH due to the positive charge on the metal oxide

surface. Analyte recovery increased when a series of injections was performed; this effect is known as
sample conditioning. Nearly complete recovery was achieved when the metal adsorptive sites were saturated with the analyte. A similar effect was achieved by conditioning the frits with phosphoric, citric
or etidronic acids, or their buffered solutions. These procedures can be utilized to mitigate analyte loss.
However, the effect is temporary, as the conditioning agent is gradually removed by the running mobile phase. Metal frits modified with hybrid organic/inorganic surface technology were shown to mitigate
analyte-to-metal surface interactions and improve recovery of acidic analytes. Quantitative recovery of
a 15–35 mer oligodeoxythymidine mixture was achieved using column hardware modified with hybrid
surface technology, without a need for column conditioning prior to analysis.
© 2021 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license
( />
1. Introduction
Liquid chromatography (LC) separations are based on analyte
interactions with the sorbent packed in the chromatographic column [1]. Typical sorbents are porous particles (or monoliths) with
large surface available for the adsorption-desorption interactions.
While the chromatographic sorbent is the dominant source of analyte retention, it has been reported that additional interactions
may participate in the chromatographic process [2–4]. Unexpected
peak tailing, loss of analyte recovery and shifts in retention time
were observed in analysis of peptides [5,6], oligonucleotides [4],
proteins [3], glycans [7] and selected small molecules [4,8] due to
undesirable interactions.
Several published reports discuss this phenomenon and its
causes. The observed behavior was linked to interactions of analytes with the metal surfaces present in LC systems (e.g. connecting capillaries and injector) or metallic column hardware (column
body, end fittings, and the column frits) [4,9]. Metal adsorption is
most apparent for acidic molecules, such as those containing phosphate, or multiple carboxylate moieties [4,6,10].

∗

Corresponding author.
E-mail address: (M. Gilar).

Nagayasu et al. [10] confirmed that adsorption of carboxylic
acids on stainless steel surfaces depends on pH and ionic strength
conditions. The adsorption strength increased with the number of
the charged moieties in the test molecule structure (1–6 carboxyl
groups). The authors concluded that the adsorption phenomenon
is driven by the charge of the metal oxide layer on the surface of
the stainless steel, which can be positively or negatively charged
(stainless steel pI~7). It is important to consider that pH also governs the analyte charge, which has a direct impact on the overall
adsorption [10].
This hypothesis about ionic analyte adsorption on metallic surfaces is supported by the observation that an increase in mobile phase ionic strength leads to decreased adsorption strength
of test analytes [10]. Sugiyama et al. studied ionic strength and
pH effects on the adsorption of proteins on stainless steel particles. The authors observed that multi-valent phosphate and citrate buffers compete for adsorption sites on the stainless steel
surface and effectively minimize the adsorption of acidic proteins
[11].
Loss of phosphopeptide recovery in LC MS was reported by
Fleitz et al. and others, most notably for multi-phosphorylated
peptides [6,12]. LC system treatment with EDTA or high pH mobile
phase (pH> 9) presumably reduced the phosphopeptide adsorption
on LC instrument and column metal hardware [6,13].

/>0021-9673/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( />

M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

10 x 155.5 ng

A

- PEEK union

AU

0.12

10 x 62.2 ng

0.08

10 x 31.1 ng
0.04

0.00
0.0

5.0

10.0

Minutes

15.0

90 %

AU

0.06

B

- stainless steel frit

0.04

60 %
0.02

0.00
0.0

5.0

10.0

Minutes

15.0

Fig. 1. Results of a MISER experiment consisting of 10 × 31.1 ng injections followed by 10 × 62.2 ng and 10 × 155.5 ng injections of 25-mer phosphorothioate oligonucleotide.
(A) The injector was connected to the detector via a PEEK union; (B) The PEEK union was replaced with a 4.6 mm stainless steel frit used in HPLC columns. The mobile
phase was 5 mM ammonium acetate, pH 6 and the peaks were detected by absorbance at 260 nm.

A

B

120

100

100

80

80

% recovery

% recovery

120

60
40
20
0

60
40
20
0

0

200

400

600

800

1000

1200

1400

0

200

400

ng injected

600

800

1000

1200

1400

ng injected

Fig. 2. Cumulative recovery of 25-mer phosphorothioate oligonucleotide injected in MISER experiments on a 2.1 mm i.d. stainless steel UPLC frit (A) or a 4.6 mm i.d. stainless
steel HPLC frit (B). Experimental conditions were the same as in Fig. 1. RSD values were estimated from n = 4 experiments for 2.1 mm frits and n = 6 experiments for
4.6 mm frits.
Table 1
Estimated metal surface of frits, columns, and sample accessible metal surface in selected LC systems.

Acquity H-class UPLC
Arc HPLC
a
b

Frit area (mm2 )

Column areaa (mm2 )

LC system areab (mm2 )

453 (2.1 mm)
3396 (4.6 mm)

1236 (50 × 2.1 mm)
8237 (100 × 4.6 mm)

513
1453

Calculated as internal column body surface + two frits.
Estimated from connecting tubing and injector surface exposed to the sample.

covery and reduce the carryover [3,7,12,16]. Despite the published

reports, many chromatographers underappreciate the effects of adsorption on metal surfaces on their analytical results. This is because the detrimental effects are not apparent for neutral or acidic
compound that are not charged at LC conditions. The adsorption
becomes very obvious for acidic analytes, particularly those containing two or more phosphate groups, when analyzed with low
ionic strength LC MS compatible mobile phases (e.g. 0.1% formic
acid and organic eluent) [4,8].
Several strategies were proposed to mitigate metal-surface adsorptive effects. Citrates, phosphates, ethylenediaminetetraacetic

Nucleotide mono- (NMP), di- (NDP), and triphosphates (NTP)
are classes of analytes susceptible to metal adsorption [3,8,14].
The peak tailing and sample loss increases in the following order:
NMP < NDP < NTP. The loss of analytes was conclusively linked
to adsorption on metal tubing and metal surfaces, including the
MS electrospray needle [4,15]. The adsorption of acidic analytes
on metal surfaces was also linked to an increased LC carryover
[16]. The undesirable adsorptive effects in LC prompted the development of chromatographic instruments and columns using nonmetallic materials. Polyether ether ketone (PEEK) lined tubing and
columns were shown to decrease peak tailing, improve sample re-

2


M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

AU

0.08

0.00

A

Ti HST pH 5

0.04

HST-modified titanium frits. We studied the sample loss in the mobile phase pH range between 5 and 9. We also evaluated the efficiency of LC frits passivation with selected solution of acids and
chelators.
2. Experimental

0.0

6.0

Minutes

12.0

18.0

2.1. Materials and reagents

AU

0.08

Ti pH 5

0.04

0.00

0.0

Ammonium acetate (for molecular biology, purity ≥ 98%), hexylamine (purity ≥ 99%), acetic acid (purity ≥ 99.8%), phosphoric
acid (ACS reagent, ≥ 85% w/v solution), etidronic acid (60% w/v
solution), ammonium citrate tribasic (purity ≥ 97%), ammonium
phosphate dibasic (purity ≥ 99%), citric acid anhydrous (purity ≥
99.5%), adenosine 5 -(α ,β -methylene) diphosphate (AMPcP, purity
≥ 99.0%), and 2-pyridinol 1-oxide (PNO, purity ≥ 99.0%) were purchased from Sigma-Aldrich (St. Louis, USA). Ammonium hydroxide (purity ≥ 28–30%, w/v, J.T. Baker) and acetonitrile, HPLC purity, were purchased from Thermo Fisher Scientific (Waltham, MA,
USA). HPLC purified 25-mer phosphorothioate oligonucleotide Trecovirsen (GEM91), sequence CTC TCG CAC CCA TCT CTC TCC TTC
T, was obtained from Nitto Denko Avecia, Inc (Milford, MA, USA).
A Milli-Q water purification system (Millipore, Bedford, MA, USA)
was used for preparation of HPLC mobile phases. For chemical
structures of analytes see Supplementary Fig. S1.

B
6.0

Minutes

12.0

18.0

AU

0.08

0.00

C

ss pH 5

0.04

0.0

6.0

Minutes

12.0

18.0

AU

0.08

ss pH 8

0.04

0.00

D

0.0

6.0

Minutes

12.0

2.2. LC instrumentation, columns, and frits
All chromatographic measurements were performed using an
ACQUITY UPLC H-class Bio system (Waters, Milford, MA, USA) consisting of a quaternary solvent manager (QSM), a column manager
(CM) module, a flow through needle (FTN) sample manager, and
an ACQUITY photodiode array (PDA) detector equipped with a 5 μL
titanium flow cell. The flow path of the instrument was modified
with HST to minimize the analyte interaction with metal surfaces.
Empower 3 software was used for data acquisition and analysis.
Multiple Injection in a Single Experimental Run (MISER) mode was
enabled by prototype instrument control software developed inhouse and installed on the SM-FTN. We utilized sample manager
FTN_V1.65.356, (MISER_HT_V13) firmware. The instrument control
software enables repetitive sample injections from one or multiple
vials while collecting the data in a single chromatogram.
˚ 1.7 μm,
An ACQUITY UPLC Oligonucleotide BEH C18, 130 A,
2.1 × 50 mm column (stainless steel hardware) and an ACQUITY
˚ 1.7 μm, 2.1 × 50 mm
PREMIER Oligonucleotide BEH C18, 130 A,
column were compared for the analysis of a 15–35 mer
oligodeoxythymidine mixture. The denotation PREMIER signifies
that the column hardware is modified with HST, as described in
Section 2.3. Columns and the MassPREPTM Oligonucleotide Standard were obtained from Waters (Milford, MA, USA).

18.0

Fig. 3. Results of MISER experiments consisting of 10 × 10 ng injections followed
by 10 × 20 ng and 10 × 50 ng injections of AMPcP on metal frits. (A) 4.6 mm
i.d. titanium frit modified with HST. (B) unmodified 4.6 mm i.d. titanium frit. (C,
D) 4.6 mm i.d. stainless steel frit. The mobile phase was 5 mM ammonium acetate
adjusted to pH 5 for experiments A, B, C and pH 8 for experiment D. The peaks
were detected by absorbance at 260 nm.

acid (EDTA), acetylacetone [17] or medronic acid have been used
as additives to the mobile phase (or sample diluent) to minimize
the sample adsorptive effects [3,8,9,18–20]. While effective, this
approach may have a negative impact on MS sensitivity.
Some laboratories perform column conditioning with series of
injections, until the signal response stabilizes at a desirable level
[21,22]. Another method is to passivate an entire LC-MS system
flow path, sometimes including the column, with phosphoric or
citric acid solutions wash [14,19].
Recently, methods involving permanent modification of metal
surfaces were developed to eliminate the analyte adsorption in LC.
Lauber et al. explored the use of a hybrid organic/inorganic barrier layer applied by vapor deposition. The method produces a permanent chemical barrier that minimizes the analyte contact with
metal surfaces. Vapor deposition readily modifies surfaces of column hardware, connecting tubing in LC system, and porous frits
embedded within the chromatographic columns [23]. The technology of ethylene bridged siloxane modified metal surface, named
hybrid surface technology (HST), was deployed for use in reversedphase (RP) and hydrophilic interaction chromatography (HILIC) applications [24,25].
The goal of this study is to evaluate the usefulness of HST metal
surface modification for LC analysis of acidic molecules such as
oligonucleotides and selected small molecules. The metal adsorption of analytes was investigated using stainless-steel, titanium and

2.3. Frit modification with hybrid surface technology
The hybrid surface technology, HST, forms an ethylene-bridged

siloxane polymer bonded to the metal oxide surface via a vapor
deposition process [23]. The resulting hybrid organic-inorganic barrier is chemically similar to that of bridged-ethylene hybrid (BEH)
chromatographic particles [26]. This process was used to modify
metal tubing and frit surfaces for the study. The vapor deposition
technique creates an effective barrier on high aspect ratio substrates, such as tubing with an internal diameter of 100 μm and
a length of 368 mm and porous materials such as LC frits. This
makes it possible to implement the technology for modification of
LC instruments and column hardware. The vapor deposition process was utilized for improved LC analysis of molecules with strong
3


M. Gilar, M. DeLano and F. Gritti

120

Journal of Chromatography A 1650 (2021) 462247

oligonucleoƟde, stainless steel

100

120

A
pH 5
pH 6
pH 7
pH 8
pH 9

40

60

pH 5
pH 6
pH 7
pH 8
pH 9

40

200

400

600

800

1000 1200 1400 1600

200

400

600

800

1000 1200 1400 1600

0

120

D

100

120

AMPcP, Ɵtanium

E

40

20

60

pH 5
pH 6
pH 7
pH 8
pH 9

40

20

0
800

1000

200

120

400

600

800

1000

0

G

400

40

20
0

H

800

1000

800

1000

PNO, HST Ɵtanium

I

80

60

pH 5
pH 6
pH 7
pH 8
pH 9

40

20
0

600

600

100

% recovery

pH 5
pH 6
pH 7
pH 8
pH 9

% recovery

60

400

200

120

PNO, Ɵtanium

80

200

pH 5

pH 6
pH 7
pH 8
pH 9

40

ng injected

100

80

F

60

ng injected

PNO, stainless steel

0

AMPcP, HST Ɵtanium

0
0

ng injected

100

1000 1200 1400 1600

20

0
600

800

80

% recovery

pH 5
pH 6
pH 7
pH 8
pH 9

% recovery

60

120

600

100

80

400

400

ng injected

100

80

200

200

ng injected

AMPcP, stainless steel

0

pH 5
pH 6
pH 7
pH 8
pH 9

40

0
0

ng injected

120

60

20

0
0

C

80

20

0

oligonucleoƟde, HST Ɵtanium

100

% recovery

60

20

% recovery

B

80

% recovery

% recovery

80

% recovery

120

oligonucleoƟde, Ɵtanium

100

60

pH 5
pH 6
pH 7
pH 8
pH 9

40

20
0

0

200

ng injected

400

600

800

1000

0

200

400

ng injected

600

800

1000

ng injected

Fig. 4. Cumulative recovery of 25-mer phosphorothioate oligonucleotide (A, B, C), AMPcP (D, E, F) and PNO (G, H, I) injected on 4.6 mm frits using mobile phase pH values
ranging from 5 to 9. All mobile phases were 5 mM ammonium acetate. HST modification of the frit minimizes the analyte adsorption on the metal surface (panels C, F, I).
0.06

A
0.04

AU

Ti frit,
oligonucleotide

0.02

0.00
0.0

5.0

Minutes

10.0

15.0

0.03

B

Ti frit,
AMPcP

AU

0.02

0.01

0.00
0.0

5.0

Minutes

10.0

15.0

Fig. 5. Results of MISER experiments consisting of three series of 5 injections separated by 2.5 min gaps. Totals of 15 injections of 155.5 ng of 25-mer phosphorothioate
oligonucleotide (A) or 15 injections of 50 ng of AMPcP (B) were performed on 4.6 mm i.d. titanium frits. The mobile phase was 5 mM ammonium acetate, pH 6 and the
peaks were detected by absorbance at 260 nm. The arrows highlight the decline of the signal between the last injection in a series and the initial injection after the time
gap.

4


M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

A - first load

AU

0.06

0.04

0.02

0.00

B - second load

AU

0.06

0.04

0.02

0.00

C - third load after basic wash

AU

0.06

0.04

0.02

0.00

0.0

5.0

Minutes

10.0

15.0

Fig. 6. Results of MISER experiments performed under the same conditions as Fig. 1 using a 4.6 mm i.d. stainless steel frit and the 25-mer phosphorothioate oligonucleotide
sample (A). A repeated experiment performed immediately after the conclusion of the first experiment illustrates the sample conditioning effect resulting in an improved
oligonucleotide recovery (B). The third experiment (C) was performed with the same frit after it was washed with 2.8% ammonia solution. The observed sample loss was
comparable to the result in panel A. The basic wash removes the sample conditioning (desorbs the sample from the frit surface). For additional discussion see Supplemental
Fig. S4.

affinity to metal surfaces as described by Lauber, DeLano and colleagues [22].

The MISER experiment consisted of one microliter injections in
0.5 min intervals, unless noted otherwise. The mobile phase was
5 mM ammonium acetate with pH adjusted to desirable value with
acetic acid or ammonium hydroxide (for pH see figure captions).
Mobile phase flow rate was 0.2 mL/min and the experimental temperature was 25 °C. The typical experiment included ten injections
of 2 pmole (31.1 ng) of 25-mer GEM91 oligonucleotide sample, followed by ten injections of 4 pmole (62.2 ng) and ten injections
of 10 pmol (155.5 ng) of oligonucleotide. Due to the absence of
a chromatographic column, each sample injection is detected as a
single peak. Fig. 1A illustrates the resulting MISER “chromatogram”
of thirty peaks spaced at an 0.5 min intervals performed with a
PEEK union. The experiments with AMPcP or PNO analytes were
performed in a similar fashion using 10 × 10 ng injections, followed by 10 × 20 ng and 10 × 50 ng sample injections. For all
MISER experiments the sample solvent was consistent with the
running mobile phase. Sample Manager Wash and Sample Manager Purge lines were primed with the solvent matching the running mobile phase (60 s prime for Wash solvent line, 50 cycles
prime for Purge solvent line) to ensure that the frits were exposed only to the solvents consistent with the running mobile
phase.
Fig. 1B shows that when a stainless steel frit was placed in the
sample flow-path, a loss of analyte was detected. The amount of
analyte loss was estimated by comparing the data to the control

2.4. MISER experimental procedures
MISER experimental setup was utilized with the goal to speed
up the multiple injection of the sample. The samples were injected
typically every 0.5 min by an FTN sample manager onto a single
chromatographic frit placed in a custom-made holder. The frits are
identical to those used in chromatographic columns. The sample
signal was recorded using a PDA detector connected to a holder
with 30 cm × 75 μm PEEK capillary. When evaluating analyte adsorption on the frit metal surface, stainless steel, titanium or HST
surface modified titanium frits (2.1 or 4.6 mm i.d.) were placed in

the sample flow path. A control experiment was performed by replacing the frit with the PEEK union. The control experiment (no
frit in the flow path) provided the 100% recovery value used for
relative quantitation.
The flow path of an ACQUITY H–Class Bio system was modified using HST to minimize the metal surface area exposed to the
sample. Therefore, the dominant metal surfaces available for analyte adsorption were the investigated frits. The PDA detector flow
cell and its internal tubing were built from unmodified titanium.
The amount of area available for sample adsorption in the PDA
flow cell was two orders of magnitude smaller compared to the
surface area of a 4.6 mm metal frit.
5


M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

experiment in which a PEEK union was used in place of the frit
(Fig. 1A).
Additional MISER experiments were performed with 155.5 ng
(oligonucleotide) or 50 ng (small molecules) injections using a series of five injections separated by the time gap. The gap was inserted to evaluate the amount of sample desorption from the frit
metal surface during the mobile phase wash. Another experiment
consisted of multiple series of ten injections separated by a 10 min
gap. The design of the MISER experiments can be best understood
from the presented chromatograms.

3.2. Effects of pH on analyte adsorption on metal surfaces
Fig. 3 illustrates the results of similar experiments as described
in the previous section performed with various types of 4.6 mm
i.d. frits and a range of mobile phase pH. It was reported that the
adsorption of acidic molecules is stronger at acidic pH [4,12,15,22].

Therefore, we used 5 mM ammonium acetate buffer adjusted to
pH 5 with acetic acid for the following experiment. Three separate frits were evaluated side-by-side: a titanium frit modified with
HSB technology (Fig. 3A), an unmodified titanium frit (Fig. 3B), and
a stainless steel frit (Fig. 3C).
AMPcP analyte, structurally similar to adenosine diphosphate,
but hydrolytically stable in acidic solutions, was selected for the
MISER experiment. Ten injections of 10 ng AMPcP were followed
with ten injections of 20 ng and ten injections of 50 ng of analyte. Fig. 3A shows that the hybrid surface technology modified
titanium frit does not show any adsorptive losses of the AMPcP
(compared with a PEEK union experiment, not shown). An unmodified titanium frit exhibited distinct adsorption of the analyte at pH
5 and some peak tailing can be observed (Fig. 3B). The stainless
steel frit had the most pronounced adsorption and extensive peak
tailing (Fig. 3C). When the experiment was repeated with a new
stainless steel frit and mobile phase adjusted to pH 8, reduced adsorption and tailing were detected (Fig. 3D). The observed behavior
confirms that analyte adsorption on metal surfaces is detectable for
acidic small molecules. The degree of adsorption depends on the
metal nature and mobile phase pH.
Because the analyte adsorption is related to the surface charge
on the metal [15], and also the charge of the analyte [10], we
extended the study to a wide range of mobile phase pH and investigated the adsorption of three analytes. Fig. 4 summarizes the
recovery of oligonucleotide, AMPcP and 2-pyridinol 1-oxide, PNO
for 5 mM ammonium acetate mobile phases adjusted to pH 5–9.
Oligonucleotide and AMPcP are strongly acidic molecules, expected
to be charged in the pH range 5–9. PNO is a weak acid with pKa
~ 6.2; it carries an average charge of −1 at pH >8, −0.87 at pH 7,
−0.4 at pH 6 and −0.06 at pH 5 (values calculated with ChemAxon
software). Fig. 4 illustrates several trends: The adsorption losses
for the oligonucleotide and AMPcP are most significant at acidic
pH, when the titanium and stainless steel surfaces are predominantly positively charged. The recovery of all samples improves at
basic pH, presumably due to the alteration of the surface charge

on the metal oxide layer from positive to negative. This observation is consistent with published reports and our own experience
with phosphopeptide analysis using basic mobile phases [18]. Another conclusion from the results shown in Fig. 4 is that adsorption of the oligonucleotide and AMPcP is most significant for the
stainless steel frit, followed by titanium, and HST modified titanium frits. Because the HST modification hinders the analyte access to the metal surface, this frit shows limited sample adsorption. The minor losses observed at pH 5 and 6 are likely due to
sample adsorption on parts of the LC system that were not modified with HSB (detector cell tubing); they are comparable with the
magnitude of integration error (~ 1%).
Some “dips” in the trends in Fig. 4 (most notably in Fig. 4H)
were caused by significant peak tailing that precluded an accurate
peak integration, especially when the sample mass load changed
such as in the 11th and 21st injections. The peak tailing is related
to desorption of analyte trapped on the frit surface; we will discuss
this phenomenon in the next section. The peak tailing due to metal
surfaces is illustrated in Supplemental Fig. S5.
PNO shows somewhat different behavior than the oligonucleotide and AMPcP. The lowest recoveries of PNO are observed at
pH 6 for the titanium and stainless steel frit. This is due to PNO
weaker acidity; the PNO is mostly neutral at pH 5, which leads to
improved recovery (Fig. 4G, H) and peak shape (data not shown).

3. Results and discussion
3.1. Evaluation of oligonucleotide adsorption on stainless steel using
MISER
Table 1 illustrates that the surface area of the selected HPLC
or UPLC systems accessible to the sample is relatively small compared to the surface area of the chromatographic column. Due to
the porous nature of metal frits in chromatographic columns, they
are the dominant source of the adsorption. Therefore, we utilized
2.1 and 4.6 mm LC column frits as surrogates to study the analyte
adsorption on metal surfaces.
Fig. 1 outlines the performed MISER experiment. Ten injections
of 25 mer GEM91 oligonucleotide sample (31.1 ng) were followed
with ten injections of 62.2 ng and finally ten injections of 155.5 ng
of sample. Fig. 1A shows the results of this experiment, where

the injector was connected to the detector with PEEK tubing and
union. Little or no sample loss was observed in this experiment;
the peak areas in Fig. 1A were used for estimation of oligonucleotide recovery in the subsequent experiments with the frits.
Fig. 1B shows the scenario where the PEEK union was substituted
with a 4.6 mm i.d. stainless steel frit. An apparent loss of oligonucleotide peak area was observed in the initial injections followed
by a gradual signal buildup in later injections (Fig. 1B). Approximately 60% oligonucleotide recovery was observed in the twentieth injection, while the recovery increased to 90% of the expected
peak area for the thirtieth injection. The experimental results support several conclusions. (i) The stainless steel surface adsorbs the
oligonucleotide (acidic) sample. (ii) When a sufficient mass of the
sample is injected, the frit adsorption sites are partially or completely saturated, and the analyte recovery improves. These observations are consistent with the experience of chromatographers
analyzing samples susceptible to metal adsorption. The so-called
“sample conditioning” protocol is often utilized prior to analysis
to mitigate sample losses, consisting of repetitive injections of the
sample of interest or a sacrificial compound of a similar nature.
We repeated the experiment shown in Fig. 1 with a 2.1 mm
i.d. stainless steel frit. Because the 2.1 mm frit surface is 7.5 fold
smaller compared to the 4.6 mm frit, proportionally lesser adsorption loss was observed Fig. 1B). Nearly 100% signal was observed
after a cumulative load of 500 ng of the oligonucleotide (Fig. 2A).
In contrast, the 4.6 mm i.d. frit required a mass load of 1300 ng
for the signal to reach approximately 80% recovery (Fig. 2B).
In the next experiment we investigated whether the sample adsorption on stainless steel frits is repeatable for multiple frits. The
experiment was repeated with six 4.6 mm and four 2.1 mm i.d.
new frits that were not previously exposed to the sample. The error bars in Fig. 2 illustrate that some variability exists between the
frits, presumably due to surface area and surface oxide layer variation. Raw chromatograms obtained in this experiment are provided
as Supplemental material, Fig. S2 and S3.
The presented chromatograms confirm that the MISER experimental setup is useful for rapid evaluation of sample losses on
metal surfaces. In the next section we utilize the method to study
a wider set of analytes, LC conditions, and types of metal surfaces.
6

M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

3.3. Impact of analyte adsorption-desorption on peak tailing and LC
analysis

3.4. LC system conditioning using acid wash and chelators
Alternative approaches for mitigation of sample adsorption on
metal surfaces involve an addition of chelating compounds (citric,
medronic or ethylenediaminetetraacetic acid) to the mobile phase
or sample [9,18,19,30]. Another approach is to passivate the LC system with phosphoric acid solutions [19].
We employed the MISER method to investigate the conditioning
(passivation) effectiveness for stainless steel frits. Unused 4.6 mm
i.d. frits were sonicated for 15 min at 50 °C in an aqueous solution
of the selected acids, buffers, or chelating agents. After the treatment the frits were placed into the holder, equilibrated with mobile phase, and tested for adsorption using the protocol described
in Fig. 1.
Fig. 7 visually represents the oligonucleotide recovery for various frit conditioning treatments. The conditioning in phosphoric,
citric or etidronic acid solutions was successful; little or no loss of
oligonucleotide due to metal adsorption observed. The treatments
with formic acid and EDTA did not condition the frits successfully,
while the nitric acid treatment enhanced the adsorptive loss of the
sample; no oligonucleotide signal was observed for thirty injections.
Phosphate, citrate and etidronate salts solutions were used with
the pH between 5.7 and 7. Fig. 7 illustrates that the results are
comparable with the acid solutions. This result leads us to conclude that the conditioning mechanism is the adsorption of multivalent phosphate, citrate, or etidronate ions on the metal surface,
rather than acid treatment of the frit. Nitrate or formate ions have
presumably weaker adsorption to the metal and do not act as efficient conditioning agents.
Birdsall and colleagues observed improved peak shapes for
acidic peptides after LC system treatment with phosphoric acid

[19]. However, the treatment durability was limited; the tailing
gradually worsened over several hours of run time. This observation supports our hypothesis that competitive adsorption of multivalent ions (phosphate, citrate, etidronate) reduces the interaction of acidic analytes with metallic components of the LC system/column. The conditioning with selected agents (Fig. 7) is helpful, but not a permanent solution. In a separate experiment we
investigated the stability of conditioned frits and found that deconditioning is accelerated at elevated mobile phase pH (wash
with 2.8% ammonia solution, Supplemental Fig. S6).
Fig. 8 illustrates the temporary nature of stainless steel surface
conditioning with ammonium etidronate. A new 4.6 mm i.d. stainless steel frit was treated with 10 × 10 μL injections of 0.42 M
ammonium etidronate, pH 7. The conditioning by repetitive injections of conditioning solution is an alternative to off-line frit sonication in ammonium etidronate buffer; this process is equivalent
to the sample conditioning method described in Section 3.1. After
the conditioning we immediately proceeded with the MISER experiment presented in Fig. 8. Ten series of ten injections of AMPcP were performed spaced by a 10-minute gap after each series.
Initially nearly quantitative recovery was observed (98%). However,
the signal decreased during the experiment and the peak tailing
became more pronounced. Multiple processes are contributing to
the observed loss of conditioning: (i) injected AMPcP competitively
displaces a portion of the etidronate ions from the frit, (ii) AMPcP itself is partially desorbed during the 10-minute gaps between
the injection series, (iii) running mobile phase buffer is competing
with etidronate ions adsorbed on the metal surface and the conditioning agent is gradually displaced.

Fig. 5 illustrates the results of a MISER experiment where three
series of five sample injections were separated by gaps of 2.5 min.
The design of the experiment was adjusted in order to investigate
the analyte desorption during the time gap when the frit is continuously washed with the running mobile phase. The mechanism of
tailing in chromatography can be explained by simultaneous participation of weak and strong interactions [27,28], or mass overload [29]. In the case of frits the peak shape is a result of flow
through (no interaction) and relatively strong analyte interaction
with charged metal surface.
Fig. 5 shows the initial sample recovery loss, and a gradual signal build up after titanium surface conditioning with the analyte.
The oligonucleotide peaks (Fig. 5A) show only minor tailing and
small losses of recovery (peak height) were observed after the time
gaps indicated by the arrows in Fig. 5A. This result suggests that
the oligonucleotide is strongly adsorbed to the metal surface and
the desorption kinetics is slow. Only a small portion of the oligonucleotide is desorbed from the titanium surface during the 2.5 min

wash with the mobile phase. In contrast, the experiment with AMPcP reveals severe peak tailing and an apparent loss of signal after the time gaps (see arrows in Fig. 5B). These results suggest
that AMPcP is rapidly desorbed from the titanium surface at pH 6,
which means that the effect of sample conditioning is only temporary. Fig. 5 illustrates that sample conditioning is a viable strategy
for LC analysis of the GEM91 oligonucleotide, but not for AMPcP.
The peak tailing seen for AMPcP in Fig. 5B can explain the carryover phenomenon linked to metal adsorption [16]. After the sample is injected onto the column (column frits), the adsorbed analyte slowly desorbs from the metal surface as the mobile phase
flows through the column. This leads to enrichment of the desorbed analyte on the chromatographic sorbent and contributes to
the analyte signal in the subsequent injections.
To investigate the stability of sample conditioning, we performed additional experiments illustrated in Fig. 6. First, a new
4.6 mm i.d. stainless steel frit was conditioned with the oligonucleotide sample using the experimental conditions described in
Fig. 1. After frit conditioning (Fig. 6A), the frit was washed
for 2 min with running mobile phase and the experiment was
restarted. The result for this repeated experiment with an additional thirty injections is shown in Fig. 6B. Improved oligonucleotide signal was observed in Fig. 6B, which is consistent with
successful sample conditioning of the frit in the previous run. After the second experiment the frit was disconnected from the LC
system and 0.2 mL of 2.8% ammonia aqueous solution was pushed
through the frit using a syringe, followed with 0.6 mL of 5 mM
ammonium acetate, pH 6. The frit in the holder was reconnected to
the LC system, equilibrated for 2 min with the mobile phase, and
an additional thirty injections of sample were performed (Fig 6C).
The observed loss of signal was comparable to Fig. 6A confirming
that the frit conditioning was effectively removed with the ammonia wash. This finding is consistent with the results in Fig. 4; the
basic mobile phase reduces the acidic analyte adsorption on the
metal surface by altering the metal oxide layer charge.
In a separate experiment we pre-loaded a stainless steel frit
with the oligonucleotide sample and subsequently executed five
10 μl injections of 2.8% ammonia solution on the frit. Recorded
UV data confirmed that the injections of the high pH solution effectively desorbed the oligonucleotide from the metal frit (Supplemental Fig. S4).

7

M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

0.17 M (NH4)2HPO4
pH 5.7

0.17 M H3PO4

0.52 M (NH4)3 citrate
pH 7

0.52 M citric acid

0.42 M NH4 etidronate
pH 7

0.42 M etidronic acid

2.65 M formic acid

1 mM EDTA

2.24 M nitric acid
0.0

5.0

Minutes

10.0

15.0

0.0

5.0

Minutes

10.0

15.0

Fig. 7. Results of MISER experiments executed using the same conditions as Fig. 1. Prior to the experiment each individual 4.6 mm i.d. stainless steel frit was conditioned
for 15 min at 50 °C in acid or salt solution. Blue chromatograms represent the experiments with successful conditioning with nearly complete oligonucleotide recovery. Black
chromatograms show loss of oligonucleotide recovery, which indicates an unsuccessful conditioning with formic, nitric, or EDTA acid solutions.

98 %
89 %

0.03

61 %
AU

0.02

A

0.01
0.00

0.0

30.0

60.0

90.0

120.0

150.0

Minutes
Fig. 8. Results of a MISER experiment consisting of ten series of ten injections of 50 ng of AMPcP; the series of injections were separated by ten minute gaps inserted to
allow for an extended mobile phase wash of the frit. Prior to the experiment the 4.6 mm i.d. stainless steel frit was conditioned with ten injections of 10 μL of 0.42 M
ammonium etidronate, pH 7. The decline in the AMPcP signal is suggestive of a gradual loss of etidronate.

3.5. Application of HST modified columns to the analysis of
oligonucleotides

10 pmol of each oligonucleotide was injected on column). Adsorption of oligonucleotides on conventional stainless steel column hardware resulted in low peak areas, most notably in the
first injection. Minor peaks present in the sample eluting between the dominant oligonucleotides are impurities created in
the oligonucleotide synthesis. The minor impurities were not observed in the first injection and only partially in the second injection on the conventional column. Ten injections of the oligonucleotide standard (10 pmol/injection for each dominant oligonucleotide) were required to completely condition the standard column (data not shown). Alternatively, one or two injections of
500 pmol 35 mer oligodeoxythymidine provided sufficient column
conditioning. However, no conditioning is required when we used
column hardware modified with HST. 97–100% recovery was observed in the first 10 pmol injection of 15, 20, 25, 30 and 35 mer
oligodeoxythymidine mixture (Fig 9B; for recovery see Fig. S8). The

same batch of chromatographic sorbent was used in both Fig. 9 ex-

In the previous sections we investigated several strategies for
mitigation of analyte loss due to metal surface adsorption in LC.
While helpful, neither sample conditioning nor washes of LC hardware with conditioning agents provide a permanent solution. Some
LC manufacturers introduced columns with non-metallic hardware,
most commonly constructed from polyether ether ketone (PEEK)
or PEEK-lined stainless steel. Due to limitations in mechanical
strength and chemical compatibility of PEEK, other solutions were
explored. Columns with metal hardware modified with hybrid surface technology were introduced for the analysis of metal-sensitive
analytes, including oligonucleotides, acidic peptides and certain
small molecules [22,25,31].
Fig. 9 shows the effect of column hardware on the recovery of
oligonucleotides (15, 20, 25, 30 and 35 mer oligodeoxythymidines,
8


M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

Fig. 9. Analyses of 15–35 mer oligodeoxythymidine sample using 50 × 2.1 mm, 1.7 μm Oligonucleotide BEH C18 columns with conventional stainless steel hardware (A)
and HST modified column hardware (B). Mobile phase A was 25 mM hexylammonium acetate, pH 6 and mobile phase B was prepared by mixing mobile phase A and
acetonitrile in a 1:1 ratio (v:v). A linear gradient was carried out from 50 to 86% B in 12 min, with a flow rate of 0.4 mL/min, and a column temperature of 60 °C. 10 pmol
of each oligonucleotide was injected on column. The peaks were detected by absorbance at 260 nm. For both columns three consecutive injections were made (only first
two injections are showed), followed by injection of 500 pmol of 35-mer oligodeoxythymidine and the forth injection of 10 pmol 15–35 mer oligodeoxythymidine sample
(labelled sample conditioned).

(v) Modification of metal column hardware with hybrid surface technology provides nearly quantitative recovery for
oligonucleotides and metal-sensitive small molecules. Expected signal was observed from the first injection with

no column conditioning required prior to analysis. Hybrid
organic/inorganic surface technology holds promise for LC
analysis of analytes susceptible to adsorption on metal surfaces.

periments. The improved recovery obtained with the LC column
with HST modified hardware is consistent with the MISER experiment using an HST modified titanium frit (Supplemental Fig. S7).
No loss of oligonucleotide was detected using a 4.6 mm i.d. frit
modified with HST. Due to the covalent nature of the surface modification produced using HST, the metal surface protection is stable
in most LC conditions. Accelerated stability studies of HST modified
frits under acidic and basic conditions were described in a recent
report [22].

Declaration of Competing Interest
4. Conclusion

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

In this study we were able to directly observe and quantify the
loss of analytes injected on metal frits. We suggest several strategies for mitigation of acidic analytes loss in liquid chromatography.

CRediT authorship contribution statement
(i) Loss of acidic analytes observed in LC is predominantly due
to ionic adsorption on positively charged metal oxide surfaces.
(ii) The extent of the loss depends on the nature of the metal
and mobile phase pH. The observed analyte losses were
greater on stainless steel than on titanium frits. Titanium
frits modified with hybrid surface technology exhibited negligible sample losses.
(iii) Repetitive exposure of the metal hardware to acidic samples or conditioning with the solutions of multivalent acids
(phosphate, citrate, etc.) reduced the analyte adsorptive

losses.
(iv) Although the sample conditioning and multivalent acid
washes improve the recovery and peak shape of acidic analytes in LC, the effect is temporary. LC hardware exposure
to mobile phase, and high pH buffers desorb the conditioning molecules from the metal surface.

Martin Gilar: Conceptualization, Methodology, Investigation,
Writing – original draft, Visualization. Mathew DeLano: Methodology, Investigation. Fabrice Gritti: Conceptualization, Writing – review & editing.
Acknowledgement
The authors wish to thank Thomas McDonald and Patrick Brophy for their help with setting up the LC system for MISER experiment.
Supplementary materials
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.chroma.2021.462247.
9


M. Gilar, M. DeLano and F. Gritti

Journal of Chromatography A 1650 (2021) 462247

References

[16] H. Sakamaki, T. Uchida, L.W. Lim, T. Takeuchi, Evaluation of column carryover of phosphorylated peptides and fumonisins by duplicated solvent gradient method in liquid chromatography/tandem mass spectrometry, Anal. Sci. 31
(2015) 91–97.
[17] D. Siegel, H. Permentier, R. Bischoff, Controlling detrimental effects of metal
cations in the quantification of energy metabolites via ultrahigh pressure-liquid chromatography-electrospray-tandem mass spectrometry by employing
acetylacetone as a volatile eluent modifier, J. Chromatogr. A 1294 (2013) 87–97.
[18] Y.Q. Yu, I. Campuzano, M. Gilar, Enhanced nanoLC/MS analysis of phosphopeptides, Waters Application Note (2006) 720001458en.
[19] R.E. Birdsall, J. Kellett, Y.Q. Yu, W. Chen, Application of mobile phase additives to reduce metal-ion mediated adsorption of non-phosphorylated peptides
in RPLC/MS-based assays, J. Chromatogr. B Analyt. Technol. Biomed. Life. Sci.
1126-1127 (2019) 121773.

[20] J.J. Hsiao, O.G. Potter, T.W. Chu, H. Yin, Improved LC/MS Methods for the Analysis of Metal-Sensitive Analytes Using Medronic Acid as a Mobile Phase Additive, Anal. Chem. 90 (2018) 9457–9464.
[21] M. Gilar, Z. Li, J. MacLean, Improved Chromatographic Analysis of Oligonucleotides with ACQUITY PREMIER Oligonucleotide BEH C18 Columns, Waters
Application Note 720 0 07012en 2020.
[22] M.A. Lauber, T.H. Walter, M. Gilar, D. DeLano, C. Boissel, K. Smith, R.E. Birdsall,
P. Rainville, J. Belanger, K. Wyndham, Low Adsorption HPLC Columns Based
on MaxPeak High Performance Surfaces, White paper, Waters 720 0 06930EN
(2021).
[23] A.C. Jones, M.L. Hithchman, Chemical Vapour Deposition. Precursors, Processes
and Applications, RSC Publishing, 2008.
[24] M.A. Lauber, S.M. Koza, S.A. McCall, B.A. Alden, P.C. Iraneta, K.J. Fountain, High-resolution peptide mapping separations with MS-friendly mobile phases and
charge-surface-modified C18, Anal. Chem. 85 (2013) 6936–6944.
[25] M. DeLano, T.H. Walter, M.A. Lauber, M. Gilar, M.C. Jung, J.M. Nguyen, C. Boissel, A.V. Patel, A. Bates-Harrison, K.D. Wyndham, Using hybrid organic-inorganic surface technology to mitigate analyte interactions with metal surfaces
in UHPLC, Anal. Chem. 93 (2021) 5773–5781.
[26] K.D. Wyndham, J.E. O’Gara, T.H. Walter, K.H. Glose, N.L. Lawrence, B.A. Alden,
G.S. Izzo, C.J. Hudalla, P.C. Iraneta, Characterization and evaluation of C18 HPLC
stationary phases based on ethyl-bridged hybrid organic/inorganic particles,
Anal. Chem. 75 (2003) 6781–6788.
[27] T. Fornstedt, G.M. Zhong, G. Guiochon, Peak tailing and mass transfer kinetics
in linear chromatography, J. Chromatogr. A 741 (1996) 1–12.
[28] G. Gotmar, T. Fornstedt, G. Guiochon, Peak tailing and mass transfer kinetics
in linear chromatography dependence on the column length and the linear
velocity of the mobile phase, J. Chromatogr. A 831 (1999) 17–35.
[29] J. Bagge, M. Enmark, M. Les´ ko, F. Limé, T. Fornstedt, J. Samuelsson, Impact
of stationary-phase pore size on chromatographic performance using oligonucleotide separation as a model, J. Chromatogr. A 1634 (2020) 461653.
[30] T. Nakamura, K.T. Myint, Y. Oda, Ethylenediaminetetraacetic acid increases
identification rate of phosphoproteomics in real biological samples, J. Proteome
Res. 9 (2010) 1385–1391.
[31] M.A. Lauber, M.H. Delano, S.A. McCall, J.L. Belanger, T.A. Dourdeville, K.M.
Smith, P.D. Rainville, D.D. Shah, S.J. Shiner, C. Doneanu, M. Donegan, Waters
Technologies Corporation, US 20190086371, 2018.

[1] U.D. Neue, HPLC columns: theory, technology, and Practice, Wiley-VCH, New
York, 1997.
[2] J.J. Hsiao, T.W. Chu, O.G. Potter, G.O. Staples, D.R. Stoll, Troubleshooting LC separations of biomolecules, part ii: passivation and mobile-phase additives, LC
GC N. Am. 38 (2020) 431–434.
[3] J.J. Hsiao, G.O. Staples, D.R. Stoll, Troubleshooting LC separations of
biomolecules, part i: background, and the meaning of inertness, LC GC N. Am.
38 (2020) 146–150.
[4] R. Tuytten, F. Lemière, E. Witters, W. Van Dongen, H. Slegers, R.P. Newton, H. Van Onckelen, E.L. Esmans, Stainless steel electrospray probe: a dead
end for phosphorylated organic compounds? J. Chromatogr. A 1104 (2006)
209–221.
[5] M. Gilar, Y.Q. Yu, J. Ahn, J. Fournier, J.C. Gebler, Mixed-mode chromatography
for fractionation of peptides, phosphopeptides, and sialylated glycopeptides, J.
Chromatogr. A 1191 (2008) 162–170.
[6] A. Fleitz, E. Nieves, C. Madrid-Aliste, S.J. Fentress, L.D. Sibley, L.M. Weiss,
R.H. Angeletti, F.Y. Che, Enhanced detection of multiply phosphorylated peptides and identification of their sites of modification, Anal. Chem. 85 (2013)
8566–8576.
[7] K. Sandra, J. Vandenbussche, P. Sandra, Analyzing phosphorylated N-glycans
with full recovery on bio-inert LC systems and PEEK-lined HILIC columns, LC
GC Eur. 31 (2018) 566–571.
[8] J.C. Heaton, D.V. McCalley, Some factors that can lead to poor peak shape in
hydrophilic interaction chromatography, and possibilities for their remediation,
J. Chromatogr. A 1427 (2016) 37–44.
[9] D. Winter, J. Seidler, Y. Ziv, Y. Shiloh, W.D. Lehmann, Citrate boosts the performance of phosphopeptide analysis by UPLC-ESI-MS/MS, J. Proteome Res. 8
(2009) 418–424.
[10] T. Nagayasu, C. Yoshioka, K. Imamura, K. Nakanishi, Effects of carboxyl groups
on the adsorption behavior of low-molecular-weight substances on a stainless
steel surface, J. Colloid Interface Sci. 279 (2004) 296–306.
[11] H. Sugiyama, T. Hagiwara, H. Watanabe, T. Sakiyama, Effects of ionic substances
on the adsorption of egg white proteins to a stainless steel surface, Biosci.

Biotechnol. Biochem. 76 (2012) 467–472.
[12] H. Sakamaki, T. Uchida, L.W. Lim, T. Takeuchi, Evaluation of column hardware
on liquid chromatography-mass spectrometry of phosphorylated compounds,
J. Chromatogr. A 1381 (2015) 125–131.
[13] C. Song, M. Ye, G. Han, X. Jiang, F. Wang, Z. Yu, R. Chen, H. Zou, Reversed-phase-reversed-phase liquid chromatography approach with high orthogonality for multidimensional separation of phosphopeptides, Anal. Chem.
82 (2010) 53–56.
[14] A. Wakamatsu, K. Morimoto, M. Shimizu, S. Kudoh, A severe peak tailing of
phosphate compounds caused by interaction with stainless steel used for liquid chromatography and electrospray mass spectrometry, J. Sep. Sci. 28 (2005)
1823–1830.
[15] M.A. Iqubal, R. Sharma, Kamaluddin, Surface interaction of ribonucleic acid
constituents with spinel ferrite nanoparticles: a prebiotic chemistry experiment, RSC Adv. 6 (2016) 68574–68583.

10