The impact of low adsorption surfaces for the analysis of DNA and RNA oligonucleotides

The goal of the present paper was to comprehensively evaluate various types of bioinert materials used in ion-pairing reversed-phase (IPRPLC) and hydrophilic interaction chromatography (HILIC) to mitigate this issue for 15- to 100-mer DNA and RNA oligonucleotides.

Journal of Chromatography A 1677 (2022) 463324

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/chroma

Honorine Lardeuxa, b, Alexandre Goyonc, Kelly Zhangc, Jennifer M Nguyend,

Matthew A Lauberd, Davy Guillarmea, b, Valentina D’Atria, b, ∗

a Institute of Pharmaceutical Sciences of Western Switzerland (ISPSO), University of Geneva, CMU-Rue Michel Servet 1, Geneva 4 1211, Switzerland

b School of Pharmaceutical Sciences, University of Geneva, CMU-Rue Michel Servet 1, Geneva 4 1211, Switzerland

c Small Molecule Pharmaceutical Sciences, Genentech Inc., DNA Way, South San Francisco, CA 94080, USA

d Waters Corporation, 34 Maple Street, Milford, MA 01757, USA

a r t i c l e i n f o

Article history:

Received 18 March 2022

Revised 7 July 2022

Accepted 8 July 2022

Available online 9 July 2022

Keywords:

Oligonucleotides

Ion-pairing reversed-phase chromatography

(IP-RPLC)

Hydrophilic interaction chromatography

(HILIC)

Bioinert surfaces

Low adsorption surfaces

a b s t r a c t

As interest in oligonucleotide (ON) therapeutics is increasing, there is a need to develop suitable ana- lytical methods able to properly analyze those molecules However, an issue exists in the adsorption of ONs on different parts of the instrumentation during their analysis The goal of the present paper was

to comprehensively evaluate various types of bioinert materials used in ion-pairing reversed-phase (IP- RPLC) and hydrophilic interaction chromatography (HILIC) to mitigate this issue for 15- to 100-mer DNA and RNA oligonucleotides The whole sample flow path was considered under both conditions, including chromatographic columns, ultra-high-performance liquid chromatography (UHPLC) system, and ultravio- let (UV) flow cell It was found that a negligible amount of non-specific adsorption might be attributable

to the chromatographic instrumentation However, the flow cell of a detector should be carefully sub- jected to sample-based conditioning, as the material used in the UV flow cell was found to significantly impact the peak shapes of the largest ONs (60- to 100-mer) Most importantly, we found that the choice

of column hardware had the most significant impact on the extent of non-specific adsorption Depend- ing on the material used for the column walls and frits, adsorption can be more or less pronounced It was proved that any type of bioinert RPLC/HILIC column hardware offered some clear benefits in terms

of adsorption in comparison to their stainless-steel counterparts Finally, the evaluation of a large set of ONs was performed, including a DNA duplex and DNA or RNA ONs having different base composition, furanose sugar, and modifications occurring at the phosphate linkage or at the sugar moiety This work represents an important advance in understanding the overall ON adsorption, and it helps to define the best combination of materials when analyzing a wide range of unmodified and modified 20-mer DNA and RNA ONs

© 2022 The Authors Published by Elsevier B.V This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/)

1 Introduction

Therapeutic oligonucleotides have gained increasing attention

thanks to their high potential to treat a large variety of diseases

[ 1, 2] By the end of 2021, 16 oligonucleotide-drug therapies have

been approved by Food and Drug Administration (FDA) or Euro-

pean Medicines Agency (EMA), with twelve of them having re-

ceived approval since 2016 [3] Milasen, a personalized oligonu-

cleotide specifically developed for a single patient suffering from

∗ Corresponding author at: Institute of Pharmaceutical Sciences of Western

Switzerland (ISPSO), University of Geneva, CMU-Rue Michel Servet 1, Geneva 4 1211,

Switzerland

E-mail address: valentina.datri@unige.ch (V D’Atri)

Batten disease, is a promising example of oligonucleotide-based customized medicine [ 4, 5]

To support the development of these complex drugs, robust and sensitive analytical methods are required Ion-pairing reversed- phase liquid chromatography (IP-RPLC), also known as ion-pair chromatography, is recognized as the gold standard method for the characterization of oligonucleotide products and related im- purities [ 6, 7] Being complex amphiphilic molecules, ONs present hydrophilic and negatively-charged backbone Therefore, they are not sufficiently retained on hydrophobic RPLC stationary phases For this reason, ion-pairing (IP) agents such as N-alkyl amines are added to the mobile phase, forming oligonucleotide ion-pairs that may be separated based on their related hydrophobicity At ele- vated temperatures applied to these separations, oligonucleotides

https://doi.org/10.1016/j.chroma.2022.463324

0021-9673/© 2022 The Authors Published by Elsevier B.V This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )

adopt a linear form such that a length-based separation is primar-

ily observed [8–10] Even if the ion-pair formation in solution is

a commonly depicted mechanism, the chromatographic separation

may also be explained by the initial adsorption of the IP agent on

the hydrophobic stationary phase via its alkyl chains, followed by

an ion-exchange process between the charged surface and the an-

alyte [11] It is generally accepted that both mechanisms coexist to

explain the retention model of the so-called “ion-pair chromatog-

raphy” [ 12, 13]

IP-RPLC of oligonucleotides has been widely explored over the

years First achieved by Fritz et al in 1978, IP-RPLC separation

of oligonucleotides historically used triethylamine (TEA) as the IP

agent [ 8, 14, 15] To overcome some limitations in terms of ESI-MS

sensitivity, Apffel et al suggested in 1997 the addition of hexaflu-

oroisopropanol (HFIP) to facilitate the ESI process as well as the IP

efficiency [ 11, 16–21] While alternative IP agents have been widely

evaluated [22–31], TEA-HFIP mobile phase remains widely used for

oligonucleotide characterization [ 6, 7, 16, 32]

The highly polar nature of ONs makes it possible to also con-

sider hydrophilic interaction chromatography (HILIC) In a HILIC

separation, charged oligonucleotides can be separated on a polar

stationary phase, usually bonded with polar groups such as amide

or diol moieties, using a highly-organic mobile phase that contains

salts to enhance retention capabilities and selectivity The separa-

tion mechanism involves the partitioning of analytes between the

bulk mobile phase and a water-rich layer immobilized on the sta-

tionary phase surface Retention is further achieved through ionic

and hydrogen bonding interactions [33]

HILIC was first introduced by Alpert in 1990, but HILIC for ON

analysis has grown exponentially in the last few years [34] The

popularization of MS-friendly, ion-pairing free buffers such as am-

monium acetate or formate that substitute the previous use of tri-

ethylammonium acetate in HILIC mode further encourages the de-

velopment of HILIC [35–44]

It has been widely reported that oligonucleotides, because of

their electron-rich backbone, suffer from undesired, and often ad-

sorptive, interactions with materials traditionally used in chro-

matographic analyses The main construction material of chro-

matographic systems and columns is stainless-steel to ensure pres-

sure resistance Despite its mechanical strength, easy manufac-

turability and compatibility with most eluents, stainless-steel was

found to be susceptible to corrosion with many diverse chromato-

graphic eluents [ 45, 46] The resulting positively-charged metal ox-

ide layer at the surface of the metallic components may cause

problems of metal leaching, impacting the chromatographic and

MS performance, as well as leading to irreversible adsorption of

analytes [47] This non-specific adsorption is even more critical

when working at low to neutral pH, being that these are condi-

tions under which metals are more electropositive and most likely

to cause ionic interactions with negatively-charged species such

as oligonucleotides [ 42, 48–51] These unwanted ionic interactions

with the oligonucleotides are further increased as the stainless-

steel surface becomes more corroded and as the number of phos-

phate groups increases [ 51, 52] Non-specific adsorption may also

be a result of polarity-based interactions between hydrophobic

ion-pairs and hydrophobic materials from the flow path

This phenomenon negatively impacts chromatographic perfor-

mance by reducing recovery and altering peak shapes (tailing,

asymmetry) Consequently, sensitive detection as well as accurate

quantitation are hindered, and reliability and reproducibility be-

come compromised [ 53, 54] Several approaches have traditionally

been used in an attempt to minimize oligonucleotide adsorption

In one case, a strong acid or a sacrificial sample can be used to

mask active sites of metallic surfaces and thereby passivate a chro-

matographic system or column [ 55, 56] Chelators such as ethylene-

diaminetetraacetic acid (EDTA) may also be used to trap metal ions

and prevent adsorption However, their use can come with certain drawbacks, such as ion suppression and persistence in the system

In addition, these techniques are time-consuming and not long- standing [57–60]

In the last few years, chromatographic instrument manufactur- ers have focused their developments on strategies to permanently mitigate adsorption of problematic analytes Low adsorption sys- tems and columns have been offered and are based on the use

of novel surface technologies Made of bioinert and/or biocompat- ible materials, they provide a solution to suppress interactions of oligonucleotides with surfaces [ 53–55, 60–63] In general, the term

“bioinert” refers to a surface that hampers adsorption, while the term “biocompatible” is used to define a corrosion-resistant mate- rial [ 61, 64] Therefore, oligonucleotide analyses require the use of bioinert materials, which are also biocompatible

In this work, we present a comprehensive evaluation of bioinert strategies to prevent non-specific adsorption of oligonucleotides in IP-RPLC and HILIC In IP-RPLC mode, three columns made of dif- ferent bioinert hardware ( i.e titanium-lined, PEEK-lined and hy- brid organic/inorganic surface columns) were compared to their stainless-steel counterparts using model oligonucleotide samples (DNA and RNA oligonucleotides ranging from 15- to 100-mer) Sim- ilarly, bioinert HILIC columns were compared to their stainless- steel counterparts As bioinert HILIC columns are just recently emerging, only PEEK-lined and hybrid organic/inorganic surface columns are available To our knowledge, HILIC columns in a titanium-lined hardware are not yet offered Finally, the impact of instrumentation hardware was also investigated Three chromato- graphic systems with different fluidic path material ( i.e stainless-

steel, MP35N and titanium, and hybrid organic/inorganic surface) were considered and the impact of the UV flow cell was also high- lighted The last part of the study deals with the extension of our observations with the analysis of a wide range of unmodified and modified 20-mer oligonucleotides

To our knowledge, a systematic comparison of the impact of bioinert columns consisting of different column hardware has never been reported before More importantly, the evaluation of non-specific adsorption of DNA and RNA oligonucleotides in HILIC mode has been here comprehensively investigated for the first time by using bioinert column hardware

This work was essential to understand the contribution of each hardware parameter on the overall oligonucleotide adsorption and conclude on a combination of materials to preferentially use in fu- ture studies

2 Experimental

2.1 Chemicals and reagents

Oligonucleotides were purchased from Eurogentec (Seraing, Bel- gium) and Integrated DNA Technologies (IDT, Leuven, Belgium) Type 1 water was obtained from a Milli-Q purification system from Millipore (Bedford, MA, USA) LC-MS grade methanol (art M/4062/17) and acetonitrile (art A/0638/17) were purchased from Thermo Fisher Scientific (Reinach, Switzerland) Ammonium ac- etate ( ≥98%, art 32301), 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP,

≥99%, art 105228), triethylamine (TEA, ≥99.5%, art 90340), and RNase-free water (art 95289) were purchased from Sigma-Aldrich (Buchs, Switzerland)

2.2 Sample preparation

Eppendorf DNA LoBind® tubes, Eppendorf Dualfilter T.I.P.S® and polypropylene vials were systematically used during this work to eliminate any risk of additional adsorption that would bias our re- sults 100-μM oligonucleotide aliquots were initially prepared by

H Lardeux, A Goyon, K Zhang et al Journal of Chromatography A 1677 (2022) 463324

Table 1

Sequences, molecular masses and modification types of investigated oligonucleotides Phosphorothioate (PS) linkages are indicated by a ∗ , 2’- O -methoxyethyl modifications (MOE) by a X, 2’-O-methyl (OMe) modifications by a Y, and locked nucleic acids (LNA) by a Z

Compound Name Length (mer) DNA/ RNA Sequence (5’-3’) Molecular weight (g.mol −1 ) Modification

dT15-35 :

Equimolar mixture of

dT15, dT20, dT25, dT30,

dT35

7542.9

dT40-100 :

Equimolar mixture of

dT40, dT60, dT80, dT100

rU15-30 :

Equimolar mixture of

rU15, rU20, rU30

reconstituting lyophilized material in the appropriate volume of

RNase-free water and stored at – 20 °C (DNA oligonucleotides) or

– 80 °C (RNA oligonucleotides) Oligonucleotides samples were pre-

pared by diluting the oligonucleotide material to 2 μM in RNase-

free water or 10:90 H 2O/ACN prior to RPLC or HILIC analysis, re-

spectively Equimolar oligonucleotide mixtures were prepared by

mixing aliquots and diluting the oligonucleotide product to 2 μM

Table1lists investigated oligonucleotides and their characteristics

A 20-mer double-stranded DNA oligonucleotide was also stud-

ied in HILIC mode A single-stranded ON, having 5’-TTC GCC

TCG CAG TGC GCC TT-3’ sequence and molecular weight of

12 237 g.mol −1, and its complementary sequence were annealed

to form the duplex by using the following protocol 100- μM single-

stranded oligonucleotides were mixed in an annealing buffer con-

sisting of 100 mM ammonium acetate in H 2O, heated at 85 °C for

5 min, then allowed to slowly cool down to room temperature and

aliquoted The final concentration of the duplex was 50 μM Du-

plex sample was then prepared by diluting the material to 2 μM in

10:90 H 2O/ACN

2.3 Instrumentation and columns

2.3.1 H-Class

The Acquity UPLC TMH-Class system (Waters, Milford, MA, USA)

was equipped with a quaternary solvent delivery pump, an au-

tosampler including a 10-μL flow-through-needle injector and a

tunable ultraviolet (TUV) detector with a 500-nL analytical flow

cell (10 mm path length) The flow path of the instrument was

made of stainless-steel

2.3.2 H-Class Bio

The Waters Acquity UPLC TM H-Class Bio system was equipped

with a biocompatible quaternary solvent delivery pump, an au-

tosampler including a 10-μL flow-through-needle injector The TUV

detector was equipped with a 1500-nL titanium flow cell (5 mm

path length) The flow path of this instrument primarily consists

of an MP35N alloy

2.3.3 Premier system

The Waters Acquity Premier TMsystem was equipped with a bi-

nary solvent delivery pump and an autosampler including a 10-μL

flow-through-needle injector The TUV detector was equipped with

a 500-nL analytical flow cell (10 mm path length) The flow path

of this instrument contains hybrid surface technology (HST), which

is described as MaxPeak TM High Performance Surfaces by Waters

It is an ethylene bridge hybrid siloxane surface that is applied to materials by vapor deposition

For sake of consistency and proper data comparison, the same TUV detector was used on the three instruments while being equipped with different flow cells In all cases, absorbance data were acquired at 260 nm Data acquisition and instrument control were performed by Empower 3 software (Waters)

2.3.4 Columns

RPLC and HILIC columns used for this study, including bioin- ert columns and their stainless-steel analogs, have been listed in Table2 The complete history of RPLC and HILIC column injections

is reported in Table S1 and Table S2, respectively

2.4 Chromatographic conditions

Mobile phases for IP-RPLC analyses were composed of 15 mM TEA, 400 mM HFIP in water, pH 7.9 (mobile phase A) and a mix- ture of 50:50 mobile phase A and methanol (mobile phase B) The flow rate was set at 0.4 mL/min and the column tempera- ture at 70 °C A gradient of 40–50%B in 15 min was used for dT40–

100, while a gradient change of 20%B in 20 min was used for the other oligonucleotides Gradients of 30–50%B were used for dT15-

35, dT20, dA20, dG20, dC20, dT20-PS, rU20-MOE, dT20-LNA; and 20-40%B for rU15-30, rU20, rU20-PS, rU20-OMe

Mobile phases for HILIC analyses were composed of 50 mM am- monium acetate in water, pH 6.9 (mobile phase A, no adjustment

of pH) and acetonitrile (mobile phase B) The flow rate was set at 0.3 mL/min and the column temperature at 40 °C Gradient condi- tions were the same for all oligonucleotides A gradient of 55–25%B

in 30 min was used for Waters columns, while a gradient of 75– 45%B in 30 min was used for YMC columns

It should be noted that the goal of this work was to eval- uate the contribution of each hardware parameter on the over- all oligonucleotide adsorption, therefore possible effects of mobile phase compositions and pH were not investigated

Column dimensions

˚ A)

High

Olig

High

All oligonucleotides were concentrated at 2 μM and injection volume was 2 μL Gradient conditions were optimized during pre- liminary studies All gradients were systematically followed by an 8-min re-equilibration to the initial conditions

3 Results and discussions

3.1 Impact of column material

Column hardware represents more than 70% of the sample ac- cessible surfaces during an analysis [60] Generally composed of

a stainless-steel tube and stainless-steel frits, it is often assumed that the LC column will introduce the most significant source of adsorption problems Several strategies exist to minimize sample losses and distorted peaks While a “sample conditioning” proto- col is often used prior to analysis, bioinert columns have recently become available, and they are meant to be a permanent solution

to non-specific adsorption on column hardware Columns featur- ing titanium, PEEK and hybrid organic/inorganic surface technolo- gies are now commercially available as alternatives to conventional stainless-steel columns

To examine currently available techniques, we compared three different bioinert RPLC columns (one of each technology) to their stainless-steel analog The first technology was a hybrid surface technology (HST) column described as MaxPeak TM High Perfor- mance Surfaces by Waters (Wpr), which is the bioinert version of the BEH C18 column (Wss) The second one is a PEEK-lined column from YMC (Ypk), which is the bioinert version of the YMC-Triart C18 column (Yss) Last one is a titanium-lined column described

as bioZen TMby Phenomenex (Pti), which is the bioinert version of the Kinetex® EVO C18 (Pss)

Besides, bioinert HILIC alternatives are still emerging To the best of our knowledge, commercially available bioinert HILIC columns are currently limited to hybrid organic/inorganic surface and PEEK technologies Therefore, we compared the two existing bioinert HILIC column hardware types with their stainless-steel analogs The first one is a HST column from Waters (Wpr_H) that

is the bioinert version of a BEH Amide column (Wss_H) The sec- ond one is a PEEK-lined column from YMC (Ypk_H), which is the bioinert version of the YMC-Triart Diol (Yss_H)

A Premier system comprised of hybrid surface flow path com- ponents was systematically used for this column comparison study

3.1.1 Sample conditioning of columns

A sample conditioning step was incorporated into the com- parison of these columns The dT15-35 sample (mixture of poly- deoxythymidylic acids of 15-, 20-, 25-, 30- and 35-mer), which has many times been used for column and system performance test- ing, was chosen as the “conditioning” sample [ 53, 60] Consecutive injections of a solution containing 4 pmol of each oligodeoxyri- bonucleotide (ODN) were made on each brand-new column until consistent peak areas were achieved

For the IP-RPLC mode, results have been reported in Fig.1for Waters (Wss vs Wpr, Fig.1A, B), YMC (Yss vs Ypk, Fig.1C, D) and Phenomenex (Pss vs Pti, Fig.1E, F) columns For each column, the first injection was taken as the reference value (100%) and relative peak areas of each ODN were expressed as % to plot the evolu- tion of peak areas over injections Fig.1also presents the overlaid chromatograms of the first and last injections of the sample con- ditioning protocol

With stainless-steel columns, very low peak areas were ob- tained for the first injection on the columns, with early eluting peaks particularly affected, as shown in Fig.1A and C for the Wss and Yss columns This was even worse for the Pss column ( Fig.1E) which resulted in nearly complete sample loss regardless of the ODN Because of non-specific adsorption, the use of a brand-new

H Lardeux, A Goyon, K Zhang et al Journal of Chromatography A 1677 (2022) 463324

Fig 1 Monitoring of peak area increases during sample conditioning in IP-RPLC mode using the Premier system Overlaid chromatograms from the first injection (before

conditioning) and the last injection (after conditioning) of the mixture dT15-35 when using (A) stainless-steel Waters (Wss), (B) bioinert Waters (Wpr), (C) stainless-steel YMC (Yss), (D) PEEK-lined YMC (Ypk), (E) stainless-steel Phenomenex (Pss), or (F) titanium-lined Phenomenex (Pti) columns 15 injections (300 pmol) and 4 injections (80 pmol) were required for sample-based conditioning of the stainless-steel and bioinert columns, respectively 100% peak area corresponds to first injection (inj 1)

column without conditioning cannot give reliable results In the

subsequent injections, the peak areas gradually increased with a

plateau reached after 15 injections This corresponds to an ac-

tual mass load of 300 pmol, required to mask the active sites of

stainless-steel material from the columns We can also notice dif-

ferences in terms of adsorption behavior between ODNs The short-

est oligodeoxythymidines (dT15 and dT20) systematically showed

the greatest increase in peak areas over conditioning time, with

relative peak areas up to 4200% in the 15th injection This may be

explained by their elution order rather than length Early eluting

compounds will sacrificially saturate adsorption sites as they go

through the column, leading to later eluting oligonucleotides be-

ing less adsorbed to the column hardware [54] As a result, dT35

relative peak area showed a 175 to 580% value in the last injec-

tion, which is moderately high compared to dT15 or dT20 Despite

differences in adsorption behavior, these results demonstrate the

need to carefully condition stainless-steel columns

Contrary to their stainless-steel analogs, nearly full recovery for

all oligonucleotides was achieved upon the first injection on bioin-

ert RPLC columns To verify this observation, four successive in-

jections of the ODN mixture were performed, corresponding to 80

pmol loaded on column Fig 1B, D and 1F showed that relative

peak areas over the four injections varied from 98 to 104% regard-

less of the ON length and type of bioinert column Adsorption was

reduced upon the first injection, and reliable results were obtain-

able without sacrificing time or samples In addition, resolution of

minor peaks from failure sequences can be seen to be afforded

with the bioinert columns upon the first injection However, this

was not achieved with brand-new stainless-steel columns, where sample conditioning was needed to detect such minor impurities The same sample conditioning protocol was applied to both stainless-steel HILIC ( Fig 2A and C) and bioinert HILIC ( Fig 2B and D) columns Different than with IP-RPLC, each HILIC column showed different behavior and did not require the same amount

of sample to be effectively passivated As shown in Fig.2C, Yss_H columns showed the greatest number of injections required to reach a plateau in terms of peak area, with 20 injections cor- responding to 400 pmol of oligonucleotide in total In any case the utility of sample conditioning the stainless-steel columns to limit undesired interactions with the metallic column hardware was again demonstrated

However, despite the conditioning protocol, the increase in peak areas over conditioning is not as pronounced as in IP-RPLC mode (up to 440% vs up to 4200%, respectively) Surprisingly, bioin- ert HILIC columns also required some conditioning injections to achieve consistent peak areas for the dT15-35 sample ( Fig.2B and D) while it was barely necessary in IP-RPLC Indeed, 10 and 14 in- jections (200/280 pmol) were performed to reach the plateau of peak area on Wpr_H and Ypk_H columns, respectively, demonstrat- ing that this sample conditioning protocol was quite slow with a very gradual increase in peak areas

Finally, these findings highlight the benefits of using a bioinert column for the analysis of metal-sensitive analytes from a “sam- ple conditioning” point of view Sample conditioning was found

to be required with standard RPLC columns, while bioinert RPLC columns seem to show ready to use performance upon the first

Fig 2 Monitoring of peak area increases during sample conditioning in HILIC mode using the Premier system Overlaid chromatograms from the first injection (before

conditioning) and the last injection (after conditioning) of the mixture dT15-35 when using (A) stainless-steel Waters (Wss_H), (B) bioinert Waters (Wpr_H) column, (C) stainless-steel YMC (Yss_H), (D) PEEK-lined YMC (Ypk_H) columns 15 injections (300 pmol) and 10 injections (200 pmol) were required for sample-based conditioning of the Wss_H and Wpr_H columns, respectively, while 20 injections (400 pmol) and 14 injections (280 pmol) were required for sample-based conditioning of the Wss_H and Ypk_H, respectively 100% peak area corresponds to the first injection (inj 1)

injection Besides, the use of bioinert HILIC columns did not com-

pletely suppress non-specific adsorption, but required a reduced

number of conditioning injections in comparison with stainless-

steel HILIC columns

3.1.2 Analysis of unmodified oligonucleotides

In addition to the initial column conditioning of the different

columns employed in this work, the behavior of these columns

was evaluated for the analysis of three different types of model

oligonucleotide (ON) samples These included the previously ana-

lyzed dT15-35 sample, but also a mixture of larger ODNs (dT40-

100, a mixture of poly-deoxythymidylic acids of 40-, 60-, 80-

and 100-mer), and finally a mixture of small oligoribonucleotides

(ORNs, rU15-30, a mixture of poly-uridylic acids of 15-, 20- and

30-mer)

To obtain a consistent comparison and draw reliable conclu-

sions, the six RPLC columns and the four HILIC columns shared

the same history in terms of usage before the injections of the

three model ON samples were carried out (as reported in Table S1

and Table S2) In Fig.3, the relative peak areas of the different ON

products are provided, while the chromatograms related to these

data for bioinert columns are reported in Fig S1 For each individ-

ual ON and in both IP-RPLC and HILIC modes, the Waters Premier

column (Wpr and Wpr_H, respectively) was taken as the reference

value (100%) for the calculations The columns were already condi-

tioned, and a plateau was reached as described in Section3.1.1and

reported in Figs.1and 2

Concerning the results obtained in IP-RPLC mode ( Fig 3A–C),

no significant differences were observed between the stainless-

steel and bioinert (Wss and Wpr) Waters columns when analyz-

ing the dT15-35 sample ( Fig 3A) The behavior of the two other

bioinert columns (Ypk and Pti) was also in line with our expec-

tations, and values of 105–110% (not significantly different from

100%) were experimentally observed, which means that all three

bioinert type columns produced the same peak areas for the dT15-

35 mixture On the other hand, despite the column conditioning,

the two remaining stainless-steel columns (Yss and Pss) produced

a peak area equal to only 60–82%, with increasing values for the larger ODNs such as dT35 (similar behavior to what was already explained in Section 3.2.1.) This confirms that adsorption of ONs (equal to 20–40%) is still taking place on these two columns, de- spite the column conditioning procedure This could be due to dif- ferences in metal surface areas, microsite corrosion across the vari- ous materials or column manufacturing procedures, but ultimately

it seems to be an indication of the number of active sites where ONs can adsorb Batch-to-batch testing of different hardware lots was not possible here, and it could be equally possible that the behavior of stainless-steel columns in this regard is highly variable

As reported in Fig.3B, more pronounced differences were observed between the RPLC columns when analyzing the dT40–100 sample, most likely due to the increasing sizes of the ONs Some differ- ences were observed between the reference bioinert hybrid surface (Wpr) column and its stainless-steel counterpart (Wss) Indeed, relative areas varied from 90% for dT40 to only 40% for dT100, and the first eluted peak was not the one to show the lowest re- covery In addition, the PEEK column (Ypk) behaves very well for dT40 (relative peak area of 101%), but adsorption was significantly more pronounced when increasing the ON size (relative peak area

of 46% for dT100) On the contrary, the stainless-steel column from the same provider (Yss) has a relatively constant behavior indepen- dent of the ON size, with relative peak area comprised between 70 and 82% Finally, the titanium column (Pti) has the exact opposite behavior to the PEEK column (Ypk), with a significant reduction

of adsorption from dT40 (relative peak area of 44%) to dT100 (rel- ative peak area of 97%) The stainless-steel column from the same provider (Pss) showed similar behavior, but adsorption was slightly less pronounced, in particular for the smaller ONs of this sample (dT40 and dT60) Interpretation of these results is quite difficult since there is likely to be an interplay between several different factors (mostly related to the larger size of these ONs and their molecular masses ranging between 10 and 30 kDa) However, it

is also important to keep in mind that the columns were initially

H Lardeux, A Goyon, K Zhang et al Journal of Chromatography A 1677 (2022) 463324

Fig 3 RPLC bioinert columns composed of hybrid surfaces, PEEK- or titanium-lined hardware (Wpr, Ypk and Pti) in comparison with their stainless-steel analogs (Wss, Yss,

Pss), and HILIC bioinert columns composed of hybrid surfaces and PEEK-lined hardware (Wpr_H, Ypk_H) in comparison with their stainless-steel analogs (Wss_H, Yss_H).One injection of each mixture, namely dT15-35 (A, D), dT40-100 (B, E), and rU15-30 (C, F), was performed on each previously conditioned column using the Premier system Relative peak areas for each oligonucleotide are reported 100% corresponds to peak area using the Wpr and Wpr_H column for the IP-RPLC and HILIC mode, respectively

conditioned with a dT15–35 sample (see Section 3.1.1.) Since the

chemical nature of the dT40-100 sample is different from the ma-

terial used for conditioning, the column adsorption sites may not

have been perfectly masked and lead to partial adsorption of larger

ONs ( Fig.3B) This itself highlights an inherent drawback to having

to rely on sample-based column conditioning Fig 3C shows the

adsorption data experimentally obtained for small ORNs, namely

the rU15–30 sample In this case, the differences observed between

the RPLC columns were more pronounced than for the dT15-35

sample ( Fig.3A), but less than for the dT40-100 sample ( Fig.3B) It

is important to mention that two bioinert columns (Wpr and Ypk)

showed comparable behavior, with no issue related to adsorption

for the small RNA products The third bioinert column (Pti) offered

very good performance in terms of adsorption, with relative peak

area values ranging from 75 to 90%, and there was no problem

with peak shape for this class of molecules (see Fig S1) For the

three stainless-steel columns (Wss, Yss and Pss), undesired adsorp-

tion became systematically more pronounced versus their bioinert

counterparts, with an average increase of 20 to 40% As expected,

losses appeared to be increasingly less pronounced for the larger

ON species

Concerning the results obtained in HILIC mode ( Fig.3D–F), sam-

ple conditioning was not sufficient to permanently mitigate ad-

sorption For the dT15-35 sample ( Fig.3D), the relative peak areas

from the stainless-steel columns (Wss_H and Yss_H) varied from

20 to 90% and from 30 to 40% respectively That means that non-

specific adsorption was still significant on the HILIC stainless-steel

columns, even after a conditioning procedure (see previous sec-

tion) However, these two columns behave quite differently Indeed,

the Yss_H column offers consistent analyte recovery whatever the

ON size, while sample losses become more and more significant with increasing ON size when using the Wss_H

Some significant differences were also observed between the bioinert columns (HST or PEEK) and those made from stainless- steel On average, the HST material (Wpr_H) offered 20–30% better recovery than the PEEK-coated column (Ypk_H) that was found to result in almost stable relative peak areas whatever the ON length Despite some differences in terms of peak areas, peak shapes were excellent on the two bioinert columns for the dT15-35 sample, as illustrated in Fig S1

Similar results were obtained for the mixture of small ORNs (rU15-30, relative peak areas presented in Fig.3F) vs small ODNs ( Fig 3D), but the recoveries experimentally obtained were im- proved compared to what was observed for all the columns for the small ODNs ORN relative peak areas were measured to be 50–80% for the Wss_H, 50–60% for the Yss_H and 80–95% for the Ypk_H column Values for the small ODNs were 20–90% for the Wss_H, 30–40% for the Yss_H and 70–80% for the Ypk_H col- umn These results demonstrate that ORNs are less prone to ad- sorption than ODNs under HILIC conditions Besides some changes

in peak areas, it is important to notice that peak shapes were again excellent on the two bioinert columns (Fig S1) Finally, some larger ODNs (dT40–100) were also analyzed on the four HILIC columns, and these results are reported in Fig 3E Here, the performance of the columns ranked the same as with small ONs The Wpr_H always showed better results compared to the other ones in terms of adsorption (reference column, relative peak area values of 100%) followed by the other bioinert HILIC col-

umn (Ypk_H, values of 70–85%), the Waters stainless-steel col-

umn (Wss_H, values of 60–70%) and the YMC one (Yss_H, val-

ues of 40–50%) Non-specific adsorption did not vary according

to the length of oligonucleotides across any of the tested HILIC

columns At most, there was 10–15% variation of relative peak

area with a given column This behavior is in line with the pre-

viously obtained results, except on the Wss_H column where

the decrease of relative peak areas with oligonucleotide size was

not any longer observed with large oligonucleotides (dT40-100)

These results suggest that sample losses on the Wss_H column

are very dependent on the size of the oligonucleotide between

a 15- and 40-mer length, but that differences in size beyond 40

residues might have a diminishing effect As illustrated in Fig S1,

all peaks remained symmetrical on the two bioinert columns for

the mixture of large ODNs Broader peaks were observed on the

Wpr_H vs the Ypk_H column Nevertheless, selectivity and reso-

lution were always greater on the Wpr_H column, which might

in part be tied to its stationary phase and its corresponding

retentivity

In summary, it appears that the smaller DNA and RNA samples

(15- to 35-mer) behaved quite similarly in terms of non-specific

adsorption in both modes, with adsorption being generally more

pronounced under HILIC conditions The larger oligodeoxyribonu-

cleotides (40- to 100-mer) showed different behavior on the differ-

ent columns investigated in this work Most of the observed chro-

matographic differences are related to the size of these ON species,

which are clearly more difficult to characterize

3.2 Impact of instrumentation

The instrument could also be responsible for non-specific ad-

sorption and its impact might very well be different under HILIC

vs IP-RPLC conditions since mobile phase compositions are quite

different For this part, a previously conditioned bioinert Waters

column (Wpr or Wpr_H) was systematically employed to minimize

as much as possible non-specific adsorption within the column and

to thereby more sensitively investigate the impact of the UV flow

cell and UHPLC instrumentation

Indeed, in the last few years, all LC instrument manufactur-

ers have released several different chromatographic systems, which

have been referred to as bioinert, biocompatible and iron-free

[ 55, 61, 64] These systems have been designed to minimize non-

specific adsorption losses due to metal interactions and/or to offer

a better compatibility with mobile phases containing high amount

of salts Historically, bioinert HPLC systems were made of PEEK,

but with the emergence of UHPLC conditions, advanced materi-

als such as titanium and MP35N alloys have also been applied to

build instrumentation that can withstand elevated pressures In ad-

dition, a new bioinert UHPLC system was recently released by Wa-

ters where the flow path is covered with a hybrid surface tech-

nology that is created through the vapor deposition of ethylene

bridged hybrid inorganic/organic surfaces To have a clear view

of what can be done with the currently available UHPLC instru-

ments for the analysis of ON products, three different systems from

the same manufacturer were compared The first one was a regu-

lar stainless-steel UHPLC instrument (Waters Acquity H-Class) The

second one was a biocompatible UHPLC system (Waters Acquity H-

Class Bio), where the flow path is composed of corrosion resistant

MP35N alloy Finally, the last one was a new UHPLC system (Wa-

ters Acquity Premier) constructed with hybrid surface flow path

components

Besides the evaluation of three different UHPLC instruments,

two different UV flow cells were also tested The first one is in-

cluded on the commercial Waters Acquity H-Class and Waters Ac-

quity Premier instruments It is a regular light-guided analytical

flow cell with a path length of 10 mm and a volume of only 500

nL PEEK is used as the tubing material on the inside of the flow cell apparatus that is itself mostly made of Teflon The second one

is commonly used with the Waters Acquity H-Class Bio system to limit adsorption of biopharmaceutical products during their anal- ysis It is a titanium flow cell having a path length of only 5 mm and a volume of 1500 nL

Fig.4shows the results obtained for the dT15-35, dT40-100 and dT15-35 samples with each combination of UHPLC instrument and

UV flow cell material when analyzing the mixtures in HILIC mode Results concerning IP-RPLC mode have been instead reported in the Supplementary Information (Fig S2)

3.2.1 Impact of UV flow cell conditioning

Some preliminary HILIC experiments were performed with the

UV analytical flow cell mounted on a Waters Premier instrument before and after oligonucleotide sample conditioning Fig.5shows HILIC chromatograms of the small (dT15-35) and large (dT40–100) ODNs samples as obtained with the original and then the condi- tioned UV flow cell The experiments with the original UV flow cell were performed on a brand-new Premier instrument, as received from manufacturing This means that the UV flow cell had not seen any ON sample before the experiment reported in Fig.5(blue trace) On the other hand, the conditioned UV flow cell corre- sponds to a part that had been used for about one month and exposed to numerous injections of ONs This corresponds to the black trace in Fig.5

As shown, some clear differences were observed between the black and the blue traces and were seen upon zooming in on the baseline (bottom chromatograms in Fig 5) Differences were not drastic for small ODNs of around 15–20 nucleotides (nt), even though some minor species (probably corresponding to shortmers and longmers) were less resolved and/or hardly visible on the UV flow cell that was not conditioned (blue trace) Differences be- tween the UV flow cells were amplified for the larger ODNs of the dT15-35 sample corresponding to dT25 to dT35, where there was

a significant loss of resolution between minor species and a sig- nificant drift in baseline The situation was at its worst with the mixture of large ODNs (dT40-100) Here, weakly retained compo- nents in the sample were hardly detected when using the uncon- ditioned, original UV cell In addition, the four main peaks (dT40, dT60, dT80 and dT100) were poorly resolved and a strong base- line drift was observed On the contrary, the sample-conditioned

UV flow cell provided better signal intensity, improved peak sym- metry, higher resolution and less baseline drift These observa- tions clearly demonstrate the need to properly condition the UV flow cell before its first use This also proves that non-specific adsorption within the UV flow cell was most pronounced with large ONs

3.2.2 Impact of instrumentation material

The impact of UHPLC instrumentation on the non-specific ad- sorption of ONs was assessed using three different Waters chro- matographic systems The H-Class and Premier instrument are originally equipped with an analytical UV flow cell (mostly made

of Teflon wetted parts) that was already conditioned with the dT15-35 sample (see Section3.2.1.), while the H-Class Bio is origi- nally equipped with a titanium UV flow cell Therefore, the original instrument configurations were compared with the use of the ana- lytical flow cell for the H-Class Bio instrument Importantly, to have adsorption data that can be reliably compared between the three instruments, the same UV detector was used on the three instru- ments Differences in sensitivity due to the detector itself and in particular the UV lamp were therefore avoided

HILIC experimental results have been summarized in Fig.4for the three model mixtures of ONs ON relative peak areas were plot- ted for the four different systems, and the instrument configura-

H Lardeux, A Goyon, K Zhang et al Journal of Chromatography A 1677 (2022) 463324

Fig 4 Comparison of instruments namely H-Class, H-Class Bio and Premier equipped with the same UV detector with an analytical flow cell, except for the H-Class Bio

where the use of a titanium flow cell was also discussed Letter M indicates that the configuration is the one that is commercially available Histograms corresponding to HILIC-UV chromatograms of the three mixtures using the Waters bioinert column (Wpr_H)

Fig 5 HILIC-UV chromatograms of DNA mixtures of oligonucleotides showing the impact of sample conditioning of the UV analytical flow cell on oligonucleotide adsorption

a The UV analytical flow cell was used for the first time b The same analytical flow cell was used after being sample conditioned A previously conditioned Wpr_H column was used on the Premier system

tion combining the Premier system with the conditioned analytical

flow cell was taken as the reference (100%) Regardless of the UH-

PLC instrument, the peak areas obtained with the titanium UV flow

cell were about 2-fold lower than with the analytical UV flow cell

(as a result of its 5 vs 10 mm path length) Since the analytical UV

flow cell was already conditioned, no significant differences were

observed between ONs varying in size and type As illustrated in

Fig.4, there were almost no differences for large ODNs and small

ORNs between the three UHPLC instruments equipped with the

analytical flow cell (relative peak areas values ranged from 95 to

105%)

However, some slight differences were observed between the

Premier system and the two remaining UHPLC instruments when

analyzing small ODNs (dT15-35) In this particular case, relative

peak areas on the two other instruments were equal to 85-90%

vs 100% on the Premier system The Premier system having be-

ing used for the column conditioning studies, it therefore saw sig-

nificantly higher quantities of samples and in particular the dT15-

35 sample This might explain the slightly improved recoveries of

short ODNs with as much likelihood as the hybrid surfaces of the

instrument flow path inherently contributing such a sizable effect

In the end, it appears that the non-specific adsorption of oligonu-

cleotides on any type of UHPLC instrument can be negligible, at

least with the mass loads applied here for the sake of sample char-

acterization In this type of work, care should at least be taken to

condition the UV flow cell

The impact of the instrumentation on non-specific adsorption

of ONs was also evaluated in IP-RPLC mode Fig S2 shows the re-

sults obtained for the dT15-35, dT40-100, and rU15-30 samples

with each combination of UHPLC instruments and UV flow cell

material (3 systems and 2 UV flow cells) As in HILIC mode, the

two-fold decrease of UV signal observed for each ON when mod-

ifying the analytical UV flow cell for a titanium flow cell can be

attributed to the shorter path length (5 mm vs 10 mm) When

looking at chromatograms of 15- to 35-mer ONs (Fig S2A and S2C),

peak shapes remain strictly identical whatever the type of instru-

ment and UV flow cell material employed meaning that no signif-

icant adsorption issues were observed, even when using the Wa-

ters Acquity H-Class system which is composed of stainless-steel

Fig S2B shows the results obtained for the mixture of larger ODNs

(dT40-100 sample) Here, the instrument had a clear impact on ad-

sorption and above all peak shapes of the largest ONs Interest-

ingly, the peak shapes of the largest ONs (60- to 100-mer) were

strongly degraded with severe tailing and broadening observed

when using the regular vs titanium UV flow cell To understand

this behavior, it is relevant to mention that there are some signif-

icant differences between the two UV flow cells in terms of their

surface exposed materials Indeed, the analytical flow cell is mostly

composed of Teflon which is a hydrophobic material where the

complexes of ON and TEA (which are also hydrophobic) can adsorb,

despite their short residence time, and that can lead to peak shape

distortion

Besides the UV flow cell, some additional (more limited) differ-

ences were also observed for dT80 and dT100 between the differ-

ent UHPLC systems, especially when using the analytical flow cell

Indeed, the H-Class system offered worse performance (asymmetry

at 10% for dT100 was 4.39) vs the H-Class Bio (asymmetry at 10%

was 1.85) or the Premier instrument (asymmetry at 10% was 2.18)

This confirms that large ONs (60- to 100-mer) are more prone to

adsorption and that the latter two UHPLC systems should be pref-

erentially used

Based on these findings, it is clear that the analytical UV flow

cell of 10 mm is suitable for use with small ONs (15- to 40-mer) in

order to achieve maximum sensitivity (namely a 2-fold improve-

ment) On the contrary, despite the more limited sensitivity, the

titanium flow cell, even with its 5 mm path length, should be

preferably used in IP-RPLC mode to achieve suitable peak shapes for large ONs (60- to 100-mer)

3.3 Application to the analysis of unmodified and modified 20-mer oligonucleotides

Differences in the general features of the oligonucleotide have shown to impact non-specific adsorption, and therefore, chemical modifications of the oligonucleotide structure may influence their recovery

Indeed, it should be noted that therapeutic ONs have to be chemically-modified to ensure proper pharmacokinetic properties and sufficient activity in vivo [65] In this context, modifications often involve the phosphate linkage and the furanose sugar moi- ety (deoxyribose in DNA and ribose in RNA) Among the modifi- cations involving the phosphodiester backbone, the most widely used is a phosphorothioate (PS) bond, in which a sulfur replaces one of the non-bridging oxygen atoms of the phosphate linkage (reported in blue in Fig 6) [66] In addition, modifications ap- plied to the furanose sugar moiety (reported in red in Fig.6) in- clude substitutions in the 2’-position, with 2’-O-methyl (OMe), 2’- O-methoxyethyl (MOE), and locked nucleic acid (LNA) [66] Based

on the performed chemical modifications and base composition of the ONs (U/T, C, A, G, reported in grey in Fig.6), a change in the general features of the ON occurs and potentially impacts non- specific adsorption

The influence of chemical modifications on the 20-mer ON ad- sorption in both IP-RPLC and HILIC modes was therefore investi- gated and the list of the evaluated ONs is reported in Table1

As a result of previous findings, the bioinert hybrid surface columns (Wpr and Wpr_H) and their stainless-steel counterparts (Wss and Wss_H) were used in combination with the best LC sys- tem configuration (consisting of the Premier instrument equipped with an analytical flow cell) to evaluate the impact of ON chem- istry on adsorption Corresponding chromatograms are reported in Figs.7and8 The main chromatographic descriptors and percent- age recoveries (calculated as the ratio between the stainless-steel reference column and the bioinert column areas) are summarized

in Table3 First of all, and as reported in Table 3, a higher recovery

of all the ONs was obtained with the bioinert HST columns (Wpr/Wpr_H) as compared to their stainless-steel counterparts (Wss/Wss_H) It is worth mentioning that the electropositive metal oxide layer on the surface of stainless-steel columns is generally thought to result in ionic interactions with the negatively-charged backbone of the ONs and therefore be the cause of non-specific ad- sorption [ 62, 67] This metal oxide is masked in the case of the hy- brid surface (Wpr/Wpr_H) columns Indeed, this hypothesis is sup- ported by the chromatograms of all ONs shown in Figs 7and 8, that correspond to the first and second set of 20-mer ON samples, respectively As reported in Table3, a better recovery can be ob- served, indicating that ionic interactions between column surfaces and the ONs are attenuated with the bioinert columns

Concerning the detailed impact of modifications, differences

in the ON sugar (deoxyribose/ribose) and base composition (U/T,

C, A, G) were first evaluated The IP-RPLC-UV and HILIC-UV chromatograms of this set of samples were reported in Fig 7A and Fig 7B, respectively For this part of the work, 20-mer homomolecular oligodeoxyribonucleotides (ODNs) were consid- ered, namely an oligodeoxyadenosine (dA20), an oligodeoxycyti- dine (dC20), and an oligodeoxyguanosine (dG20) to be compared with an oligodeoxythymidine (dT20) A change on the sugar com- position was also applied and a 20-mer oligoribonucleotide (ORN), namely oligouridine (rU20), was analyzed By fixing the length of the ONs at 20-mer, and therefore the number of the ON phos- phate groups, it was possible to examine the extent of adsorption